trevon/yams
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 2
yams/tests/unit/daemon/daemon_metrics_status_test.cpp

7325 lines
307 KiB
C++
Raw Permalink Blame History

// Daemon metrics and status test suite (Catch2)
// Consolidates status/metrics tests: embedding status, plugin degradation, WAL metrics, FSM states
// Covers: DaemonMetrics, StatusResponse serialization, plugin degradation, WAL metrics
// pi-lens-ignore: fatal error
#include <catch2/catch_test_macros.hpp>
#include <catch2/generators/catch_generators.hpp>
#include <algorithm>
#include <array>
#include <chrono>
#include <cstdlib>
#include <filesystem>
#include <fstream>
#include <future>
#include <mutex>
#include <random>
#include <span>
#include <stdexcept>
#include <thread>
#include <unordered_map>
#include <unordered_set>
#include "../../common/metadata_test_db.h"
#include "../../common/test_helpers_catch2.h"
#include <spdlog/spdlog.h>
#include <yams/app/services/graph_query_service.hpp>
#include <yams/app/services/session_service.hpp>
#include <yams/crypto/hasher.h>
#include <yams/daemon/components/DaemonLifecycleFsm.h>
#include <yams/daemon/components/DaemonMetrics.h>
#include <yams/daemon/components/DatabaseManager.h>
#include <yams/daemon/components/RequestDispatcher.h>
#include <yams/daemon/components/ResourceGovernor.h>
#include <yams/daemon/components/ServiceManager.h>
#include <yams/daemon/components/StateComponent.h>
#include <yams/daemon/components/WorkCoordinator.h>
#include <yams/daemon/daemon.h>
#include <yams/daemon/daemon_lifecycle.h>
#include <yams/daemon/ipc/fsm_metrics_registry.h>
#include <yams/daemon/ipc/ipc_protocol.h>
#include <yams/daemon/ipc/proto_serializer.h>
#include <yams/daemon/ipc/request_context.h>
#include <yams/daemon/ipc/request_context_registry.h>
#include <yams/daemon/metric_keys.h>
#include <yams/daemon/resource/external_plugin_host.h>
#include <yams/daemon/resource/model_provider.h>
#include <yams/daemon/resource/plugin_host.h>
#include <yams/metadata/knowledge_graph_store.h>
#include <yams/metadata/metadata_repository.h>
#include <yams/metadata/path_utils.h>
#include <yams/vector/vector_database.h>
#include <yams/version.hpp>
#include <boost/asio/co_spawn.hpp>
#include <boost/asio/io_context.hpp>
#include <boost/asio/use_future.hpp>
using namespace yams;
using namespace yams::daemon;
// =============================================================================
// Test Helpers
// =============================================================================
namespace {
class StubLifecycle : public IDaemonLifecycle {
public:
explicit StubLifecycle(LifecycleState state = LifecycleState::Ready) {
snapshot_.state = state;
}
LifecycleSnapshot getLifecycleSnapshot() const override {
++getLifecycleSnapshotCalls_;
if (throwOnGetLifecycleSnapshotCall_ > 0 &&
getLifecycleSnapshotCalls_ == throwOnGetLifecycleSnapshotCall_) {
throw std::runtime_error("getLifecycleSnapshot boom");
}
if (throwOnGetLifecycleSnapshotCount_ > 0) {
--throwOnGetLifecycleSnapshotCount_;
throw std::runtime_error("getLifecycleSnapshot boom");
}
return snapshot_;
}
void setSubsystemDegraded(const std::string& subsystem, bool degraded,
const std::string& reason) override {
if (throwOnSetSubsystemDegraded_) {
throw std::runtime_error("setSubsystemDegraded boom");
}
lastSubsystem_ = subsystem;
lastDegraded_ = degraded;
lastReason_ = reason;
setSubsystemDegradedCalls_++;
}
void onDocumentRemoved(const std::string& hash) override { removedHashes_.push_back(hash); }
void requestShutdown(bool, bool) override {}
const std::string& lastSubsystem() const { return lastSubsystem_; }
bool lastDegraded() const { return lastDegraded_; }
const std::string& lastReason() const { return lastReason_; }
int setSubsystemDegradedCalls() const { return setSubsystemDegradedCalls_; }
void resetDegradedState() {
lastSubsystem_.clear();
lastDegraded_ = false;
lastReason_.clear();
setSubsystemDegradedCalls_ = 0;
}
void setThrowOnSetSubsystemDegraded(bool enabled) { throwOnSetSubsystemDegraded_ = enabled; }
void setThrowOnGetLifecycleSnapshot(bool enabled) {
throwOnGetLifecycleSnapshotCount_ = enabled ? 1 : 0;
}
void setThrowOnGetLifecycleSnapshotCount(int count) {
throwOnGetLifecycleSnapshotCount_ = count;
}
void setThrowOnGetLifecycleSnapshotCall(int call) { throwOnGetLifecycleSnapshotCall_ = call; }
const std::vector<std::string>& removedHashes() const { return removedHashes_; }
private:
LifecycleSnapshot snapshot_{};
std::string lastSubsystem_;
bool lastDegraded_{false};
std::string lastReason_;
int setSubsystemDegradedCalls_{0};
bool throwOnSetSubsystemDegraded_{false};
mutable int getLifecycleSnapshotCalls_{0};
mutable int throwOnGetLifecycleSnapshotCount_{0};
int throwOnGetLifecycleSnapshotCall_{0};
std::vector<std::string> removedHashes_;
};
// Minimal stub provider for embedding tests
class StubModelProvider : public IModelProvider {
public:
explicit StubModelProvider(size_t dim, std::string path)
: dim_(dim), path_(std::move(path)), available_(true) {}
Result<std::vector<float>> generateEmbedding(const std::string& text) override {
lastTextArg_ = text;
if (auto it = singleTextErrors_.find(text); it != singleTextErrors_.end()) {
return it->second;
}
if (generateEmbeddingError_) {
return *generateEmbeddingError_;
}
if (!singleEmbeddingResult_.empty()) {
return singleEmbeddingResult_;
}
return std::vector<float>(dim_, 0.25f);
}
Result<std::vector<std::vector<float>>>
generateBatchEmbeddings(const std::vector<std::string>& texts) override {
lastBatchSizeArg_ = texts.size();
if (!texts.empty()) {
lastTextArg_ = texts.front();
}
if (generateBatchError_ && !isWarmupBatch(texts)) {
return *generateBatchError_;
}
if (!batchEmbeddingResult_.empty()) {
return batchEmbeddingResult_;
}
return std::vector<std::vector<float>>(lastBatchSizeArg_, std::vector<float>(dim_, 0.5f));
}
Result<std::vector<float>> generateEmbeddingFor(const std::string&,
const std::string& text) override {
lastTextArg_ = text;
return generateEmbedding(text);
}
Result<std::vector<std::vector<float>>>
generateBatchEmbeddingsFor(const std::string&, const std::vector<std::string>& texts) override {
lastBatchSizeArg_ = texts.size();
if (!texts.empty()) {
lastTextArg_ = texts.front();
}
return generateBatchEmbeddings(texts);
}
Result<void> loadModel(const std::string& modelName) override {
if (loadError_) {
return *loadError_;
}
lastOptionsModel_.clear();
lastOptionsJson_.clear();
loaded_.push_back(modelName);
return Result<void>();
}
Result<void> loadModelWithOptions(const std::string& modelName,
const std::string& optionsJson) override {
if (loadError_) {
return *loadError_;
}
lastOptionsModel_ = modelName;
lastOptionsJson_ = optionsJson;
loaded_.push_back(modelName);
return Result<void>();
}
Result<void> unloadModel(const std::string& modelName) override {
auto it = std::find(loaded_.begin(), loaded_.end(), modelName);
if (it == loaded_.end()) {
return ErrorCode::NotFound;
}
loaded_.erase(it);
return Result<void>();
}
bool isModelLoaded(const std::string& modelName) const override {
if (throwOnIsModelLoadedCount_ > 0) {
--throwOnIsModelLoadedCount_;
throw std::runtime_error("isModelLoaded boom");
}
return std::find(loaded_.begin(), loaded_.end(), modelName) != loaded_.end();
}
std::vector<std::string> getLoadedModels() const override { return loaded_; }
size_t getLoadedModelCount() const override { return loaded_.size(); }
Result<ModelInfo> getModelInfo(const std::string& modelName) const override {
if (failModelInfo_) {
return ErrorCode::NotFound;
}
if (!isModelLoaded(modelName))
return ErrorCode::NotFound;
ModelInfo mi;
mi.name = modelName;
mi.path = path_;
mi.embeddingDim = dim_;
return mi;
}
size_t getEmbeddingDim(const std::string&) const override { return dim_; }
std::shared_ptr<vector::EmbeddingGenerator>
getEmbeddingGenerator(const std::string& = "") override {
return nullptr;
}
std::string getProviderName() const override { return "StubProvider"; }
std::string getProviderVersion() const override { return "vtest"; }
bool isAvailable() const override { return available_; }
size_t getMemoryUsage() const override {
if (throwOnMemoryUsage_) {
throw std::runtime_error("memory boom");
}
return 0;
}
void releaseUnusedResources() override {}
void shutdown() override {}
void setAvailable(bool available) { available_ = available; }
const std::string& lastOptionsModel() const { return lastOptionsModel_; }
const std::string& lastOptionsJson() const { return lastOptionsJson_; }
void setLoadError(Error error) { loadError_ = std::move(error); }
void clearLoadError() { loadError_.reset(); }
void setThrowOnMemoryUsage(bool enabled) { throwOnMemoryUsage_ = enabled; }
void setFailModelInfo(bool enabled) { failModelInfo_ = enabled; }
void setThrowOnIsModelLoaded(bool enabled) { throwOnIsModelLoadedCount_ = enabled ? 1 : 0; }
void setThrowOnIsModelLoadedCount(int count) { throwOnIsModelLoadedCount_ = count; }
void setSingleEmbeddingResult(std::vector<float> embedding) {
singleEmbeddingResult_ = std::move(embedding);
}
void setBatchEmbeddingResult(std::vector<std::vector<float>> embeddings) {
batchEmbeddingResult_ = std::move(embeddings);
}
void setGenerateEmbeddingError(Error error) { generateEmbeddingError_ = std::move(error); }
void setGenerateBatchError(Error error) { generateBatchError_ = std::move(error); }
void setSingleTextError(const std::string& text, Error error) {
singleTextErrors_[text] = std::move(error);
}
void clearSingleTextErrors() { singleTextErrors_.clear(); }
private:
static bool isWarmupBatch(const std::vector<std::string>& texts) {
static const std::array<std::string, 4> kWarmupTexts = {
"warmup text one", "warmup text two", "warmup text three", "warmup text four"};
return texts.size() == kWarmupTexts.size() &&
std::equal(texts.begin(), texts.end(), kWarmupTexts.begin());
}
size_t dim_;
std::string path_;
bool available_;
std::vector<std::string> loaded_;
std::string lastOptionsModel_;
std::string lastOptionsJson_;
std::optional<Error> loadError_;
std::optional<Error> generateEmbeddingError_;
std::optional<Error> generateBatchError_;
bool throwOnMemoryUsage_{false};
bool failModelInfo_{false};
mutable int throwOnIsModelLoadedCount_{0};
std::vector<float> singleEmbeddingResult_;
std::vector<std::vector<float>> batchEmbeddingResult_;
std::unordered_map<std::string, Error> singleTextErrors_;
mutable std::string lastTextArg_;
mutable std::size_t lastBatchSizeArg_{0};
};
using yams::test::ScopedEnvVar;
class TuneAdvisorListGuard {
public:
TuneAdvisorListGuard(uint32_t inflightLimit, uint32_t admissionWaitMs)
: originalInflightLimit_(TuneAdvisor::listInflightLimit()),
originalAdmissionWaitMs_(TuneAdvisor::listAdmissionWaitMs()) {
TuneAdvisor::setListInflightLimit(inflightLimit);
TuneAdvisor::setListAdmissionWaitMs(admissionWaitMs);
}
~TuneAdvisorListGuard() {
TuneAdvisor::setListInflightLimit(originalInflightLimit_);
TuneAdvisor::setListAdmissionWaitMs(originalAdmissionWaitMs_);
}
private:
uint32_t originalInflightLimit_;
uint32_t originalAdmissionWaitMs_;
};
template <typename Fn> class ScopeExit {
public:
explicit ScopeExit(Fn fn) : fn_(std::move(fn)) {}
~ScopeExit() { fn_(); }
private:
Fn fn_;
};
std::shared_ptr<metadata::ConnectionPool> getPruneTestPool() {
static std::shared_ptr<metadata::ConnectionPool> pool = [] {
metadata::ConnectionPoolConfig cfg{};
cfg.enableWAL = false;
cfg.enableForeignKeys = false;
auto p = std::make_shared<metadata::ConnectionPool>(":memory:", cfg);
(void)p->initialize();
return p;
}();
return pool;
}
std::filesystem::path makeTempMetadataDbPath(const std::string& prefix) {
return yams::test::migrated_metadata_db_template().clone(prefix);
}
std::shared_ptr<metadata::MetadataRepository>
makeReadyMetadataRepo(const std::shared_ptr<metadata::ConnectionPool>& pool) {
return std::make_shared<metadata::MetadataRepository>(
*pool, nullptr, metadata::MetadataRepository::SchemaBootstrapMode::AssumeReady);
}
class StubPruneMetadataRepository : public metadata::MetadataRepository {
public:
StubPruneMetadataRepository() : metadata::MetadataRepository(*getPruneTestPool()) {}
void addDocument(const metadata::DocumentInfo& doc) {
std::lock_guard<std::mutex> lk(mu_);
docs_.push_back(doc);
docsByHash_[doc.sha256Hash] = doc;
docsById_[doc.id] = doc;
}
void setQueryError(Error error) { queryError_ = std::move(error); }
void setThrowOnQuery(std::string message) { throwOnQuery_ = std::move(message); }
void setMissingHash(const std::string& hash) { missingHashes_.insert(hash); }
void setThrowOnLookupHash(const std::string& hash, std::string message) {
throwOnLookup_[hash] = std::move(message);
}
void setDeleteError(int64_t id, Error error) { deleteErrors_[id] = std::move(error); }
void setSnapshots(std::vector<std::string> snapshots) { snapshots_ = std::move(snapshots); }
void setSnapshotError(Error error) { snapshotError_ = std::move(error); }
void setSnapshotDocuments(std::vector<metadata::DocumentInfo> docs) {
snapshotDocs_ = std::move(docs);
}
void setSnapshotDocumentsError(Error error) { snapshotDocsError_ = std::move(error); }
void setExactPathDocument(std::optional<metadata::DocumentInfo> doc) {
std::lock_guard<std::mutex> lk(mu_);
exactPathDocument_ = std::move(doc);
}
void setExactPathError(Error error) {
std::lock_guard<std::mutex> lk(mu_);
exactPathError_ = std::move(error);
}
void setAllMetadata(int64_t id,
std::unordered_map<std::string, metadata::MetadataValue> values) {
std::lock_guard<std::mutex> lk(mu_);
metadataById_[id] = std::move(values);
}
void setAllMetadataError(int64_t id, Error error) {
std::lock_guard<std::mutex> lk(mu_);
metadataByIdErrors_[id] = std::move(error);
}
void setPathTreeDocuments(std::vector<metadata::DocumentInfo> docs) {
std::lock_guard<std::mutex> lk(mu_);
pathTreeDocs_ = std::move(docs);
}
void setMetadataValueCounts(
std::unordered_map<std::string, std::vector<metadata::MetadataValueCount>> counts) {
metadataValueCounts_ = std::move(counts);
}
void setMetadataValueCountsError(Error error) { metadataValueCountsError_ = std::move(error); }
std::size_t deleteCallCount() const {
std::lock_guard<std::mutex> lk(mu_);
return deleteCallCount_;
}
std::vector<int64_t> deletedIds() const {
std::lock_guard<std::mutex> lk(mu_);
return deletedIds_;
}
Result<std::vector<metadata::DocumentInfo>>
queryDocuments(const metadata::DocumentQueryOptions&) override {
std::lock_guard<std::mutex> lk(mu_);
if (throwOnQuery_) {
throw std::runtime_error(*throwOnQuery_);
}
if (queryError_) {
return *queryError_;
}
return docs_;
}
Result<std::vector<metadata::DocumentInfo>>
findDocumentsBySnapshot(const std::string&) override {
std::lock_guard<std::mutex> lk(mu_);
if (snapshotDocsError_) {
return *snapshotDocsError_;
}
return snapshotDocs_;
}
Result<std::vector<std::string>> getSnapshots() override {
std::lock_guard<std::mutex> lk(mu_);
if (snapshotError_) {
return *snapshotError_;
}
return snapshots_;
}
Result<std::unordered_map<std::string, std::vector<metadata::MetadataValueCount>>>
getMetadataValueCounts(const std::vector<std::string>&,
const metadata::DocumentQueryOptions&) override {
std::lock_guard<std::mutex> lk(mu_);
if (metadataValueCountsError_) {
return *metadataValueCountsError_;
}
return metadataValueCounts_;
}
Result<std::optional<metadata::DocumentInfo>>
getDocumentByHash(const std::string& hash) override {
std::lock_guard<std::mutex> lk(mu_);
if (auto it = throwOnLookup_.find(hash); it != throwOnLookup_.end()) {
throw std::runtime_error(it->second);
}
if (missingHashes_.count(hash) > 0) {
return std::optional<metadata::DocumentInfo>(std::nullopt);
}
if (auto it = docsByHash_.find(hash); it != docsByHash_.end()) {
return std::optional<metadata::DocumentInfo>(it->second);
}
return std::optional<metadata::DocumentInfo>(std::nullopt);
}
Result<std::unordered_map<std::string, metadata::MetadataValue>>
getAllMetadata(int64_t documentId) override {
std::lock_guard<std::mutex> lk(mu_);
if (auto it = metadataByIdErrors_.find(documentId); it != metadataByIdErrors_.end()) {
return it->second;
}
if (auto it = metadataById_.find(documentId); it != metadataById_.end()) {
return it->second;
}
return std::unordered_map<std::string, metadata::MetadataValue>{};
}
Result<std::optional<metadata::DocumentInfo>>
findDocumentByExactPath(const std::string&) override {
std::lock_guard<std::mutex> lk(mu_);
if (exactPathError_) {
return *exactPathError_;
}
if (exactPathDocument_.has_value()) {
return exactPathDocument_;
}
return std::optional<metadata::DocumentInfo>(std::nullopt);
}
Result<std::vector<metadata::DocumentInfo>>
findDocumentsByPathTreePrefix(std::string_view, bool = true, int limit = 0) override {
std::lock_guard<std::mutex> lk(mu_);
if (limit > 0 && static_cast<std::size_t>(limit) < pathTreeDocs_.size()) {
return std::vector<metadata::DocumentInfo>(pathTreeDocs_.begin(),
pathTreeDocs_.begin() + limit);
}
return pathTreeDocs_;
}
Result<void> deleteDocument(int64_t id) override {
std::lock_guard<std::mutex> lk(mu_);
deleteCallCount_++;
if (auto it = deleteErrors_.find(id); it != deleteErrors_.end()) {
return it->second;
}
deletedIds_.push_back(id);
if (auto it = docsById_.find(id); it != docsById_.end()) {
docsByHash_.erase(it->second.sha256Hash);
docsById_.erase(it);
}
docs_.erase(
std::remove_if(docs_.begin(), docs_.end(),
[id](const metadata::DocumentInfo& doc) { return doc.id == id; }),
docs_.end());
return Result<void>();
}
private:
mutable std::mutex mu_;
std::vector<metadata::DocumentInfo> docs_;
std::unordered_map<std::string, metadata::DocumentInfo> docsByHash_;
std::unordered_map<int64_t, metadata::DocumentInfo> docsById_;
std::optional<Error> queryError_;
std::optional<Error> snapshotError_;
std::optional<Error> snapshotDocsError_;
std::optional<Error> metadataValueCountsError_;
std::optional<std::string> throwOnQuery_;
std::unordered_set<std::string> missingHashes_;
std::unordered_map<std::string, std::string> throwOnLookup_;
std::unordered_map<int64_t, Error> deleteErrors_;
std::optional<metadata::DocumentInfo> exactPathDocument_;
std::optional<Error> exactPathError_;
std::unordered_map<int64_t, std::unordered_map<std::string, metadata::MetadataValue>>
metadataById_;
std::unordered_map<int64_t, Error> metadataByIdErrors_;
std::vector<metadata::DocumentInfo> pathTreeDocs_;
std::vector<std::string> snapshots_;
std::vector<metadata::DocumentInfo> snapshotDocs_;
std::unordered_map<std::string, std::vector<metadata::MetadataValueCount>> metadataValueCounts_;
std::size_t deleteCallCount_{0};
std::vector<int64_t> deletedIds_;
};
class StubContentStore : public api::IContentStore {
public:
void setBlob(const std::string& hash, const std::string& content) {
blobs_[hash] = std::vector<std::byte>(reinterpret_cast<const std::byte*>(content.data()),
reinterpret_cast<const std::byte*>(content.data()) +
content.size());
}
void setRemoveResult(const std::string& hash, Result<bool> result) {
removeResults_[hash] = std::move(result);
}
bool hasBlob(const std::string& hash) const { return blobs_.contains(hash); }
std::vector<std::string> removedHashes() const { return removedHashes_; }
Result<api::StoreResult> storeBytes(std::span<const std::byte> data,
const api::ContentMetadata&) override {
std::string hash = "stub-hash-" + std::to_string(blobs_.size() + 1);
blobs_[hash] = std::vector<std::byte>(data.begin(), data.end());
api::StoreResult result;
result.contentHash = hash;
result.bytesStored = blobs_[hash].size();
return result;
}
Result<std::vector<std::byte>> retrieveBytes(const std::string& hash) override {
auto it = blobs_.find(hash);
if (it == blobs_.end()) {
return Error{ErrorCode::NotFound, "content not found"};
}
return it->second;
}
Result<std::vector<std::byte>> retrieveBytesPrefix(const std::string& hash,
std::size_t maxBytes) override {
auto bytes = retrieveBytes(hash);
if (!bytes) {
return bytes.error();
}
const auto prefixSize = std::min(maxBytes, bytes.value().size());
bytes.value().resize(prefixSize);
return bytes;
}
Result<api::IContentStore::RawContent> retrieveRaw(const std::string& hash) override {
auto bytes = retrieveBytes(hash);
if (!bytes) {
return bytes.error();
}
api::IContentStore::RawContent raw;
raw.data = std::move(bytes.value());
return raw;
}
std::future<Result<api::IContentStore::RawContent>>
retrieveRawAsync(const std::string& hash) override {
return std::async(std::launch::deferred, [this, hash]() { return retrieveRaw(hash); });
}
Result<api::StoreResult> store(const std::filesystem::path& path, const api::ContentMetadata&,
api::ProgressCallback) override {
std::ifstream ifs(path, std::ios::binary);
if (!ifs.good()) {
return Error{ErrorCode::FileNotFound, "stub file not found"};
}
std::vector<std::byte> bytes;
char ch = 0;
while (ifs.get(ch)) {
bytes.push_back(static_cast<std::byte>(ch));
}
if (!ifs.eof()) {
return Error{ErrorCode::IOError, "stub read failed"};
}
return storeBytes(bytes, {});
}
Result<api::RetrieveResult> retrieve(const std::string&, const std::filesystem::path&,
api::ProgressCallback) override {
return ErrorCode::NotImplemented;
}
Result<api::StoreResult> storeStream(std::istream&, const api::ContentMetadata&,
api::ProgressCallback) override {
return ErrorCode::NotImplemented;
}
Result<api::RetrieveResult> retrieveStream(const std::string&, std::ostream&,
api::ProgressCallback) override {
return ErrorCode::NotImplemented;
}
Result<bool> exists(const std::string&) const override { return ErrorCode::NotImplemented; }
Result<bool> remove(const std::string& hash) override {
if (auto it = removeResults_.find(hash); it != removeResults_.end()) {
return it->second;
}
removedHashes_.push_back(hash);
blobs_.erase(hash);
return true;
}
Result<api::ContentMetadata> getMetadata(const std::string&) const override {
return ErrorCode::NotImplemented;
}
Result<void> updateMetadata(const std::string&, const api::ContentMetadata&) override {
return ErrorCode::NotImplemented;
}
std::vector<Result<api::StoreResult>>
storeBatch(const std::vector<std::filesystem::path>&,
const std::vector<api::ContentMetadata>&) override {
return {};
}
std::vector<Result<bool>> removeBatch(const std::vector<std::string>&) override { return {}; }
api::ContentStoreStats getStats() const override { return {}; }
api::HealthStatus checkHealth() const override { return {}; }
Result<void> verify(api::ProgressCallback) override { return ErrorCode::NotImplemented; }
Result<void> compact(api::ProgressCallback) override { return ErrorCode::NotImplemented; }
Result<void> garbageCollect(api::ProgressCallback) override {
return ErrorCode::NotImplemented;
}
private:
std::unordered_map<std::string, std::vector<std::byte>> blobs_;
std::unordered_map<std::string, Result<bool>> removeResults_;
std::vector<std::string> removedHashes_;
};
class StubGraphQueryService : public app::services::IGraphQueryService {
public:
void addConnectedDocument(std::string hash, std::string relation, int distance = 1,
double relevance = 0.0) {
app::services::GraphConnectedNode node{};
node.nodeMetadata.documentHash = std::move(hash);
node.distance = distance;
node.relevanceScore = relevance;
app::services::GraphEdgeDescriptor edge{};
edge.relation = std::move(relation);
node.connectingEdges.push_back(std::move(edge));
response_.allConnectedNodes.push_back(std::move(node));
}
Result<app::services::GraphQueryResponse>
query(const app::services::GraphQueryRequest&) override {
return response_;
}
Result<app::services::ListSnapshotsResponse>
listSnapshots(const app::services::ListSnapshotsRequest&) override {
return app::services::ListSnapshotsResponse{};
}
Result<app::services::PathHistoryResponse>
getPathHistory(const app::services::PathHistoryRequest&) override {
return app::services::PathHistoryResponse{};
}
Result<std::optional<std::int64_t>> resolveToNodeId(const std::string&) override {
return std::optional<std::int64_t>(std::nullopt);
}
private:
app::services::GraphQueryResponse response_{};
};
std::filesystem::path makeTempDir(const std::string& prefix);
metadata::DocumentInfo makeGraphDispatcherDocument(const std::filesystem::path& path,
const std::string& hash) {
metadata::DocumentInfo info;
info.filePath = path.string();
info.fileName = path.filename().string();
info.fileExtension = path.extension().string();
info.fileSize = static_cast<std::int64_t>(std::filesystem::file_size(path));
info.sha256Hash = hash;
info.mimeType = "text/plain";
const auto now = std::chrono::floor<std::chrono::seconds>(std::chrono::system_clock::now());
info.createdTime = now;
info.modifiedTime = now;
info.indexedTime = now;
info.contentExtracted = true;
info.extractionStatus = metadata::ExtractionStatus::Success;
metadata::populatePathDerivedFields(info);
return info;
}
struct GraphDispatcherFixture {
GraphDispatcherFixture() {
static const bool kSuppressInfoLogs = [] {
spdlog::set_level(spdlog::level::warn);
return true;
}();
(void)kSuppressInfoLogs;
testDir = makeTempDir("yams_graph_dispatcher_");
dbPath = makeTempMetadataDbPath("yams_graph_dispatcher_db_");
cfg.dataDir = testDir;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
svc = std::make_unique<ServiceManager>(cfg, state, lifecycleFsm);
dispatcher = std::make_unique<RequestDispatcher>(&lifecycle, svc.get(), &state);
}
~GraphDispatcherFixture() {
dispatcher.reset();
svc.reset();
if (writeWork) {
writeWork.reset();
}
writeIo.stop();
if (writeThread.joinable()) {
writeThread.join();
}
kgStore.reset();
repo.reset();
pool.reset();
std::error_code ec;
std::filesystem::remove_all(testDir, ec);
yams::test::remove_sqlite_artifacts(dbPath);
}
void initMetadata(bool withKg = true) {
metadata::ConnectionPoolConfig poolConfig{};
poolConfig.minConnections = 1;
poolConfig.maxConnections = 2;
pool = std::make_shared<metadata::ConnectionPool>(dbPath.string(), poolConfig);
REQUIRE(pool->initialize().has_value());
repo = makeReadyMetadataRepo(pool);
svc->__test_setMetadataRepo(repo);
if (withKg) {
metadata::KnowledgeGraphStoreConfig kgConfig{};
kgConfig.enable_alias_fts = true;
kgConfig.enable_wal = false;
auto kgResult = metadata::makeSqliteKnowledgeGraphStore(*pool, kgConfig);
REQUIRE(kgResult.has_value());
kgStore = std::shared_ptr<metadata::KnowledgeGraphStore>(std::move(kgResult).value());
repo->setKnowledgeGraphStore(kgStore);
writeIo.restart();
writeWork = std::make_unique<
boost::asio::executor_work_guard<boost::asio::io_context::executor_type>>(
writeIo.get_executor());
writeThread = std::thread([this] { writeIo.run(); });
auto writeCoord = std::make_unique<WriteCoordinator>(writeIo, kgStore, repo);
writeCoord->start();
svc->__test_setWriteCoordinator(std::move(writeCoord));
}
}
int64_t upsertNode(const std::string& key, const std::string& label, const std::string& type,
const std::string& properties = {}) {
REQUIRE(kgStore);
metadata::KGNode node;
node.nodeKey = key;
node.label = label;
node.type = type;
if (!properties.empty()) {
node.properties = properties;
}
auto result = kgStore->upsertNode(node);
REQUIRE(result.has_value());
return result.value();
}
void addEdge(int64_t srcNodeId, int64_t dstNodeId, const std::string& relation,
float weight = 1.0f, const std::string& properties = {}) {
REQUIRE(kgStore);
metadata::KGEdge edge;
edge.srcNodeId = srcNodeId;
edge.dstNodeId = dstNodeId;
edge.relation = relation;
edge.weight = weight;
if (!properties.empty()) {
edge.properties = properties;
}
REQUIRE(kgStore->addEdge(edge).has_value());
}
void execSql(const std::string& sql) {
REQUIRE(pool);
auto result = pool->withConnection(
[&](metadata::Database& db) -> Result<void> { return db.execute(sql); });
REQUIRE(result.has_value());
}
std::filesystem::path testDir;
std::filesystem::path dbPath;
DaemonConfig cfg;
StubLifecycle lifecycle;
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
std::unique_ptr<ServiceManager> svc;
std::unique_ptr<RequestDispatcher> dispatcher;
boost::asio::io_context writeIo;
std::unique_ptr<boost::asio::executor_work_guard<boost::asio::io_context::executor_type>>
writeWork;
std::thread writeThread;
std::shared_ptr<metadata::ConnectionPool> pool;
std::shared_ptr<metadata::MetadataRepository> repo;
std::shared_ptr<metadata::KnowledgeGraphStore> kgStore;
};
std::filesystem::path makeTempDir(const std::string& prefix) {
auto base = std::filesystem::temp_directory_path();
auto dir = base / (prefix + std::to_string(std::random_device{}()));
std::filesystem::create_directories(dir);
return dir;
}
std::filesystem::path writeTempFile(const std::string& prefix, const std::string& name,
const std::string& content) {
auto path = makeTempDir(prefix) / name;
std::ofstream out(path, std::ios::binary);
REQUIRE(out.good());
out << content;
out.close();
REQUIRE(out.good());
return path;
}
std::filesystem::path createMockExternalPlugin(const std::filesystem::path& baseDir,
const std::string& name,
bool requireTrustedConfig = false) {
auto pluginDir = baseDir / name;
std::filesystem::create_directories(pluginDir);
{
std::ofstream pluginFile(pluginDir / "plugin.py");
pluginFile << R"PY(#!/usr/bin/env python3
import json
import sys
EXPECTED_MODE = )PY"
<< (requireTrustedConfig ? R"PY("trusted")PY" : "None") << R"PY(
def handle_request(req):
method = req.get("method", "")
if method == "handshake.manifest":
return {
"name": "dispatcher_external_plugin",
"version": "1.0.0",
"interfaces": ["content_extractor_v1"]
}
if method == "plugin.init":
if EXPECTED_MODE and req.get("params", {}).get("mode") != EXPECTED_MODE:
raise ValueError("missing configured mode")
return {"status": "initialized"}
if method == "plugin.health":
return {"status": "ok"}
if method == "plugin.shutdown":
return {"status": "ok"}
raise ValueError(f"Method not found: {method}")
for line in sys.stdin:
line = line.strip()
if not line:
continue
req_id = None
try:
req = json.loads(line)
req_id = req.get("id")
response = {"jsonrpc": "2.0", "id": req_id, "result": handle_request(req)}
except Exception as exc:
response = {
"jsonrpc": "2.0",
"id": req_id,
"error": {"code": -32603, "message": str(exc)}
}
print(json.dumps(response), flush=True)
)PY";
}
{
std::ofstream manifestFile(pluginDir / "yams-plugin.json");
manifestFile << R"JSON({
"name": "dispatcher_external_plugin",
"version": "1.0.0",
"interfaces": ["content_extractor_v1"],
"entry": {
"fallback_cmd": )JSON";
#ifdef _WIN32
manifestFile << R"JSON(["python", "-u", "${plugin_dir}/plugin.py"])JSON";
#else
manifestFile << R"JSON(["/usr/bin/env", "python3", "-u", "${plugin_dir}/plugin.py"])JSON";
#endif
manifestFile << R"JSON(,
"env": {
"PYTHONUNBUFFERED": "1"
}
}
})JSON";
}
return pluginDir;
}
std::string makePythonShimPath() {
auto shimDir = makeTempDir("yams_python_shim_");
auto shim = shimDir / "python";
{
std::ofstream shimFile(shim);
shimFile << "#!/bin/sh\n"
"exec /usr/bin/env python3 \"$@\"\n";
}
std::filesystem::permissions(shim,
std::filesystem::perms::owner_read |
std::filesystem::perms::owner_write |
std::filesystem::perms::owner_exec,
std::filesystem::perm_options::replace);
if (const char* currentPath = std::getenv("PATH"); currentPath && *currentPath) {
return shimDir.string() + ":" + currentPath;
}
return shimDir.string() + ":/usr/bin:/bin";
}
Response dispatchRequest(RequestDispatcher& dispatcher, const Request& req) {
boost::asio::io_context ioc;
auto fut = boost::asio::co_spawn(ioc, dispatcher.dispatch(req), boost::asio::use_future);
ioc.run();
return fut.get();
}
std::shared_ptr<SpscQueue<InternalEventBus::StoreDocumentTask>>
getStoreDocumentTaskChannel(std::size_t capacity = 0) {
auto existing = InternalEventBus::instance().get_channel<InternalEventBus::StoreDocumentTask>(
"store_document_tasks");
if (existing) {
return existing;
}
if (capacity == 0) {
capacity = static_cast<std::size_t>(TuneAdvisor::storeDocumentChannelCapacity());
}
return InternalEventBus::instance().get_or_create_channel<InternalEventBus::StoreDocumentTask>(
"store_document_tasks", capacity);
}
void drainStoreDocumentTasks() {
auto channel = InternalEventBus::instance().get_channel<InternalEventBus::StoreDocumentTask>(
"store_document_tasks");
if (!channel) {
return;
}
InternalEventBus::StoreDocumentTask task;
while (channel->try_pop(task)) {
}
}
template <typename Predicate> bool waitForCondition(Predicate&& predicate, int attempts = 100) {
for (int i = 0; i < attempts; ++i) {
if (predicate()) {
return true;
}
std::this_thread::sleep_for(std::chrono::milliseconds(20));
}
return predicate();
}
} // namespace
// =============================================================================
// Embedding Status Tests
// =============================================================================
TEST_CASE("DaemonMetrics: Embedding provider status", "[daemon][metrics][embedding]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_metrics_embed_");
ServiceManager svc(cfg, state, lifecycleFsm);
SECTION("Snapshot includes embedding backend information") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/test-model.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setAdoptedProviderPluginName("stub");
// Model provider is now directly used - no need to ensure generator
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
// Backend should be "plugin:<name>" format or "unknown"
bool validBackend =
(snap->embeddingBackend == "unknown" || snap->embeddingBackend.starts_with("plugin:"));
REQUIRE(validBackend);
}
SECTION("Metrics snapshot doesn't crash without provider") {
// No provider set
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
REQUIRE(!snap->embeddingBackend.empty());
}
}
TEST_CASE("DaemonMetrics reports effective embedding policy before provider adoption",
"[daemon][metrics][embedding][config][snapshot]") {
const auto dataDir = makeTempDir("yams_metrics_embed_policy_");
const auto configPath = yams::test::write_file(dataDir / "config.toml", R"toml(
[embeddings]
backend = "simeon"
preferred_model = "simeon-default"
embedding_dim = 1024
preload_on_startup = false
)toml");
yams::test::ScopedEnvVar configPathEnvironment{"YAMS_CONFIG_PATH", std::nullopt};
yams::test::ScopedEnvVar configEnvironment{"YAMS_CONFIG", std::nullopt};
yams::test::ScopedEnvVar backendEnvironment{"YAMS_EMBED_BACKEND", std::nullopt};
yams::test::ScopedEnvVar modelEnvironment{"YAMS_PREFERRED_MODEL", std::nullopt};
yams::test::ScopedEnvVar dimensionEnvironment{"YAMS_EMBED_DIM", std::nullopt};
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = dataDir;
cfg.configFilePath = configPath;
ServiceManager svc(cfg, state, lifecycleFsm);
REQUIRE(svc.initialize().has_value());
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
const auto snapshot = metrics.getSnapshot();
REQUIRE((snapshot != nullptr));
CHECK((snapshot->embeddingBackend == "simeon"));
CHECK((snapshot->embeddingModel == "simeon-default"));
CHECK((snapshot->embeddingDim == 1024U));
const auto policy = svc.getResolvedEmbeddingConfig();
REQUIRE((policy != nullptr));
CHECK((policy->policyIdentity == "simeon:simeon-default:1024"));
CHECK((policy->effectiveConfigPath == configPath));
}
// =============================================================================
// WAL Metrics Tests
// =============================================================================
TEST_CASE("DaemonMetrics: WAL metrics in GetStats", "[daemon][metrics][wal]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_metrics_wal_");
YamsDaemon daemon(cfg);
DaemonLifecycleAdapter lifecycleAdapter(&daemon);
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycleAdapter, &svc, &state);
SECTION("GetStats includes WAL metric keys") {
GetStatsRequest req;
Request r = req;
boost::asio::io_context ioc;
auto fut = boost::asio::co_spawn(ioc, dispatcher.dispatch(r), boost::asio::use_future);
ioc.run();
auto resp = fut.get();
REQUIRE(std::holds_alternative<GetStatsResponse>(resp));
auto stats = std::get<GetStatsResponse>(resp);
// WAL metrics should be present
REQUIRE(stats.additionalStats.count("wal_active_transactions") > 0);
REQUIRE(stats.additionalStats.count("wal_pending_entries") > 0);
}
}
TEST_CASE("RequestDispatcher: status includes canonical readiness flags",
"[daemon][status][readiness]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_readiness_flags_");
YamsDaemon daemon(cfg);
DaemonLifecycleAdapter lifecycleAdapter(&daemon);
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycleAdapter, &svc, &state);
StatusRequest req;
req.detailed = false;
Request r = req;
boost::asio::io_context ioc;
auto fut = boost::asio::co_spawn(ioc, dispatcher.dispatch(r), boost::asio::use_future);
ioc.run();
auto resp = fut.get();
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& status = std::get<StatusResponse>(resp);
const std::vector<std::string_view> requiredCoreReadinessKeys = {
readiness::kIpcServer,
readiness::kContentStore,
readiness::kDatabase,
readiness::kMetadataRepo,
readiness::kSearchEngine,
readiness::kModelProvider,
readiness::kVectorIndex,
readiness::kVectorDb,
readiness::kPlugins,
readiness::kVectorDbInitAttempted,
readiness::kVectorDbReady,
readiness::kVectorDbDim,
readiness::kEmbeddingReady,
readiness::kEmbeddingDegraded,
readiness::kPluginsReady,
readiness::kPluginsDegraded,
readiness::kContentExtractorsReady,
readiness::kSymbolExtractorsReady,
readiness::kEntityExtractorsReady,
readiness::kTitleExtractorReady,
readiness::kRepairService,
readiness::kSearchEngineBuildReasonInitial,
readiness::kSearchEngineBuildReasonRebuild,
readiness::kSearchEngineBuildReasonDegraded,
readiness::kSearchEngineLexicalReady,
readiness::kSearchEngineVectorUsable,
readiness::kSearchEngineHybridUsable,
readiness::kSearchEngineLexicalEnhancementConfigured,
readiness::kSearchEngineLexicalEnhancementReady,
readiness::kSearchEngineLexicalEnhancementBuilding,
readiness::kSearchEngineFragmentGeometryReady};
for (const auto key : requiredCoreReadinessKeys) {
INFO("missing readiness key: " << key);
REQUIRE(status.readinessStates.count(std::string(key)) > 0);
}
const std::array<std::string_view, 5> requiredCapabilityMetricKeys = {
metrics::kContentExtractorsLoaded, metrics::kSymbolExtractorsLoaded,
metrics::kEntityExtractorsLoaded, metrics::kTitleExtractorEnabled,
metrics::kPluginSkippedCount,
};
for (const auto key : requiredCapabilityMetricKeys) {
INFO("missing capability metric key: " << key);
REQUIRE(status.requestCounts.count(std::string(key)) > 0);
}
}
TEST_CASE("RequestDispatcher: compact and detailed status agree on FSM readiness",
"[daemon][status][readiness]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_fsm_readiness_parity_");
YamsDaemon daemon(cfg);
DaemonLifecycleAdapter lifecycleAdapter(&daemon);
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycleAdapter, &svc, &state);
auto fetch = [&](bool detailed) {
StatusRequest req;
req.detailed = detailed;
Request r = req;
boost::asio::io_context ioc;
auto fut = boost::asio::co_spawn(ioc, dispatcher.dispatch(r), boost::asio::use_future);
ioc.run();
auto resp = fut.get();
REQUIRE(std::holds_alternative<StatusResponse>(resp));
return std::get<StatusResponse>(resp);
};
const auto compact = fetch(false);
const auto detailed = fetch(true);
for (const auto key : {readiness::kEmbeddingReady, readiness::kEmbeddingDegraded,
readiness::kPluginsReady, readiness::kPluginsDegraded}) {
const std::string k(key);
INFO("readiness key must agree between compact and detailed: " << k);
REQUIRE(compact.readinessStates.count(k) > 0);
REQUIRE(detailed.readinessStates.count(k) > 0);
REQUIRE(compact.readinessStates.at(k) == detailed.readinessStates.at(k));
}
}
TEST_CASE("RequestDispatcher: status includes repair metrics and flags",
"[daemon][status][repair]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_repair_metrics_");
YamsDaemon daemon(cfg);
DaemonLifecycleAdapter lifecycleAdapter(&daemon);
StateComponent state;
state.stats.repairQueueDepth.store(7);
state.stats.repairBatchesAttempted.store(11);
state.stats.repairEmbeddingsGenerated.store(5);
state.stats.repairEmbeddingsSkipped.store(2);
state.stats.repairFailedOperations.store(1);
state.stats.repairIdleTicks.store(99);
state.stats.repairBusyTicks.store(33);
state.stats.repairCurrentOperationCode.store(10);
state.stats.repairCurrentOperationStartedMs.store(
static_cast<uint64_t>(std::chrono::duration_cast<std::chrono::milliseconds>(
std::chrono::steady_clock::now().time_since_epoch())
.count()) -
1500);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycleAdapter, &svc, &state);
StatusRequest req;
req.detailed = true;
Request r = req;
boost::asio::io_context ioc;
auto fut = boost::asio::co_spawn(ioc, dispatcher.dispatch(r), boost::asio::use_future);
ioc.run();
auto resp = fut.get();
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& status = std::get<StatusResponse>(resp);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairQueueDepth)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairBatchesAttempted)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairEmbeddingsGenerated)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairEmbeddingsSkipped)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairFailedOperations)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairIdleTicks)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairBusyTicks)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairRunning)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairInProgress)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairCurrentOperationCode)) > 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairCurrentOperationElapsedMs)) > 0);
CHECK(status.requestCounts.at(std::string(metrics::kRepairCurrentOperationCode)) == 10);
CHECK(status.requestCounts.at(std::string(metrics::kRepairCurrentOperationElapsedMs)) >= 1);
}
TEST_CASE("RequestDispatcher: status includes the current request in requestsProcessed",
"[daemon][status][requests]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_request_count_");
YamsDaemon daemon(cfg);
DaemonLifecycleAdapter lifecycleAdapter(&daemon);
StateComponent state;
state.stats.requestsProcessed.store(41, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycleAdapter, &svc, &state);
auto resp = dispatchRequest(dispatcher, Request{StatusRequest{}});
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& status = std::get<StatusResponse>(resp);
CHECK(status.requestsProcessed == 42);
CHECK(state.stats.requestsProcessed.load(std::memory_order_relaxed) == 42);
}
TEST_CASE("RequestDispatcher: status responds even when lifecycle not ready",
"[daemon][status][readiness]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_not_ready_");
YamsDaemon daemon(cfg);
DaemonLifecycleAdapter lifecycleAdapter(&daemon);
// Daemon is not started; lifecycle state is expected to be non-ready.
{
const auto snap = daemon.getLifecycle().snapshot();
REQUIRE(snap.state != LifecycleState::Ready);
REQUIRE(snap.state != LifecycleState::Degraded);
}
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycleAdapter, &svc, &state);
StatusRequest req;
req.detailed = true;
Request r = req;
boost::asio::io_context ioc;
auto fut = boost::asio::co_spawn(ioc, dispatcher.dispatch(r), boost::asio::use_future);
ioc.run();
auto resp = fut.get();
REQUIRE(std::holds_alternative<StatusResponse>(resp));
}
TEST_CASE("RequestDispatcher: tree diff handler validates inputs and missing metadata repo",
"[daemon][tree-diff][dispatcher]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_tree_diff_dispatcher_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
SECTION("returns invalid argument when snapshot ids are missing") {
ListTreeDiffRequest req;
req.targetSnapshotId = "target-only";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message.find("baseSnapshotId") != std::string::npos);
}
SECTION("returns internal error when metadata repository is unavailable") {
ListTreeDiffRequest req;
req.baseSnapshotId = "base";
req.targetSnapshotId = "target";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Metadata repository not available") != std::string::npos);
}
}
TEST_CASE("RequestDispatcher: prune handler reports missing metadata repo",
"[daemon][prune][dispatcher]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_prune_dispatcher_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
PruneRequest req;
req.dryRun = true;
req.extensions = {"log"};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Metadata repository unavailable") != std::string::npos);
}
TEST_CASE("RequestDispatcher: prune handler covers parsing and execution branches",
"[daemon][prune][dispatcher]") {
auto makeDoc = [](int64_t id, std::string filePath, std::string ext, int64_t size,
std::string hash, int64_t ageDays) {
metadata::DocumentInfo doc{};
doc.id = id;
doc.filePath = std::move(filePath);
doc.fileName = std::filesystem::path(doc.filePath).filename().string();
doc.fileExtension = std::move(ext);
doc.fileSize = size;
doc.sha256Hash = std::move(hash);
doc.mimeType = "text/plain";
const auto now =
std::chrono::time_point_cast<std::chrono::seconds>(std::chrono::system_clock::now());
doc.modifiedTime = now - std::chrono::hours(ageDays * 24);
doc.indexedTime = now;
return doc;
};
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_prune_dispatcher_live_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto store = std::make_shared<StubContentStore>();
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
repo->addDocument(makeDoc(1, "build/output.log", "log", 2048, std::string(64, 'a'), 10));
repo->addDocument(
makeDoc(2, "logs/server.log", ".log", 3 * 1024 * 1024, std::string(64, 'b'), 14));
repo->addDocument(makeDoc(3, "notes/readme.txt", "txt", 64, std::string(64, 'c'), 2));
repo->addDocument(makeDoc(4, "cache/old.tmp", "tmp", 4096, std::string(64, 'd'), 500));
SECTION("dry run parses age and size filters across units") {
PruneRequest req;
req.dryRun = true;
req.olderThan = "1w";
req.largerThan = "1KB";
req.smallerThan = "5MB";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.filesDeleted == 0);
CHECK(pruneResp.filesFailed == 0);
CHECK(pruneResp.totalBytesFreed == (2048 + 3 * 1024 * 1024 + 4096));
REQUIRE(pruneResp.categoryCounts.count("build-object") == 1);
REQUIRE(pruneResp.categoryCounts.count("logs") == 1);
REQUIRE(pruneResp.categoryCounts.count("temp") == 1);
CHECK(pruneResp.categoryCounts.at("build-object") == 1);
CHECK(pruneResp.categoryCounts.at("logs") == 1);
CHECK(pruneResp.categoryCounts.at("temp") == 1);
}
SECTION("dry run accepts month and year age units") {
PruneRequest monthReq;
monthReq.dryRun = true;
monthReq.olderThan = "1m";
auto monthResp = dispatchRequest(dispatcher, Request{monthReq});
REQUIRE(std::holds_alternative<PruneResponse>(monthResp));
const auto& monthPruneResp = std::get<PruneResponse>(monthResp);
REQUIRE(monthPruneResp.categoryCounts.count("temp") == 1);
CHECK(monthPruneResp.categoryCounts.at("temp") == 1);
CHECK(monthPruneResp.categoryCounts.count("logs") == 0);
PruneRequest yearReq;
yearReq.dryRun = true;
yearReq.olderThan = "1y";
auto yearResp = dispatchRequest(dispatcher, Request{yearReq});
REQUIRE(std::holds_alternative<PruneResponse>(yearResp));
const auto& yearPruneResp = std::get<PruneResponse>(yearResp);
CHECK(yearPruneResp.filesDeleted == 0);
CHECK(yearPruneResp.totalBytesFreed == 4096);
REQUIRE(yearPruneResp.categoryCounts.count("temp") == 1);
CHECK(yearPruneResp.categoryCounts.at("temp") == 1);
}
SECTION("dry run accepts gigabyte size units") {
PruneRequest req;
req.dryRun = true;
req.largerThan = "1GB";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.totalBytesFreed == 0);
CHECK(pruneResp.categoryCounts.empty());
}
SECTION("dry run falls back on unknown age unit and raw byte sizes") {
PruneRequest req;
req.dryRun = true;
req.olderThan = "3q";
req.largerThan = "2000";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.totalBytesFreed == (2048 + 3 * 1024 * 1024 + 4096));
REQUIRE(pruneResp.categoryCounts.count("build-object") == 1);
REQUIRE(pruneResp.categoryCounts.count("logs") == 1);
REQUIRE(pruneResp.categoryCounts.count("temp") == 1);
CHECK(pruneResp.categoryCounts.at("build-object") == 1);
CHECK(pruneResp.categoryCounts.at("logs") == 1);
CHECK(pruneResp.categoryCounts.at("temp") == 1);
}
SECTION("dry run filters by extension and categories") {
PruneRequest req;
req.dryRun = true;
req.categories = {"build"};
req.extensions = {"log"};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.totalBytesFreed == 2048);
CHECK(pruneResp.categoryCounts.size() == 1);
REQUIRE(pruneResp.categoryCounts.count("build-object") == 1);
CHECK(pruneResp.categoryCounts.at("build-object") == 1);
}
SECTION("dry run matches dotted extensions when category is logs") {
PruneRequest req;
req.dryRun = true;
req.categories = {"logs"};
req.extensions = {"log"};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.totalBytesFreed == 3 * 1024 * 1024);
CHECK(pruneResp.categoryCounts.size() == 1);
REQUIRE(pruneResp.categoryCounts.count("logs") == 1);
CHECK(pruneResp.categoryCounts.at("logs") == 1);
}
SECTION("dry run handles query failures from repository") {
repo->setQueryError(Error{ErrorCode::InternalError, "query boom"});
auto resp = dispatchRequest(dispatcher, Request{PruneRequest{}});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.errorMessage.empty());
CHECK(pruneResp.totalBytesFreed == 0);
CHECK(pruneResp.categoryCounts.empty());
}
SECTION("dry run handles repository exceptions") {
repo->setThrowOnQuery("query exception");
auto resp = dispatchRequest(dispatcher, Request{PruneRequest{}});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.errorMessage == "Failed to query candidates: query exception");
}
SECTION("apply mode counts deleted missing failed and exception cases") {
repo->setMissingHash(std::string(64, 'b'));
repo->setDeleteError(4, Error{ErrorCode::InternalError, "delete failed"});
repo->setThrowOnLookupHash(std::string(64, 'c'), "lookup boom");
PruneRequest req;
req.dryRun = false;
req.olderThan = "1d";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.filesDeleted == 0);
CHECK(pruneResp.filesFailed == 0);
CHECK(pruneResp.totalBytesFreed == (2048 + 3 * 1024 * 1024 + 64 + 4096));
REQUIRE(pruneResp.categoryCounts.count("build-object") == 1);
REQUIRE(pruneResp.categoryCounts.count("logs") == 1);
REQUIRE(pruneResp.categoryCounts.count("temp") == 1);
CHECK(pruneResp.categoryCounts.at("build-object") == 1);
CHECK(pruneResp.categoryCounts.at("logs") == 1);
CHECK(pruneResp.categoryCounts.at("temp") == 1);
CHECK_FALSE(pruneResp.statusMessage.empty());
REQUIRE(waitForCondition([&] { return repo->deleteCallCount() == 2; }));
CHECK(repo->deleteCallCount() == 2);
CHECK(repo->deletedIds() == std::vector<int64_t>{1});
const auto removed = store->removedHashes();
CHECK(std::find(removed.begin(), removed.end(), std::string(64, 'd')) != removed.end());
}
SECTION("apply mode removes content through the document deletion service") {
const auto targetHash = std::string(64, 'a');
store->setBlob(targetHash, "indexed log bytes");
PruneRequest req;
req.dryRun = false;
req.categories = {"build"};
req.extensions = {"log"};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.filesDeleted == 0);
CHECK(pruneResp.filesFailed == 0);
CHECK(pruneResp.totalBytesFreed == 2048);
CHECK_FALSE(pruneResp.statusMessage.empty());
REQUIRE(waitForCondition([&] { return repo->deleteCallCount() == 1; }));
CHECK_FALSE(store->hasBlob(targetHash));
CHECK(store->removedHashes() == std::vector<std::string>{targetHash});
CHECK(repo->deletedIds() == std::vector<int64_t>{1});
}
SECTION("apply mode yields during large prune batches") {
auto fullHashFor = [](int i) {
static constexpr char kHex[] = "0123456789abcdef";
std::string hash(64, 'e');
hash[62] = kHex[(i / 16) & 0x0f];
hash[63] = kHex[i & 0x0f];
return hash;
};
for (int i = 0; i < 105; ++i) {
repo->addDocument(makeDoc(1000 + i, "cache/bulk_" + std::to_string(i) + ".tmp", "tmp",
2048, fullHashFor(i), 30));
}
PruneRequest req;
req.dryRun = false;
req.categories = {"temp"};
req.extensions = {"tmp"};
req.olderThan = "1d";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.filesDeleted == 0);
CHECK(pruneResp.filesFailed == 0);
CHECK(pruneResp.totalBytesFreed == (4096 + 105 * 2048));
REQUIRE(pruneResp.categoryCounts.count("temp") == 1);
CHECK(pruneResp.categoryCounts.at("temp") == 106);
CHECK_FALSE(pruneResp.statusMessage.empty());
REQUIRE(waitForCondition([&] { return repo->deleteCallCount() == 106; }));
CHECK(repo->deleteCallCount() == 106);
CHECK(repo->deletedIds().size() == 106);
}
SECTION("dry run leaves empty size filters as zero") {
PruneRequest req;
req.dryRun = true;
req.largerThan = "";
req.smallerThan = "";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PruneResponse>(resp));
const auto& pruneResp = std::get<PruneResponse>(resp);
CHECK(pruneResp.totalBytesFreed == (2048 + 3 * 1024 * 1024 + 64 + 4096));
REQUIRE(pruneResp.categoryCounts.count("build-object") == 1);
REQUIRE(pruneResp.categoryCounts.count("logs") == 1);
REQUIRE(pruneResp.categoryCounts.count("temp") == 1);
CHECK(pruneResp.categoryCounts.at("build-object") == 1);
CHECK(pruneResp.categoryCounts.at("logs") == 1);
CHECK(pruneResp.categoryCounts.at("temp") == 1);
}
}
TEST_CASE("RequestDispatcher: model handlers cover validation and status branches",
"[daemon][model][dispatcher]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_model_dispatcher_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
SECTION("load model rejects missing name when provider exists") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/model.onnx");
svc.__test_setModelProvider(provider);
LoadModelRequest req;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidData);
CHECK(err.message == "modelName is required");
}
SECTION("load model with options succeeds and records provider state") {
auto provider = std::make_shared<StubModelProvider>(768, "/tmp/provider-model.onnx");
svc.__test_setModelProvider(provider);
LoadModelRequest req;
req.modelName = "dispatcher-model";
req.optionsJson = R"({"mode":"offline"})";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
const auto& loadResp = std::get<ModelLoadResponse>(resp);
CHECK(loadResp.success);
CHECK(loadResp.modelName == "dispatcher-model");
CHECK(provider->isModelLoaded("dispatcher-model"));
CHECK(provider->lastOptionsModel() == "dispatcher-model");
CHECK(provider->lastOptionsJson() == req.optionsJson);
}
SECTION("load model failure marks embedding degraded") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/failing-model.onnx");
provider->setLoadError(Error{ErrorCode::InvalidState, "provider refused load"});
svc.__test_setModelProvider(provider);
lifecycle.resetDegradedState();
LoadModelRequest req;
req.modelName = "failing-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "provider refused load");
CHECK(lifecycle.setSubsystemDegradedCalls() == 1);
CHECK(lifecycle.lastSubsystem() == "embedding");
CHECK(lifecycle.lastDegraded());
CHECK(lifecycle.lastReason() == "provider_load_failed");
}
SECTION("load model sanitizes invalid utf8 in provider errors") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/utf8-model.onnx");
std::string invalidMessage = "prefix ";
invalidMessage.push_back(static_cast<char>(0xC2));
invalidMessage.push_back(static_cast<char>(0xA9));
invalidMessage.push_back(static_cast<char>(0xE2));
invalidMessage.push_back(static_cast<char>(0x82));
invalidMessage.push_back(static_cast<char>(0xAC));
invalidMessage.push_back(static_cast<char>(0xF0));
invalidMessage.push_back(static_cast<char>(0x9F));
invalidMessage.push_back(static_cast<char>(0x92));
invalidMessage.push_back(static_cast<char>(0xA9));
invalidMessage.push_back(static_cast<char>(0xC0));
invalidMessage.push_back(static_cast<char>(0xFF));
provider->setLoadError(Error{ErrorCode::InvalidState, invalidMessage});
svc.__test_setModelProvider(provider);
LoadModelRequest req;
req.modelName = "utf8-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
std::string expected = "prefix ";
expected.push_back(static_cast<char>(0xC2));
expected.push_back(static_cast<char>(0xA9));
expected.push_back(static_cast<char>(0xE2));
expected.push_back(static_cast<char>(0x82));
expected.push_back(static_cast<char>(0xAC));
expected.push_back(static_cast<char>(0xF0));
expected.push_back(static_cast<char>(0x9F));
expected.push_back(static_cast<char>(0x92));
expected.push_back(static_cast<char>(0xA9));
expected += "\xEF\xBF\xBD\xEF\xBF\xBD";
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == expected);
}
SECTION("load model returns internal error when provider throws after load") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/throwing-model.onnx");
provider->setThrowOnMemoryUsage(true);
svc.__test_setModelProvider(provider);
LoadModelRequest req;
req.modelName = "throwing-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Load model failed: memory boom") != std::string::npos);
}
SECTION("load model skips rebuild when model is already loaded") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/already-loaded-model.onnx");
REQUIRE(provider->loadModel("already-loaded-model").has_value());
svc.__test_setModelProvider(provider);
LoadModelRequest req;
req.modelName = "already-loaded-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
const auto& loadResp = std::get<ModelLoadResponse>(resp);
CHECK(loadResp.success);
CHECK(loadResp.modelName == "already-loaded-model");
CHECK(provider->getLoadedModelCount() == 1);
}
SECTION("load model honors timeout env floor") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/timeout-model.onnx");
svc.__test_setModelProvider(provider);
ScopedEnvVar timeoutGuard("YAMS_MODEL_LOAD_TIMEOUT_MS", "5");
struct ShutdownBeforeEnvRestore {
ServiceManager& svc;
~ShutdownBeforeEnvRestore() { svc.shutdown(); }
} shutdownBeforeEnvRestore{svc};
LoadModelRequest req;
req.modelName = "timeout-clamped-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
CHECK(std::get<ModelLoadResponse>(resp).success);
CHECK(provider->isModelLoaded("timeout-clamped-model"));
}
SECTION("load model survives invalid timeout env value") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/timeout-invalid.onnx");
svc.__test_setModelProvider(provider);
ScopedEnvVar timeoutGuard("YAMS_MODEL_LOAD_TIMEOUT_MS", "not-a-number");
struct ShutdownBeforeEnvRestore {
ServiceManager& svc;
~ShutdownBeforeEnvRestore() { svc.shutdown(); }
} shutdownBeforeEnvRestore{svc};
LoadModelRequest req;
req.modelName = "timeout-invalid-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
CHECK(std::get<ModelLoadResponse>(resp).success);
CHECK(provider->isModelLoaded("timeout-invalid-model"));
}
SECTION("load model rebuilds when cached search engine reports missing embedding generator") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/rebuild-health.onnx");
REQUIRE(provider->loadModel("already-loaded-health-model").has_value());
svc.__test_setModelProvider(provider);
metadata::ConnectionPoolConfig poolCfg{};
poolCfg.enableWAL = false;
poolCfg.enableForeignKeys = false;
auto pool = std::make_shared<metadata::ConnectionPool>(":memory:", poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = std::make_shared<metadata::MetadataRepository>(*pool);
search::SearchEngineConfig config;
config.vectorWeight = 1.0f;
auto unhealthyEngine =
std::make_shared<search::SearchEngine>(repo, nullptr, nullptr, nullptr, config);
svc.__test_setCachedSearchEngine(unhealthyEngine, true);
LoadModelRequest req;
req.modelName = "already-loaded-health-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
CHECK(std::get<ModelLoadResponse>(resp).success);
CHECK(provider->getLoadedModelCount() == 1);
}
SECTION("load model ignores isModelLoaded exceptions before loading") {
auto provider = std::make_shared<StubModelProvider>(4, "/tmp/isloaded-throw.onnx");
svc.__test_setModelProvider(provider);
metadata::ConnectionPoolConfig poolCfg{};
poolCfg.enableWAL = false;
poolCfg.enableForeignKeys = false;
auto pool = std::make_shared<metadata::ConnectionPool>(":memory:", poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = std::make_shared<metadata::MetadataRepository>(*pool);
yams::vector::VectorDatabaseConfig vectorCfg{};
vectorCfg.database_path = ":memory:";
vectorCfg.embedding_dim = 4;
vectorCfg.create_if_missing = true;
vectorCfg.use_in_memory = true;
auto vectorDb = std::make_shared<yams::vector::VectorDatabase>(vectorCfg);
REQUIRE(vectorDb->initialize());
search::SearchEngineConfig config;
config.vectorWeight = 1.0f;
auto healthyEngine = std::make_shared<search::SearchEngine>(
repo, vectorDb, provider->getEmbeddingGenerator(), nullptr, config);
svc.__test_setCachedSearchEngine(healthyEngine, true);
provider->setThrowOnIsModelLoadedCount(1);
LoadModelRequest req;
req.modelName = "isloaded-throw-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
const auto& loadResp = std::get<ModelLoadResponse>(resp);
CHECK(loadResp.success);
CHECK(loadResp.modelName == "isloaded-throw-model");
provider->setThrowOnIsModelLoaded(false);
CHECK(provider->isModelLoaded("isloaded-throw-model"));
}
SECTION("load model survives lifecycle degradation callback exceptions") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/lifecycle-throw.onnx");
svc.__test_setModelProvider(provider);
lifecycle.setThrowOnSetSubsystemDegraded(true);
LoadModelRequest req;
req.modelName = "lifecycle-throw-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelLoadResponse>(resp));
const auto& loadResp = std::get<ModelLoadResponse>(resp);
CHECK(loadResp.success);
CHECK(loadResp.modelName == "lifecycle-throw-model");
CHECK(provider->isModelLoaded("lifecycle-throw-model"));
lifecycle.setThrowOnSetSubsystemDegraded(false);
}
SECTION("unload model rejects missing name") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/model.onnx");
svc.__test_setModelProvider(provider);
UnloadModelRequest req;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidData);
CHECK(err.message == "modelName is required");
}
SECTION("unload model propagates provider missing-model error") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/model.onnx");
svc.__test_setModelProvider(provider);
UnloadModelRequest req;
req.modelName = "missing-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
}
SECTION("unload model reports provider not ready when no provider is available") {
auto resp = dispatchRequest(dispatcher, Request{UnloadModelRequest{"missing-provider"}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message.find("Plugin system") != std::string::npos);
}
SECTION("model status returns loaded model details and filters by name") {
auto provider = std::make_shared<StubModelProvider>(512, "/tmp/status-model.onnx");
svc.__test_setModelProvider(provider);
REQUIRE(provider->loadModel("status-model").has_value());
ModelStatusRequest req;
req.modelName = "status-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ModelStatusResponse>(resp));
const auto& statusResp = std::get<ModelStatusResponse>(resp);
REQUIRE(statusResp.models.size() == 1);
CHECK(statusResp.models.front().name == "status-model");
CHECK(statusResp.models.front().loaded);
CHECK(statusResp.models.front().embeddingDim == 512);
}
SECTION("model status handles model info lookup failure") {
auto provider = std::make_shared<StubModelProvider>(256, "/tmp/status-fallback.onnx");
REQUIRE(provider->loadModel("status-fallback").has_value());
provider->setFailModelInfo(true);
svc.__test_setModelProvider(provider);
auto resp = dispatchRequest(dispatcher, Request{ModelStatusRequest{}});
REQUIRE(std::holds_alternative<ModelStatusResponse>(resp));
const auto& statusResp = std::get<ModelStatusResponse>(resp);
REQUIRE(statusResp.models.size() == 1);
CHECK(statusResp.models.front().name == "status-fallback");
CHECK(statusResp.models.front().memoryMb == 0);
CHECK(statusResp.models.front().maxSequenceLength == 0);
CHECK(statusResp.models.front().embeddingDim == 256);
}
SECTION("model status returns empty response when provider unavailable") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/model.onnx");
provider->setAvailable(false);
svc.__test_setModelProvider(provider);
auto resp = dispatchRequest(dispatcher, Request{ModelStatusRequest{}});
REQUIRE(std::holds_alternative<ModelStatusResponse>(resp));
const auto& statusResp = std::get<ModelStatusResponse>(resp);
CHECK(statusResp.models.empty());
CHECK(statusResp.totalMemoryMb == 0);
}
SECTION("model status converts provider exceptions into internal errors") {
auto provider = std::make_shared<StubModelProvider>(384, "/tmp/model.onnx");
REQUIRE(provider->loadModel("status-throw").has_value());
provider->setThrowOnMemoryUsage(true);
svc.__test_setModelProvider(provider);
auto resp = dispatchRequest(dispatcher, Request{ModelStatusRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("memory boom") != std::string::npos);
}
}
TEST_CASE("RequestDispatcher: document handlers cover direct helper and error branches",
"[daemon][documents][dispatcher]") {
auto makeDoc = [](int64_t id, std::string filePath, std::string hash) {
metadata::DocumentInfo doc{};
doc.id = id;
doc.filePath = std::move(filePath);
doc.fileName = std::filesystem::path(doc.filePath).filename().string();
doc.fileExtension = std::filesystem::path(doc.filePath).extension().string();
doc.fileSize = 12;
doc.sha256Hash = std::move(hash);
doc.mimeType = "text/plain";
doc.setIndexedTime(1710000000 + id);
return doc;
};
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_documents_dispatcher_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
SECTION("prepareSession returns missing-session error for unknown session") {
ScopedEnvVar stateHome("XDG_STATE_HOME",
makeTempDir("yams_prepare_missing_session_").string().c_str());
RequestDispatcher::PrepareSessionOptions opts;
opts.sessionName = "missing";
CHECK(dispatcher.prepareSession(opts) == -2);
}
SECTION("prepareSession warms current session documents") {
auto stateRoot = makeTempDir("yams_prepare_session_");
ScopedEnvVar stateHome("XDG_STATE_HOME", stateRoot.string().c_str());
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(7, "src/demo.cpp", "warm-hash");
repo->setPathTreeDocuments({doc});
store->setBlob("warm-hash", "hello snippet world");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
auto sessionSvc = app::services::makeSessionService(nullptr);
REQUIRE(sessionSvc != nullptr);
sessionSvc->init("warm-session", "coverage");
sessionSvc->addPathSelector("src/**/*.cpp", {}, {});
RequestDispatcher::PrepareSessionOptions opts;
opts.sessionName = "warm-session";
opts.limit = 5;
opts.snippetLen = 5;
CHECK(dispatcher.prepareSession(opts) == 1);
auto materialized = sessionSvc->listMaterialized("warm-session");
REQUIRE(materialized.size() == 1);
CHECK(materialized.front().hash == "warm-hash");
}
SECTION("cancel request succeeds when context is registered") {
auto ctx = std::make_shared<RequestContext>();
RequestContextRegistry::instance().register_context(4242, ctx);
auto resp = dispatchRequest(dispatcher, Request{CancelRequest{4242}});
RequestContextRegistry::instance().deregister_context(4242);
REQUIRE(std::holds_alternative<SuccessResponse>(resp));
CHECK(std::get<SuccessResponse>(resp).message == "Cancel accepted");
CHECK(ctx->canceled.load(std::memory_order_relaxed));
}
SECTION("get init rejects missing hash and name") {
auto resp = dispatchRequest(dispatcher, Request{GetInitRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message.find("hash or name required") != std::string::npos);
}
SECTION("get init reports missing content store") {
GetInitRequest req;
req.hash = "missing-store-hash";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("content store unavailable") != std::string::npos);
}
SECTION("get init reports name resolution failure") {
GetInitRequest req;
req.name = "missing-name";
req.byName = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("Metadata repository not available") != std::string::npos);
}
SECTION("get init reports retrieve bytes failure") {
auto store = std::make_shared<StubContentStore>();
svc.__test_setContentStore(store);
GetInitRequest req;
req.hash = "missing-bytes";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("retrieveBytes failed") != std::string::npos);
}
SECTION("get init reports missing retrieval session manager") {
auto store = std::make_shared<StubContentStore>();
store->setBlob("needs-session-hash", "hello retrieval session");
svc.__test_setContentStore(store);
GetInitRequest req;
req.hash = "needs-session-hash";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("retrieval session manager unavailable") != std::string::npos);
}
SECTION("get init chunk and end succeed across retrieval session lifecycle") {
auto store = std::make_shared<StubContentStore>();
store->setBlob("session-flow-hash", "hello retrieval flow");
svc.__test_setContentStore(store);
svc.__test_setRetrievalSessionManager(std::make_unique<RetrievalSessionManager>());
GetInitRequest initReq;
initReq.hash = "session-flow-hash";
initReq.chunkSize = 5;
auto initResp = dispatchRequest(dispatcher, Request{initReq});
REQUIRE(std::holds_alternative<GetInitResponse>(initResp));
const auto& init = std::get<GetInitResponse>(initResp);
CHECK(init.transferId > 0);
CHECK(init.totalSize == 20);
CHECK(init.chunkSize == 5);
CHECK(init.metadata.at("hash") == "session-flow-hash");
auto chunkResp =
dispatchRequest(dispatcher, Request{GetChunkRequest{init.transferId, 0, 5}});
REQUIRE(std::holds_alternative<GetChunkResponse>(chunkResp));
const auto& chunk = std::get<GetChunkResponse>(chunkResp);
CHECK(chunk.data == "hello");
CHECK(chunk.bytesRemaining == 15);
auto endResp = dispatchRequest(dispatcher, Request{GetEndRequest{init.transferId}});
REQUIRE(std::holds_alternative<SuccessResponse>(endResp));
CHECK(std::get<SuccessResponse>(endResp).message == "OK");
auto postEndChunk =
dispatchRequest(dispatcher, Request{GetChunkRequest{init.transferId, 0, 5}});
REQUIRE(std::holds_alternative<ErrorResponse>(postEndChunk));
const auto& postEndErr = std::get<ErrorResponse>(postEndChunk);
CHECK(postEndErr.code == ErrorCode::NotFound);
CHECK(postEndErr.message.find("Invalid transferId") != std::string::npos);
}
SECTION("get chunk rejects offsets beyond total size") {
auto store = std::make_shared<StubContentStore>();
store->setBlob("offset-error-hash", "hello retrieval flow");
svc.__test_setContentStore(store);
svc.__test_setRetrievalSessionManager(std::make_unique<RetrievalSessionManager>());
GetInitRequest initReq;
initReq.hash = "offset-error-hash";
auto initResp = dispatchRequest(dispatcher, Request{initReq});
REQUIRE(std::holds_alternative<GetInitResponse>(initResp));
const auto& init = std::get<GetInitResponse>(initResp);
auto chunkResp =
dispatchRequest(dispatcher, Request{GetChunkRequest{init.transferId, 99, 5}});
REQUIRE(std::holds_alternative<ErrorResponse>(chunkResp));
const auto& err = std::get<ErrorResponse>(chunkResp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message.find("Offset beyond total size") != std::string::npos);
}
SECTION("get chunk respects max bytes cap and can return empty chunk") {
auto store = std::make_shared<StubContentStore>();
store->setBlob("cap-flow-hash", "hello retrieval flow");
svc.__test_setContentStore(store);
svc.__test_setRetrievalSessionManager(std::make_unique<RetrievalSessionManager>());
GetInitRequest initReq;
initReq.hash = "cap-flow-hash";
initReq.chunkSize = 10;
initReq.maxBytes = 7;
auto initResp = dispatchRequest(dispatcher, Request{initReq});
REQUIRE(std::holds_alternative<GetInitResponse>(initResp));
const auto& init = std::get<GetInitResponse>(initResp);
auto chunkResp =
dispatchRequest(dispatcher, Request{GetChunkRequest{init.transferId, 7, 5}});
REQUIRE(std::holds_alternative<GetChunkResponse>(chunkResp));
const auto& chunk = std::get<GetChunkResponse>(chunkResp);
CHECK(chunk.data.empty());
CHECK(chunk.bytesRemaining == 0);
}
SECTION("get chunk reports missing retrieval session manager") {
auto resp = dispatchRequest(dispatcher, Request{GetChunkRequest{9999, 0, 4}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("retrieval session manager unavailable") != std::string::npos);
}
SECTION("get chunk reports invalid transfer id") {
svc.__test_setRetrievalSessionManager(std::make_unique<RetrievalSessionManager>());
auto resp = dispatchRequest(dispatcher, Request{GetChunkRequest{9999, 0, 4}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message.find("Invalid transferId") != std::string::npos);
}
SECTION("get end succeeds even without retrieval session manager") {
auto resp = dispatchRequest(dispatcher, Request{GetEndRequest{777}});
REQUIRE(std::holds_alternative<SuccessResponse>(resp));
CHECK(std::get<SuccessResponse>(resp).message == "OK");
}
SECTION("delete rejects directory removal without recursive flag") {
DeleteRequest req;
req.directory = "cache";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message.find("recursive flag") != std::string::npos);
}
SECTION("delete recursive directory returns not found when no documents match") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto store = std::make_shared<StubContentStore>();
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
DeleteRequest req;
req.directory = "cache";
req.recursive = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message.find("No documents found matching criteria") != std::string::npos);
}
SECTION("legacy delete purge does not imply force for ambiguous names") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto store = std::make_shared<StubContentStore>();
auto firstDoc = makeDoc(19, "/tmp/delete/a/shared.txt", "delete-shared-a");
auto secondDoc = makeDoc(20, "/tmp/delete/b/shared.txt", "delete-shared-b");
repo->addDocument(firstDoc);
repo->addDocument(secondDoc);
store->setBlob(firstDoc.sha256Hash, "first");
store->setBlob(secondDoc.sha256Hash, "second");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
DeleteRequest req;
req.name = "shared.txt";
req.purge = true;
req.force = false;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidOperation);
CHECK(err.message.find("use --force") != std::string::npos);
CHECK(store->removedHashes().empty());
}
SECTION("delete reports mixed success and failure results") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto store = std::make_shared<StubContentStore>();
auto successDoc = makeDoc(21, "/tmp/delete/ok.txt", "delete-ok-hash");
auto failedDoc = makeDoc(22, "/tmp/delete/fail.txt", "delete-fail-hash");
repo->addDocument(successDoc);
repo->addDocument(failedDoc);
store->setBlob(successDoc.sha256Hash, "ok");
store->setBlob(failedDoc.sha256Hash, "fail");
store->setRemoveResult(failedDoc.sha256Hash,
Error{ErrorCode::IOError, "store remove failed"});
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
DeleteRequest req;
req.pattern = "*.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<DeleteResponse>(resp));
const auto& deleteResp = std::get<DeleteResponse>(resp);
CHECK_FALSE(deleteResp.dryRun);
CHECK(deleteResp.successCount == 1);
CHECK(deleteResp.failureCount == 1);
REQUIRE(deleteResp.results.size() == 2);
CHECK(lifecycle.removedHashes().size() == 1);
CHECK(lifecycle.removedHashes().front() == successDoc.sha256Hash);
}
SECTION("delete reports missing content store") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
svc.__test_setMetadataRepo(repo);
DeleteRequest req;
req.hash = "no-store-hash";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("Content store not available") != std::string::npos);
}
SECTION("get request maps related documents when graph view is enabled") {
auto repoPath = makeTempMetadataDbPath("yams_get_graph_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto graphQuery = std::make_shared<StubGraphQueryService>();
auto mainDoc = makeDoc(69, "/tmp/get-graph/main.txt",
"ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff");
auto relatedDoc =
makeDoc(70, "/tmp/get-graph/related.txt",
"1111111111111111111111111111111111111111111111111111111111111111");
auto mainId = repo->insertDocument(mainDoc);
auto relatedId = repo->insertDocument(relatedDoc);
REQUIRE(mainId.has_value());
REQUIRE(relatedId.has_value());
REQUIRE(repo->upsertPathTreeForDocument(mainDoc, mainId.value(), true, {}).has_value());
REQUIRE(
repo->upsertPathTreeForDocument(relatedDoc, relatedId.value(), true, {}).has_value());
graphQuery->addConnectedDocument(relatedDoc.sha256Hash, "same_content", 1, 0.8);
store->setBlob(mainDoc.sha256Hash, "main content");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
svc.__test_setGraphQueryService(graphQuery);
GetRequest req;
req.hash = mainDoc.sha256Hash;
req.showGraph = true;
req.graphDepth = 1;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<GetResponse>(resp));
const auto& getResp = std::get<GetResponse>(resp);
CHECK(getResp.graphEnabled);
CHECK(getResp.hash == mainDoc.sha256Hash);
REQUIRE(getResp.related.size() == 1);
CHECK(getResp.related.front().hash == relatedDoc.sha256Hash);
CHECK(getResp.related.front().relationship == "same_content");
}
SECTION("get request still succeeds without query trace env") {
auto repoPath = makeTempMetadataDbPath("yams_get_query_trace_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(71, "/tmp/get/query-trace.txt",
"9999999999999999999999999999999999999999999999999999999999999999");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "query trace content");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
GetRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<GetResponse>(resp));
const auto& getResp = std::get<GetResponse>(resp);
CHECK(getResp.hash == doc.sha256Hash);
CHECK(getResp.name == doc.fileName);
}
SECTION("get request maps fallback documents vector response") {
auto repoPath = makeTempMetadataDbPath("yams_get_documents_vector_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(72, "/tmp/get/documents-vector.txt",
"abababababababababababababababababababababababababababababababab");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "documents vector content");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher::__test_forceGetResponseDocumentsVectorOnce();
GetRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<GetResponse>(resp));
const auto& getResp = std::get<GetResponse>(resp);
CHECK(getResp.hash == doc.sha256Hash);
CHECK(getResp.name == doc.fileName);
CHECK(getResp.path == doc.filePath);
}
SECTION("get request returns not found when service result is forced empty") {
auto repoPath = makeTempMetadataDbPath("yams_get_empty_result_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(73, "/tmp/get/empty-result.txt",
"cdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcd");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "empty result content");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher::__test_forceGetEmptyResultOnce();
GetRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "No documents found matching criteria");
}
SECTION("file history reports missing metadata repository") {
FileHistoryRequest req;
req.filepath = "missing.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("Metadata repository not available") != std::string::npos);
}
SECTION("file history falls back to filename query and reports no snapshot versions") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto doc = makeDoc(9, "/tmp/archive/report.txt", "history-hash");
repo->addDocument(doc);
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = "report.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
CHECK_FALSE(history.found);
CHECK(history.totalVersions == 0);
CHECK(history.versions.empty());
CHECK(history.message == "File found in index but not in any snapshot");
CHECK(history.filepath.find("report.txt") != std::string::npos);
}
SECTION("file history reports file not found in index") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = "ghost.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
CHECK_FALSE(history.found);
CHECK(history.totalVersions == 0);
CHECK(history.message == "File not found in index");
}
SECTION("file history returns sorted snapshot versions from exact path") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto doc = makeDoc(11, "/tmp/archive/sorted-report.txt", "sorted-history-hash");
repo->setExactPathDocument(doc);
repo->setAllMetadata(
doc.id, {{"snapshot_id", metadata::MetadataValue("snap-alpha")},
{"snapshot_id:snap-beta", metadata::MetadataValue("")},
{"snapshot_time:snap-alpha", metadata::MetadataValue("1710000000000000")}});
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
REQUIRE(history.found);
REQUIRE(history.totalVersions == 2);
REQUIRE(history.versions.size() == 2);
CHECK(history.message == "Found 2 version(s) across snapshots");
CHECK(history.versions[0].snapshotId == "snap-beta");
CHECK(history.versions[1].snapshotId == "snap-alpha");
CHECK(history.versions[0].indexedTimestamp > history.versions[1].indexedTimestamp);
}
SECTION(
"file history falls back to document indexed time when snapshot-specific time is missing") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto doc = makeDoc(42, "/tmp/archive/fallback-time.txt", "fallback-time-hash");
repo->setExactPathDocument(doc);
repo->setAllMetadata(doc.id, {{"snapshot_id", metadata::MetadataValue("snap-fallback")}});
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
REQUIRE(history.found);
REQUIRE(history.totalVersions == 1);
REQUIRE(history.versions.size() == 1);
CHECK(history.versions.front().snapshotId == "snap-fallback");
CHECK(history.versions.front().indexedTimestamp == 1710000042);
}
SECTION("file history ignores malformed snapshot timestamps and falls back to indexed time") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto doc = makeDoc(64, "/tmp/archive/bad-snapshot-time.txt", "bad-snapshot-time-hash");
repo->setExactPathDocument(doc);
repo->setAllMetadata(doc.id, {{"snapshot_id", metadata::MetadataValue("snap-bad-time")},
{"snapshot_time", metadata::MetadataValue("not-a-number")},
{"snapshot_time:snap-bad-time",
metadata::MetadataValue("still-not-a-number")}});
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
REQUIRE(history.found);
REQUIRE(history.totalVersions == 1);
REQUIRE(history.versions.size() == 1);
CHECK(history.versions.front().snapshotId == "snap-bad-time");
CHECK(history.versions.front().indexedTimestamp == 1710000064);
}
SECTION("list request paths-only mode trims snippets metadata and tags") {
auto repoPath = makeTempMetadataDbPath("yams_list_dispatcher_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto doc = makeDoc(55, "/tmp/list/paths-only.md", "list-paths-hash");
REQUIRE(repo->insertDocument(doc).has_value());
svc.__test_setMetadataRepo(repo);
ListRequest req;
req.limit = 5;
req.pathsOnly = true;
req.showMetadata = true;
req.showTags = true;
req.showSnippets = true;
req.recent = false;
req.recentCount = 0;
req.namePattern = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ListResponse>(resp));
const auto& listResp = std::get<ListResponse>(resp);
REQUIRE(listResp.items.size() == 1);
CHECK(listResp.totalCount == 1);
CHECK(listResp.items.front().path == doc.filePath);
CHECK(listResp.items.front().snippet.empty());
CHECK(listResp.items.front().metadata.empty());
CHECK(listResp.items.front().tags.empty());
}
SECTION("list request emits dispatch timing stats by default") {
auto repoPath = makeTempMetadataDbPath("yams_list_query_trace_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto doc = makeDoc(56, "/tmp/list/query-trace.md", "list-query-trace-hash");
REQUIRE(repo->insertDocument(doc).has_value());
svc.__test_setMetadataRepo(repo);
ListRequest req;
req.limit = 3;
req.recent = false;
req.recentCount = 0;
req.namePattern = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ListResponse>(resp));
const auto& listResp = std::get<ListResponse>(resp);
REQUIRE(listResp.items.size() == 1);
CHECK(listResp.listStats.contains("phase_dispatch_total_ms"));
CHECK(listResp.listStats.contains("phase_dispatch_map_us"));
}
SECTION("list request parses filter tags and recent count") {
auto repoPath = makeTempMetadataDbPath("yams_list_filter_tags_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto doc = makeDoc(67, "/tmp/list/filter-tags.md", "list-filter-tags-hash");
REQUIRE(repo->insertDocument(doc).has_value());
svc.__test_setMetadataRepo(repo);
ListRequest req;
req.limit = 5;
req.offset = 0;
req.recent = false;
req.recentCount = 3;
req.filterTags = " alpha, beta ";
req.namePattern = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ListResponse>(resp));
const auto& listResp = std::get<ListResponse>(resp);
CHECK(listResp.totalCount >= 0);
}
SECTION("list request works when inflight limit is disabled") {
TuneAdvisorListGuard tuneGuard(0, 1);
auto repoPath = makeTempMetadataDbPath("yams_list_no_inflight_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto doc = makeDoc(68, "/tmp/list/no-inflight.md", "list-no-inflight-hash");
REQUIRE(repo->insertDocument(doc).has_value());
svc.__test_setMetadataRepo(repo);
ListRequest req;
req.limit = 5;
req.offset = 0;
req.recent = true;
req.showSnippets = true;
req.noSnippets = true;
req.namePattern = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ListResponse>(resp));
const auto& listResp = std::get<ListResponse>(resp);
REQUIRE(listResp.items.size() == 1);
CHECK(listResp.items.front().snippet.empty());
}
SECTION("list request reports missing metadata repository") {
ListRequest req;
req.limit = 2;
req.recent = false;
req.recentCount = 0;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message.find("Metadata repository not available") != std::string::npos);
}
SECTION("list request reports resource exhaustion when inflight limit is saturated") {
TuneAdvisorListGuard tuneGuard(1, 1);
auto repoPath = makeTempMetadataDbPath("yams_list_exhaustion_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto doc = makeDoc(60, "/tmp/list/exhausted.md", "list-exhausted-hash");
REQUIRE(repo->insertDocument(doc).has_value());
svc.__test_setMetadataRepo(repo);
RequestDispatcher::__test_setListInflightRequests(1);
auto resetGuard = ScopeExit([] { RequestDispatcher::__test_setListInflightRequests(0); });
ListRequest req;
req.limit = 1;
req.recent = false;
req.recentCount = 0;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::ResourceExhausted);
CHECK(err.message.find("List concurrency limit reached") != std::string::npos);
}
SECTION("list request catches std exceptions") {
RequestDispatcher::__test_forceListExceptionOnce("forced list exception");
ListRequest req;
req.limit = 2;
req.recent = false;
req.recentCount = 0;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "List failed: forced list exception");
}
SECTION("list request catches unknown exceptions") {
RequestDispatcher::__test_forceListUnknownExceptionOnce();
ListRequest req;
req.limit = 2;
req.recent = false;
req.recentCount = 0;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "List failed: unknown error");
}
SECTION("add document queues deferred work when system is under pressure") {
const bool originalGovernor = TuneAdvisor::enableResourceGovernor();
const bool originalAdmission = TuneAdvisor::enableAdmissionControl();
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setEnableResourceGovernor(originalGovernor);
TuneAdvisor::setEnableAdmissionControl(originalAdmission);
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Normal);
drainStoreDocumentTasks();
});
TuneAdvisor::setEnableResourceGovernor(true);
TuneAdvisor::setEnableAdmissionControl(true);
TuneAdvisor::setStoreDocumentChannelCapacity(64);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Emergency);
drainStoreDocumentTasks();
ScopedEnvVar retriesGuard("YAMS_INGEST_ENQUEUE_RETRIES", "not-a-number");
ScopedEnvVar baseDelayGuard("YAMS_INGEST_ENQUEUE_BASE_DELAY_MS", "-5");
ScopedEnvVar maxDelayGuard("YAMS_INGEST_ENQUEUE_MAX_DELAY_MS", "999999");
AddDocumentRequest req;
req.name = "pressure.txt";
req.content = "queued under pressure";
auto resp = dispatchRequest(dispatcher, Request{req});
INFO([&] {
if (std::holds_alternative<ErrorResponse>(resp)) {
return std::get<ErrorResponse>(resp).message;
}
return std::string("not an error");
}());
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
auto hasher = yams::crypto::createSHA256Hasher();
REQUIRE(hasher != nullptr);
CHECK(addResp.path == "pressure.txt");
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash == hasher->hash(req.content));
CHECK(addResp.message == "Queued for deferred processing (system under pressure).");
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask task;
REQUIRE(channel->try_pop(task));
CHECK(task.request.name == req.name);
CHECK(task.request.content == req.content);
}
SECTION("add document clears deferred hash when pressure-path hashing fails") {
const bool originalGovernor = TuneAdvisor::enableResourceGovernor();
const bool originalAdmission = TuneAdvisor::enableAdmissionControl();
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setEnableResourceGovernor(originalGovernor);
TuneAdvisor::setEnableAdmissionControl(originalAdmission);
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Normal);
drainStoreDocumentTasks();
});
TuneAdvisor::setEnableResourceGovernor(true);
TuneAdvisor::setEnableAdmissionControl(true);
TuneAdvisor::setStoreDocumentChannelCapacity(64);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Emergency);
drainStoreDocumentTasks();
auto filePath = writeTempFile("yams_documents_pressure_file_", "pressure-file.txt",
"pressure file content");
RequestDispatcher::__test_forceDocumentsHashFailureOnce();
AddDocumentRequest req;
req.path = filePath.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash.empty());
CHECK(addResp.message == "Queued for deferred processing (system under pressure).");
}
SECTION("add document clears deferred hash for recursive pressure path") {
const bool originalGovernor = TuneAdvisor::enableResourceGovernor();
const bool originalAdmission = TuneAdvisor::enableAdmissionControl();
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setEnableResourceGovernor(originalGovernor);
TuneAdvisor::setEnableAdmissionControl(originalAdmission);
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Normal);
drainStoreDocumentTasks();
});
TuneAdvisor::setEnableResourceGovernor(true);
TuneAdvisor::setEnableAdmissionControl(true);
TuneAdvisor::setStoreDocumentChannelCapacity(64);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Emergency);
drainStoreDocumentTasks();
AddDocumentRequest req;
req.path = makeTempDir("yams_documents_pressure_recursive_").string();
req.recursive = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash.empty());
CHECK(addResp.message == "Queued for deferred processing (system under pressure).");
}
SECTION("add document reports resource exhaustion when pressure queue is full") {
const bool originalGovernor = TuneAdvisor::enableResourceGovernor();
const bool originalAdmission = TuneAdvisor::enableAdmissionControl();
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setEnableResourceGovernor(originalGovernor);
TuneAdvisor::setEnableAdmissionControl(originalAdmission);
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Normal);
drainStoreDocumentTasks();
});
TuneAdvisor::setEnableResourceGovernor(true);
TuneAdvisor::setEnableAdmissionControl(true);
TuneAdvisor::setStoreDocumentChannelCapacity(64);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Emergency);
drainStoreDocumentTasks();
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask filler;
while (channel->try_push(filler)) {
}
AddDocumentRequest req;
req.name = "queue-full.txt";
req.content = "full";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::ResourceExhausted);
CHECK((err.message == "Ingestion queue is full. Please try again later." ||
err.message == "Ingestion queue is full; retry shortly"));
}
SECTION("add document queues async work while daemon is initializing") {
DaemonConfig notReadyCfg;
notReadyCfg.dataDir = makeTempDir("yams_documents_dispatcher_not_ready_");
StubLifecycle notReadyLifecycle(LifecycleState::Starting);
StateComponent notReadyState;
notReadyState.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
notReadyState.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm notReadyLifecycleFsm;
ServiceManager notReadySvc(notReadyCfg, notReadyState, notReadyLifecycleFsm);
RequestDispatcher notReadyDispatcher(&notReadyLifecycle, &notReadySvc, &notReadyState);
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
AddDocumentRequest req;
req.name = "initializing.txt";
req.content = "booting daemon";
auto resp = dispatchRequest(notReadyDispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
auto hasher = yams::crypto::createSHA256Hasher();
REQUIRE(hasher != nullptr);
CHECK(addResp.path == "initializing.txt");
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash == hasher->hash(req.content));
CHECK(addResp.message ==
"Ingestion accepted for asynchronous processing (daemon initializing).");
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask task;
REQUIRE(channel->try_pop(task));
CHECK(task.request.name == req.name);
CHECK(task.request.content == req.content);
}
SECTION("add document reports queue full while daemon is initializing") {
DaemonConfig notReadyCfg;
notReadyCfg.dataDir = makeTempDir("yams_documents_dispatcher_not_ready_full_");
StubLifecycle notReadyLifecycle(LifecycleState::Starting);
StateComponent notReadyState;
notReadyState.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
notReadyState.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm notReadyLifecycleFsm;
ServiceManager notReadySvc(notReadyCfg, notReadyState, notReadyLifecycleFsm);
RequestDispatcher notReadyDispatcher(&notReadyLifecycle, &notReadySvc, &notReadyState);
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
drainStoreDocumentTasks();
});
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
TuneAdvisor::setStoreDocumentChannelCapacity(64);
drainStoreDocumentTasks();
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask filler;
while (channel->try_push(filler)) {
}
AddDocumentRequest req;
req.name = "initializing-full.txt";
req.content = "booting daemon full";
auto resp = dispatchRequest(notReadyDispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::ResourceExhausted);
CHECK((err.message == "Ingestion queue is full. Please try again later." ||
err.message == "Ingestion queue is full; retry shortly"));
}
SECTION("add document clears async hash when daemon-initializing hashing fails") {
DaemonConfig notReadyCfg;
notReadyCfg.dataDir = makeTempDir("yams_documents_dispatcher_not_ready_hash_fail_");
StubLifecycle notReadyLifecycle(LifecycleState::Starting);
StateComponent notReadyState;
notReadyState.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
notReadyState.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm notReadyLifecycleFsm;
ServiceManager notReadySvc(notReadyCfg, notReadyState, notReadyLifecycleFsm);
RequestDispatcher notReadyDispatcher(&notReadyLifecycle, &notReadySvc, &notReadyState);
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
auto filePath =
writeTempFile("yams_documents_init_hash_fail_", "init-file.txt", "init file content");
RequestDispatcher::__test_forceDocumentsHashFailureOnce();
AddDocumentRequest req;
req.path = filePath.string();
auto resp = dispatchRequest(notReadyDispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash.empty());
CHECK(addResp.message ==
"Ingestion accepted for asynchronous processing (daemon initializing).");
}
SECTION("add document hashes file path while daemon is initializing") {
DaemonConfig notReadyCfg;
notReadyCfg.dataDir = makeTempDir("yams_documents_dispatcher_not_ready_file_hash_");
StubLifecycle notReadyLifecycle(LifecycleState::Starting);
StateComponent notReadyState;
notReadyState.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
notReadyState.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm notReadyLifecycleFsm;
ServiceManager notReadySvc(notReadyCfg, notReadyState, notReadyLifecycleFsm);
RequestDispatcher notReadyDispatcher(&notReadyLifecycle, &notReadySvc, &notReadyState);
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
auto filePath = writeTempFile("yams_documents_init_file_hash_", "init-file-hash.txt",
"init file hash content");
AddDocumentRequest req;
req.path = filePath.string();
auto resp = dispatchRequest(notReadyDispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
auto hasher = yams::crypto::createSHA256Hasher();
REQUIRE(hasher != nullptr);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash == hasher->hashFile(req.path));
CHECK(addResp.message ==
"Ingestion accepted for asynchronous processing (daemon initializing).");
}
SECTION("add document treats directory path as async directory ingestion") {
auto dirPath = makeTempDir("yams_documents_directory_ingest_");
drainStoreDocumentTasks();
AddDocumentRequest req;
req.path = dirPath.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.path == dirPath.string());
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash.empty());
CHECK(addResp.message == "Directory ingestion accepted for asynchronous processing.");
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask task;
REQUIRE(channel->try_pop(task));
CHECK(task.request.path == req.path);
}
SECTION("add document rejects missing path and content") {
AddDocumentRequest req;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message == "Provide either 'path' or 'content' + 'name'");
}
SECTION("add document uses async single-file path by default when ready") {
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
AddDocumentRequest req;
req.name = "async-default.txt";
req.content = "ready but async";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
auto hasher = yams::crypto::createSHA256Hasher();
REQUIRE(hasher != nullptr);
CHECK(addResp.path == "async-default.txt");
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash == hasher->hash(req.content));
CHECK(addResp.extractionStatus == "pending");
CHECK(addResp.message == "Ingestion accepted for asynchronous processing.");
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask task;
REQUIRE(channel->try_pop(task));
CHECK(task.request.name == req.name);
CHECK(task.request.content == req.content);
}
SECTION("add document hashes file paths on async path when ready") {
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
auto filePath =
writeTempFile("yams_documents_async_file_", "async-file.txt", "async file content");
AddDocumentRequest req;
req.path = filePath.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
auto hasher = yams::crypto::createSHA256Hasher();
REQUIRE(hasher != nullptr);
CHECK(addResp.path == req.path);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash == hasher->hashFile(req.path));
CHECK(addResp.message == "Ingestion accepted for asynchronous processing.");
}
SECTION("add document hashes file path on pressure path after one retry") {
const bool originalGovernor = TuneAdvisor::enableResourceGovernor();
const bool originalAdmission = TuneAdvisor::enableAdmissionControl();
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setEnableResourceGovernor(originalGovernor);
TuneAdvisor::setEnableAdmissionControl(originalAdmission);
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Normal);
RequestDispatcher::__test_setDocumentsEnqueueFailuresBeforeSuccess(0);
drainStoreDocumentTasks();
});
TuneAdvisor::setEnableResourceGovernor(true);
TuneAdvisor::setEnableAdmissionControl(true);
TuneAdvisor::setStoreDocumentChannelCapacity(64);
ResourceGovernor::instance().testing_updateScalingCaps(ResourcePressureLevel::Emergency);
drainStoreDocumentTasks();
ScopedEnvVar retriesGuard("YAMS_INGEST_ENQUEUE_RETRIES", "3");
ScopedEnvVar baseDelayGuard("YAMS_INGEST_ENQUEUE_BASE_DELAY_MS", "1");
ScopedEnvVar maxDelayGuard("YAMS_INGEST_ENQUEUE_MAX_DELAY_MS", "1");
RequestDispatcher::__test_setDocumentsEnqueueFailuresBeforeSuccess(1);
auto filePath = writeTempFile("yams_documents_pressure_retry_", "pressure-retry.txt",
"pressure retry content");
AddDocumentRequest req;
req.path = filePath.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
auto hasher = yams::crypto::createSHA256Hasher();
REQUIRE(hasher != nullptr);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash == hasher->hashFile(req.path));
CHECK(addResp.message == "Queued for deferred processing (system under pressure).");
}
SECTION("add document falls back to initializing path when later lifecycle snapshot throws") {
DaemonConfig notReadyCfg;
notReadyCfg.dataDir = makeTempDir("yams_documents_dispatcher_lifecycle_throw_late_");
StubLifecycle notReadyLifecycle(LifecycleState::Ready);
notReadyLifecycle.setThrowOnGetLifecycleSnapshotCall(2);
StateComponent notReadyState;
notReadyState.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
notReadyState.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm notReadyLifecycleFsm;
ServiceManager notReadySvc(notReadyCfg, notReadyState, notReadyLifecycleFsm);
RequestDispatcher notReadyDispatcher(&notReadyLifecycle, &notReadySvc, &notReadyState);
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
AddDocumentRequest req;
req.name = "late-throw.txt";
req.content = "late throw";
auto resp = dispatchRequest(notReadyDispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.message ==
"Ingestion accepted for asynchronous processing (daemon initializing).");
}
SECTION("add document reports queue full on async single-file path") {
const uint32_t originalCapacity = TuneAdvisor::storeDocumentChannelCapacity();
auto restore = ScopeExit([&] {
TuneAdvisor::setStoreDocumentChannelCapacity(originalCapacity);
drainStoreDocumentTasks();
});
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
TuneAdvisor::setStoreDocumentChannelCapacity(64);
drainStoreDocumentTasks();
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask filler;
while (channel->try_push(filler)) {
}
AddDocumentRequest req;
req.name = "async-full.txt";
req.content = "full async queue";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::ResourceExhausted);
CHECK((err.message == "Ingestion queue is full. Please try again later." ||
err.message == "Ingestion queue is full; retry shortly"));
}
SECTION("add document clears async hash when hashing fails on ready path") {
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "0");
drainStoreDocumentTasks();
auto filePath = writeTempFile("yams_documents_async_hash_fail_", "async-hash-fail.txt",
"async hash fail content");
RequestDispatcher::__test_forceDocumentsHashFailureOnce();
AddDocumentRequest req;
req.path = filePath.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.documentsAdded == 0);
CHECK(addResp.hash.empty());
CHECK(addResp.extractionStatus == "pending");
CHECK(addResp.message == "Ingestion accepted for asynchronous processing.");
}
SECTION("add document uses sync fallback when explicitly enabled") {
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "1");
drainStoreDocumentTasks();
svc.__test_setContentStore(std::make_shared<StubContentStore>());
AddDocumentRequest req;
req.name = "sync-add.txt";
req.content = "sync path content";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.path == "sync-add.txt");
CHECK(addResp.documentsAdded == 1);
CHECK(addResp.size == req.content.size());
CHECK(addResp.extractionStatus == "pending");
CHECK(addResp.message == "Document stored successfully.");
CHECK_FALSE(addResp.hash.empty());
auto channel = getStoreDocumentTaskChannel();
InternalEventBus::StoreDocumentTask task;
CHECK_FALSE(channel->try_pop(task));
}
SECTION("add document sync fallback forwards metadata and path inputs") {
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "true");
drainStoreDocumentTasks();
svc.__test_setContentStore(std::make_shared<StubContentStore>());
auto filePath =
writeTempFile("yams_documents_sync_file_", "sync-file.txt", "sync file content");
AddDocumentRequest req;
req.path = filePath.string();
req.mimeType = "text/custom";
req.disableAutoMime = true;
req.tags = {"alpha", "beta"};
req.metadata["task"] = "documents";
req.collection = "focus";
req.snapshotId = "snap-1";
req.snapshotLabel = "label-1";
req.sessionId = "session-1";
req.noEmbeddings = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<AddDocumentResponse>(resp));
const auto& addResp = std::get<AddDocumentResponse>(resp);
CHECK(addResp.path == req.path);
CHECK(addResp.documentsAdded == 1);
CHECK(addResp.size == std::filesystem::file_size(filePath));
CHECK_FALSE(addResp.hash.empty());
}
SECTION("add document sync fallback returns store errors") {
ScopedEnvVar syncGuard("YAMS_SYNC_SINGLE_FILE_ADD", "1");
drainStoreDocumentTasks();
svc.__test_setContentStore(std::shared_ptr<api::IContentStore>{});
AddDocumentRequest req;
req.name = "sync-error.txt";
req.content = "sync failure";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message == "Content store not available");
}
SECTION("file history skips documents when metadata lookup fails") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto failingDoc = makeDoc(57, "/tmp/archive/history-mixed.txt", "metadata-fail-hash");
auto okDoc = makeDoc(58, "/var/tmp/archive/history-mixed.txt", "metadata-ok-hash");
repo->addDocument(failingDoc);
repo->addDocument(okDoc);
repo->setAllMetadataError(failingDoc.id,
Error{ErrorCode::DatabaseError, "metadata lookup failed"});
repo->setAllMetadata(okDoc.id,
{{"snapshot_id", metadata::MetadataValue("snap-ok")},
{"snapshot_time", metadata::MetadataValue("1710000000000000")}});
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = "history-mixed.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
REQUIRE(history.found);
CHECK(history.totalVersions == 1);
REQUIRE(history.versions.size() == 1);
CHECK(history.versions.front().snapshotId == "snap-ok");
}
SECTION("file history falls back when exact path lookup errors") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto doc =
makeDoc(63, "/tmp/archive/fallback-exact-error.txt", "fallback-exact-error-hash");
repo->addDocument(doc);
repo->setExactPathError(Error{ErrorCode::DatabaseError, "exact lookup failed"});
repo->setAllMetadata(doc.id,
{{"snapshot_id", metadata::MetadataValue("snap-fallback-error")}});
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = doc.filePath;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
REQUIRE(history.found);
CHECK(history.totalVersions == 1);
REQUIRE(history.versions.size() == 1);
CHECK(history.versions.front().snapshotId == "snap-fallback-error");
}
SECTION("file history caps oversized filename matches to first hundred documents") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
for (int i = 0; i < 101; ++i) {
auto doc = makeDoc(1000 + i,
"/tmp/archive/subdir_" + std::to_string(i) + "/history-overflow.txt",
"history-overflow-hash-" + std::to_string(i));
doc.fileName = "history-overflow.txt";
repo->addDocument(doc);
repo->setAllMetadata(
doc.id, {{"snapshot_id", metadata::MetadataValue("snap-" + std::to_string(i))}});
}
svc.__test_setMetadataRepo(repo);
FileHistoryRequest req;
req.filepath = "history-overflow.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<FileHistoryResponse>(resp));
const auto& history = std::get<FileHistoryResponse>(resp);
REQUIRE(history.found);
CHECK(history.totalVersions == 100);
CHECK(history.versions.size() == 100);
}
SECTION("file history reports invalid filepath when current directory is missing") {
#ifdef _WIN32
SKIP("Windows keeps the current working directory locked, so missing-cwd simulation is "
"unreliable");
#endif
auto repo = std::make_shared<StubPruneMetadataRepository>();
svc.__test_setMetadataRepo(repo);
auto originalCwd = std::filesystem::current_path();
auto doomedDir = makeTempDir("yams_missing_cwd_");
std::filesystem::current_path(doomedDir);
std::filesystem::remove_all(doomedDir);
FileHistoryRequest req;
req.filepath = "history-from-missing-cwd.txt";
auto resp = dispatchRequest(dispatcher, Request{req});
std::filesystem::current_path(originalCwd);
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message.find("Invalid filepath") != std::string::npos);
}
SECTION("cat request reports missing content") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(58, "/tmp/cat/missing.txt", "cat-missing-hash");
repo->addDocument(doc);
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
CatRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message.find("Document not found") != std::string::npos);
}
SECTION("cat request reports missing content payload for indexed document") {
auto repoPath = makeTempMetadataDbPath("yams_cat_missing_content_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(59, "/tmp/cat/metadata-only.txt",
"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa");
REQUIRE(repo->insertDocument(doc).has_value());
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
CatRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message.find("Document content not found") != std::string::npos);
}
SECTION("cat request hits missing document branch after successful retrieval") {
auto repoPath = makeTempMetadataDbPath("yams_cat_force_missing_doc_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(61, "/tmp/cat/force-missing-doc.txt",
"bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "payload");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher::__test_forceCatMissingDocumentOnce();
CatRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "Document not found");
}
SECTION("cat request hits content unavailable branch after successful retrieval") {
auto repoPath = makeTempMetadataDbPath("yams_cat_force_missing_content_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(62, "/tmp/cat/force-missing-content.txt",
"cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "payload");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher::__test_forceCatMissingContentOnce();
CatRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Content unavailable");
}
SECTION("cat request hits native missing document branch") {
auto repoPath = makeTempMetadataDbPath("yams_cat_native_missing_doc_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(65, "/tmp/cat/native-missing-doc.txt",
"dddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "payload");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher::__test_forceCatNativeMissingDocumentOnce();
CatRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "Document not found");
}
SECTION("cat request hits native missing content branch") {
auto repoPath = makeTempMetadataDbPath("yams_cat_native_missing_content_repo_");
metadata::ConnectionPoolConfig poolCfg{};
auto pool = std::make_shared<metadata::ConnectionPool>(repoPath.string(), poolCfg);
REQUIRE(pool->initialize().has_value());
auto repo = makeReadyMetadataRepo(pool);
auto store = std::make_shared<StubContentStore>();
auto doc = makeDoc(66, "/tmp/cat/native-missing-content.txt",
"eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee");
REQUIRE(repo->insertDocument(doc).has_value());
store->setBlob(doc.sha256Hash, "payload");
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(store);
RequestDispatcher::__test_forceCatNativeMissingContentOnce();
CatRequest req;
req.hash = doc.sha256Hash;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Content unavailable");
}
}
TEST_CASE("RequestDispatcher: download handler enforces daemon policy",
"[daemon][documents][dispatcher][download]") {
ScopedEnvVar daemonDownloadEnv("YAMS_ENABLE_DAEMON_DOWNLOAD", "0");
auto dispatchDownload = [](RequestDispatcher& dispatcher, const std::string& url,
std::string outputPath = {}, std::string checksum = {}) {
DownloadRequest req;
req.url = url;
req.outputPath = std::move(outputPath);
req.checksum = std::move(checksum);
return dispatchRequest(dispatcher, Request{req});
};
auto waitForDownloadJob = [](RequestDispatcher& dispatcher, const std::string& jobId,
std::chrono::milliseconds timeout = std::chrono::seconds(2)) {
const auto deadline = std::chrono::steady_clock::now() + timeout;
DownloadResponse latest;
while (std::chrono::steady_clock::now() < deadline) {
auto resp = dispatchRequest(dispatcher, Request{DownloadStatusRequest{jobId}});
if (std::holds_alternative<DownloadResponse>(resp)) {
latest = std::get<DownloadResponse>(resp);
if (latest.state != "queued" && latest.state != "running") {
return latest;
}
}
std::this_thread::sleep_for(std::chrono::milliseconds(10));
}
return latest;
};
auto waitForFollowRedirectsCapture = [](std::chrono::milliseconds timeout =
std::chrono::seconds(2)) {
const auto deadline = std::chrono::steady_clock::now() + timeout;
std::optional<bool> latest;
while (std::chrono::steady_clock::now() < deadline) {
latest = RequestDispatcher::__test_lastDownloadServiceFollowRedirects();
if (latest.has_value()) {
return latest;
}
std::this_thread::sleep_for(std::chrono::milliseconds(10));
}
return latest;
};
RequestDispatcher::__test_resetDownloadServiceRequestCapture();
SECTION("disabled policy returns safe reminder") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_disabled_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("Daemon download is disabled") != std::string::npos);
}
SECTION("disallowed scheme is rejected before download starts") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_scheme_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "http://example.com/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("scheme 'http'") != std::string::npos);
}
SECTION("invalid URL shape is rejected before policy host checks") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_invalid_url_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "not-a-url");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error == "Invalid download URL");
}
SECTION("download URL rejects missing authority after scheme") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_missing_authority_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error == "Invalid download URL");
}
SECTION("download URL rejects empty host after credentials stripping") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_empty_host_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://user@/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error == "Invalid download URL");
}
SECTION("wildcard host policy accepts matching subdomains before checksum gate") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_wildcard_host_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = true;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*.example.com"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://api.example.com/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("checksum is required") != std::string::npos);
}
SECTION("IPv6 authority parses and matches exact allowed host") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_ipv6_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = true;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"::1"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://[::1]/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("checksum is required") != std::string::npos);
}
SECTION("invalid IPv6 authority is rejected") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_ipv6_invalid_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://[]/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error == "Invalid download URL");
}
SECTION("disallowed host is rejected before download starts") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_host_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"downloads.example.com"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("host 'example.com'") != std::string::npos);
}
SECTION("store-only policy rejects output path") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_store_only_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.storeOnly = true;
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin", "/tmp/export.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("store_only=true") != std::string::npos);
}
SECTION("invalid checksum format is rejected before download starts") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_bad_checksum_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp =
dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha1:not-allowed");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("invalid checksum format") != std::string::npos);
}
SECTION("unsupported checksum algorithm with valid hex is rejected") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_bad_checksum_algo_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp =
dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha1:abcdef12");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("invalid checksum format") != std::string::npos);
}
SECTION("malformed checksum missing separator is rejected") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_bad_checksum_shape_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha256");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("invalid checksum format") != std::string::npos);
}
SECTION("checksum-required policy rejects missing checksum") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_checksum_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = true;
cfg.downloadPolicy.allowedSchemes = {"https"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.error.find("checksum is required") != std::string::npos);
}
SECTION("valid checksum enters job tracking and handles unavailable service") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_valid_checksum_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
RequestDispatcher::__test_forceDownloadServiceUnavailableOnce();
auto resp =
dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha256:abcd1234");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK_FALSE(dl.success);
CHECK(dl.state == "queued");
CHECK_FALSE(dl.jobId.empty());
const auto finalStatus = waitForDownloadJob(dispatcher, dl.jobId);
CHECK(finalStatus.jobId == dl.jobId);
CHECK(finalStatus.state == "failed");
CHECK(finalStatus.error == "Download service not available in daemon");
}
SECTION("sha512 checksum enters job tracking and handles unavailable service") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_sha512_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
RequestDispatcher::__test_forceDownloadServiceUnavailableOnce();
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin", {},
"sha512:abcdef1234567890");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.state == "queued");
CHECK_FALSE(dl.jobId.empty());
const auto finalStatus = waitForDownloadJob(dispatcher, dl.jobId);
CHECK(finalStatus.jobId == dl.jobId);
CHECK(finalStatus.state == "failed");
CHECK(finalStatus.error == "Download service not available in daemon");
}
SECTION("md5 checksum enters job tracking and handles unavailable service") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_md5_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
RequestDispatcher::__test_forceDownloadServiceUnavailableOnce();
auto resp = dispatchDownload(dispatcher, "https://example.com/file.bin", {},
"md5:abcdef1234567890");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.state == "queued");
CHECK_FALSE(dl.jobId.empty());
const auto finalStatus = waitForDownloadJob(dispatcher, dl.jobId);
CHECK(finalStatus.jobId == dl.jobId);
CHECK(finalStatus.state == "failed");
CHECK(finalStatus.error == "Download service not available in daemon");
}
SECTION("forced successful download completes the tracked job") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_success_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
RequestDispatcher::__test_forceDownloadServiceSuccessOnce();
auto resp =
dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha256:abcd1234");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.state == "queued");
CHECK_FALSE(dl.jobId.empty());
const auto finalStatus = waitForDownloadJob(dispatcher, dl.jobId);
CHECK(finalStatus.jobId == dl.jobId);
CHECK(finalStatus.state == "completed");
CHECK(finalStatus.success);
CHECK(finalStatus.error.empty());
CHECK(finalStatus.hash == "sha256:forced-download-success");
CHECK(finalStatus.localPath.find("forced-download-success.bin") != std::string::npos);
CHECK(finalStatus.size == 64);
}
SECTION("daemon download disables redirects and surfaces blocked redirect failures") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_policy_redirect_blocked_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*"};
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
RequestDispatcher::__test_forceDownloadServiceFailureOnce(
"Redirect blocked by policy during probe (follow_redirects=false)");
auto resp =
dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha256:abcd1234");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.state == "queued");
CHECK_FALSE(dl.jobId.empty());
const auto lastFollowRedirects = waitForFollowRedirectsCapture();
REQUIRE(lastFollowRedirects.has_value());
CHECK_FALSE(*lastFollowRedirects);
const auto finalStatus = waitForDownloadJob(dispatcher, dl.jobId);
CHECK(finalStatus.jobId == dl.jobId);
CHECK(finalStatus.state == "failed");
CHECK(finalStatus.error ==
"Redirect blocked by policy during probe (follow_redirects=false)");
}
}
TEST_CASE("RequestDispatcher: download job status and list use in-memory registry",
"[daemon][documents][dispatcher][download-jobs]") {
RequestDispatcher::__test_resetDownloadJobs();
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_job_registry_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
DownloadResponse first;
first.url = "https://example.com/a.bin";
first.jobId = "download-1";
first.state = "completed";
first.createdAtMs = 100;
first.updatedAtMs = 200;
first.success = true;
DownloadResponse second;
second.url = "https://example.com/b.bin";
second.jobId = "download-2";
second.state = "failed";
second.createdAtMs = 300;
second.updatedAtMs = 400;
second.success = false;
second.error = "network boom";
RequestDispatcher::__test_seedDownloadJob(first);
RequestDispatcher::__test_seedDownloadJob(second);
SECTION("download status returns the seeded job") {
auto resp = dispatchRequest(dispatcher, Request{DownloadStatusRequest{"download-2"}});
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.jobId == "download-2");
CHECK(dl.state == "failed");
CHECK(dl.error == "network boom");
}
SECTION("download status returns not found for unknown job") {
auto resp = dispatchRequest(dispatcher, Request{DownloadStatusRequest{"missing-job"}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "Download job not found");
}
SECTION("list download jobs returns most recent first and honors limit") {
auto resp = dispatchRequest(dispatcher, Request{ListDownloadJobsRequest{1}});
REQUIRE(std::holds_alternative<ListDownloadJobsResponse>(resp));
const auto& jobs = std::get<ListDownloadJobsResponse>(resp);
REQUIRE(jobs.jobs.size() == 1);
CHECK(jobs.jobs.front().jobId == "download-2");
CHECK(jobs.jobs.front().state == "failed");
}
SECTION("list download jobs clamps zero limit to one") {
auto resp = dispatchRequest(dispatcher, Request{ListDownloadJobsRequest{0}});
REQUIRE(std::holds_alternative<ListDownloadJobsResponse>(resp));
const auto& jobs = std::get<ListDownloadJobsResponse>(resp);
REQUIRE(jobs.jobs.size() == 1);
CHECK(jobs.jobs.front().jobId == "download-2");
}
SECTION("cancel download job marks running job as canceled") {
DownloadResponse running;
running.url = "https://example.com/c.bin";
running.jobId = "download-3";
running.state = "running";
running.createdAtMs = 500;
running.updatedAtMs = 500;
running.success = false;
RequestDispatcher::__test_seedDownloadJob(running);
auto resp = dispatchRequest(dispatcher, Request{CancelDownloadJobRequest{"download-3"}});
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.jobId == "download-3");
CHECK(dl.state == "canceled");
CHECK(dl.error == "Canceled by client");
CHECK_FALSE(dl.success);
}
SECTION("cancel download job keeps terminal jobs terminal") {
auto resp = dispatchRequest(dispatcher, Request{CancelDownloadJobRequest{"download-2"}});
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
CHECK(dl.jobId == "download-2");
CHECK(dl.state == "failed");
CHECK(dl.error == "network boom");
}
SECTION("cancel download job returns not found for unknown job") {
auto resp = dispatchRequest(dispatcher, Request{CancelDownloadJobRequest{"missing-job"}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "Download job not found");
}
}
TEST_CASE("RequestDispatcher: download jobs reload from persisted registry",
"[daemon][documents][dispatcher][download-jobs]") {
RequestDispatcher::__test_resetDownloadJobs();
auto dispatchDownload = [](RequestDispatcher& dispatcher, const std::string& url,
std::string outputPath = {}, std::string checksum = {}) {
DownloadRequest req;
req.url = url;
req.outputPath = std::move(outputPath);
req.checksum = std::move(checksum);
return dispatchRequest(dispatcher, Request{req});
};
auto waitForDownloadJob = [](RequestDispatcher& dispatcher, const std::string& jobId,
std::chrono::milliseconds timeout = std::chrono::seconds(2)) {
const auto deadline = std::chrono::steady_clock::now() + timeout;
DownloadResponse latest;
while (std::chrono::steady_clock::now() < deadline) {
auto resp = dispatchRequest(dispatcher, Request{DownloadStatusRequest{jobId}});
if (std::holds_alternative<DownloadResponse>(resp)) {
latest = std::get<DownloadResponse>(resp);
if (latest.state != "queued" && latest.state != "running") {
return latest;
}
}
std::this_thread::sleep_for(std::chrono::milliseconds(10));
}
return latest;
};
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_download_job_registry_persist_");
cfg.downloadPolicy.enable = true;
cfg.downloadPolicy.requireChecksum = false;
cfg.downloadPolicy.allowedSchemes = {"https"};
cfg.downloadPolicy.allowedHosts = {"*"};
std::string jobId;
{
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
RequestDispatcher::__test_forceDownloadServiceUnavailableOnce();
auto resp =
dispatchDownload(dispatcher, "https://example.com/file.bin", {}, "sha256:abcd1234");
REQUIRE(std::holds_alternative<DownloadResponse>(resp));
const auto& dl = std::get<DownloadResponse>(resp);
REQUIRE_FALSE(dl.jobId.empty());
CHECK_FALSE(dl.success);
CHECK(dl.state == "queued");
jobId = dl.jobId;
const auto finalStatus = waitForDownloadJob(dispatcher, jobId);
CHECK(finalStatus.state == "failed");
CHECK(finalStatus.error == "Download service not available in daemon");
}
RequestDispatcher::__test_forgetDownloadJobsCache();
{
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
auto statusResp = dispatchRequest(dispatcher, Request{DownloadStatusRequest{jobId}});
REQUIRE(std::holds_alternative<DownloadResponse>(statusResp));
const auto& status = std::get<DownloadResponse>(statusResp);
CHECK(status.jobId == jobId);
CHECK(status.state == "failed");
CHECK(status.error == "Download service not available in daemon");
auto listResp = dispatchRequest(dispatcher, Request{ListDownloadJobsRequest{1}});
REQUIRE(std::holds_alternative<ListDownloadJobsResponse>(listResp));
const auto& list = std::get<ListDownloadJobsResponse>(listResp);
REQUIRE(list.jobs.size() == 1);
CHECK(list.jobs.front().jobId == jobId);
CHECK(list.jobs.front().state == "failed");
}
}
TEST_CASE("RequestDispatcher: embedding handlers cover generation and repair branches",
"[daemon][embedding][dispatcher]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_embedding_dispatcher_");
cfg.configFilePath = cfg.dataDir / "missing-config.toml";
ScopedEnvVar preferredModelGuard("YAMS_PREFERRED_MODEL", std::nullopt);
ScopedEnvVar embedBackendGuard("YAMS_EMBED_BACKEND", std::nullopt);
ScopedEnvVar configPathGuard("YAMS_CONFIG", cfg.configFilePath.string());
ScopedEnvVar xdgConfigHomeGuard("XDG_CONFIG_HOME", cfg.dataDir.string());
ScopedEnvVar homeGuard("HOME", cfg.dataDir.string());
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
SECTION("generate embedding marks provider unavailable as degraded") {
auto resp = dispatchRequest(
dispatcher, Request{GenerateEmbeddingRequest{"hello world", "embed-model", true}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message.find("Plugin system") != std::string::npos);
CHECK(lifecycle.setSubsystemDegradedCalls() == 1);
CHECK(lifecycle.lastSubsystem() == "embedding");
CHECK(lifecycle.lastDegraded());
CHECK(lifecycle.lastReason() == "provider_unavailable");
}
SECTION("generate embedding auto-loads unloaded model and returns vector") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/embed-model.onnx");
provider->setSingleEmbeddingResult({0.1f, 0.2f, 0.3f});
svc.__test_setModelProvider(provider);
GenerateEmbeddingRequest req{"hello world", "embed-model", true};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<EmbeddingResponse>(resp));
const auto& embedResp = std::get<EmbeddingResponse>(resp);
CHECK(embedResp.embedding == std::vector<float>{0.1f, 0.2f, 0.3f});
CHECK(embedResp.dimensions == 3);
CHECK(embedResp.modelUsed == "embed-model");
CHECK(provider->isModelLoaded("embed-model"));
}
SECTION("generate embedding forwards generator errors") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/embed-error.onnx");
provider->setGenerateEmbeddingError(
Error{ErrorCode::InvalidState, "generator unavailable"});
svc.__test_setModelProvider(provider);
GenerateEmbeddingRequest req{"hello world", "embed-error", true};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "generator unavailable");
}
SECTION("generate embedding sanitizes utf8 in load failures") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/embed-utf8.onnx");
std::string invalidMessage = "prefix ";
invalidMessage.push_back(static_cast<char>(0xC2));
invalidMessage.push_back(static_cast<char>(0xA9));
invalidMessage.push_back(static_cast<char>(0xE2));
invalidMessage.push_back(static_cast<char>(0x82));
invalidMessage.push_back(static_cast<char>(0xAC));
invalidMessage.push_back(static_cast<char>(0xF0));
invalidMessage.push_back(static_cast<char>(0x9F));
invalidMessage.push_back(static_cast<char>(0x92));
invalidMessage.push_back(static_cast<char>(0xA9));
invalidMessage.push_back(static_cast<char>(0xC0));
invalidMessage.push_back(static_cast<char>(0xFF));
provider->setLoadError(Error{ErrorCode::InvalidState, invalidMessage});
svc.__test_setModelProvider(provider);
GenerateEmbeddingRequest req{"hello world", "embed-utf8", true};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
std::string expected = "prefix ";
expected.push_back(static_cast<char>(0xC2));
expected.push_back(static_cast<char>(0xA9));
expected.push_back(static_cast<char>(0xE2));
expected.push_back(static_cast<char>(0x82));
expected.push_back(static_cast<char>(0xAC));
expected.push_back(static_cast<char>(0xF0));
expected.push_back(static_cast<char>(0x9F));
expected.push_back(static_cast<char>(0x92));
expected.push_back(static_cast<char>(0xA9));
expected += "\xEF\xBF\xBD\xEF\xBF\xBD";
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == expected);
}
SECTION("generate embedding honors timeout env floor while auto-loading") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/embed-timeout.onnx");
provider->setSingleEmbeddingResult({0.7f, 0.8f, 0.9f});
svc.__test_setModelProvider(provider);
ScopedEnvVar timeoutGuard("YAMS_MODEL_LOAD_TIMEOUT_MS", "5");
GenerateEmbeddingRequest req{"hello world", "embed-timeout", true};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<EmbeddingResponse>(resp));
const auto& embedResp = std::get<EmbeddingResponse>(resp);
CHECK(embedResp.embedding == std::vector<float>{0.7f, 0.8f, 0.9f});
CHECK(provider->isModelLoaded("embed-timeout"));
}
SECTION("generate embedding converts isModelLoaded exceptions into internal errors") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/embed-throw.onnx");
provider->setThrowOnIsModelLoadedCount(2);
svc.__test_setModelProvider(provider);
GenerateEmbeddingRequest req{"hello world", "embed-throw", true};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Embedding generation failed: isModelLoaded boom") !=
std::string::npos);
}
SECTION("batch embedding falls back to per-item generation") {
auto provider = std::make_shared<StubModelProvider>(2, "/tmp/batch-fallback.onnx");
provider->setGenerateBatchError(Error{ErrorCode::NotImplemented, "batch unsupported"});
provider->setSingleEmbeddingResult({0.4f, 0.6f});
svc.__test_setModelProvider(provider);
BatchEmbeddingRequest req;
req.texts = {"one", "two"};
req.modelName = "batch-fallback";
req.normalize = false;
req.batchSize = 2;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<BatchEmbeddingResponse>(resp));
const auto& batchResp = std::get<BatchEmbeddingResponse>(resp);
REQUIRE(batchResp.embeddings.size() == 2);
CHECK(batchResp.embeddings[0] == std::vector<float>{0.4f, 0.6f});
CHECK(batchResp.embeddings[1] == std::vector<float>{0.4f, 0.6f});
CHECK(batchResp.successCount == 2);
CHECK(batchResp.failureCount == 0);
CHECK(batchResp.dimensions == 2);
CHECK(batchResp.modelUsed == "batch-fallback");
}
SECTION("batch embedding counts per-item failures during fallback") {
auto provider = std::make_shared<StubModelProvider>(2, "/tmp/batch-partial.onnx");
provider->setGenerateBatchError(Error{ErrorCode::NotImplemented, "batch unsupported"});
provider->setSingleEmbeddingResult({0.9f, 0.1f});
provider->setSingleTextError("bad", Error{ErrorCode::InternalError, "single failed"});
svc.__test_setModelProvider(provider);
BatchEmbeddingRequest req;
req.texts = {"good", "bad", "good-again"};
req.modelName = "batch-partial";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<BatchEmbeddingResponse>(resp));
const auto& batchResp = std::get<BatchEmbeddingResponse>(resp);
REQUIRE(batchResp.embeddings.size() == 2);
CHECK(batchResp.successCount == 2);
CHECK(batchResp.failureCount == 1);
CHECK(batchResp.dimensions == 2);
CHECK(batchResp.modelUsed == "batch-partial");
}
SECTION("batch embedding forwards non-fallback errors") {
auto provider = std::make_shared<StubModelProvider>(2, "/tmp/batch-error.onnx");
provider->setGenerateBatchError(Error{ErrorCode::InvalidState, "batch failed"});
svc.__test_setModelProvider(provider);
BatchEmbeddingRequest req;
req.texts = {"one", "two"};
req.modelName = "batch-error";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "batch failed");
}
SECTION("batch embedding auto-loads model and returns direct batch results") {
auto provider = std::make_shared<StubModelProvider>(2, "/tmp/batch-direct.onnx");
provider->setBatchEmbeddingResult({{0.2f, 0.8f}});
svc.__test_setModelProvider(provider);
ScopedEnvVar timeoutGuard("YAMS_MODEL_LOAD_TIMEOUT_MS", "not-a-number");
BatchEmbeddingRequest req;
req.texts = {"one", "two"};
req.modelName = "batch-direct";
req.normalize = true;
req.batchSize = 2;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<BatchEmbeddingResponse>(resp));
const auto& batchResp = std::get<BatchEmbeddingResponse>(resp);
REQUIRE(batchResp.embeddings.size() == 1);
CHECK(batchResp.embeddings[0] == std::vector<float>{0.2f, 0.8f});
CHECK(batchResp.successCount == 1);
CHECK(batchResp.failureCount == 1);
CHECK(batchResp.dimensions == 2);
CHECK(batchResp.modelUsed == "batch-direct");
CHECK(provider->isModelLoaded("batch-direct"));
}
SECTION("batch embedding returns load failures for unloaded models") {
auto provider = std::make_shared<StubModelProvider>(2, "/tmp/batch-load-fail.onnx");
provider->setLoadError(Error{ErrorCode::InvalidState, "batch load refused"});
svc.__test_setModelProvider(provider);
ScopedEnvVar timeoutGuard("YAMS_MODEL_LOAD_TIMEOUT_MS", "5");
BatchEmbeddingRequest req;
req.texts = {"one", "two"};
req.modelName = "batch-load-fail";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "batch load refused");
}
SECTION("batch embedding converts isModelLoaded exceptions into internal errors") {
auto provider = std::make_shared<StubModelProvider>(2, "/tmp/batch-throw.onnx");
provider->setThrowOnIsModelLoadedCount(2);
svc.__test_setModelProvider(provider);
BatchEmbeddingRequest req;
req.texts = {"one"};
req.modelName = "batch-throw";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Batch embedding failed: isModelLoaded boom") != std::string::npos);
}
SECTION("embed documents rejects missing service manager") {
RequestDispatcher noServiceDispatcher(&lifecycle, nullptr, &state);
auto resp = dispatchRequest(noServiceDispatcher, Request{EmbedDocumentsRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message == "ServiceManager not available");
}
SECTION("embed documents reports unavailable model provider") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-provider.onnx");
provider->setAvailable(false);
svc.__test_setModelProvider(provider);
EmbedDocumentsRequest req;
req.modelName = "repair-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message == "Model provider not available");
}
SECTION("embed documents returns success when repair finds no work") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-empty.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
svc.__test_setMetadataRepo(std::make_shared<StubPruneMetadataRepository>());
EmbedDocumentsRequest req;
req.modelName = "repair-model";
req.batchSize = 4;
req.skipExisting = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<EmbedDocumentsResponse>(resp));
const auto& embedResp = std::get<EmbedDocumentsResponse>(resp);
CHECK(embedResp.requested == 0);
CHECK(embedResp.embedded == 0);
CHECK(embedResp.skipped == 0);
CHECK(embedResp.failed == 0);
}
SECTION("embed documents rejects non-empty hash lists that resolve to no daemon documents") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-missing-hashes.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
svc.__test_setMetadataRepo(std::make_shared<StubPruneMetadataRepository>());
EmbedDocumentsRequest req;
req.modelName = "repair-model";
req.documentHashes = {std::string(64, 'a')};
req.batchSize = 4;
req.skipExisting = false;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message.find("requested documents were found") != std::string::npos);
}
SECTION("embed documents rejects degraded providers") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-degraded.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setModelProviderDegraded(true, "simulated degraded state");
auto resp = dispatchRequest(dispatcher, Request{EmbedDocumentsRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "Embedding generation disabled: provider degraded");
svc.__test_setModelProviderDegraded(false);
}
SECTION("embed documents reports missing content store") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-content.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setMetadataRepo(std::make_shared<StubPruneMetadataRepository>());
EmbedDocumentsRequest req;
req.modelName = "repair-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message == "Content store not available");
}
SECTION("embed documents reports missing metadata repo") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-meta.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
EmbedDocumentsRequest req;
req.modelName = "repair-model";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message == "Metadata repository not available");
}
SECTION("embed documents reports missing configured model name") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-empty-model.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
svc.__test_setMetadataRepo(std::make_shared<StubPruneMetadataRepository>());
EmbedDocumentsRequest req;
req.modelName = "";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotInitialized);
CHECK(err.message == "No embedding model configured");
}
SECTION("embed documents uses daemon config resolver when request model is empty") {
auto provider = std::make_shared<StubModelProvider>(3, "/tmp/repair-config-model.onnx");
svc.__test_setModelProvider(provider);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
svc.__test_setMetadataRepo(std::make_shared<StubPruneMetadataRepository>());
const auto configPath = cfg.dataDir / "config.toml";
{
std::ofstream out(configPath);
REQUIRE(out.is_open());
out << "[embeddings]\n";
out << "preferred_model = \"config-repair-model\"\n";
}
cfg.configFilePath = configPath;
ScopedEnvVar sectionConfigPathGuard("YAMS_CONFIG", configPath.string());
DaemonLifecycleFsm configLifecycleFsm;
ServiceManager configSvc(cfg, state, configLifecycleFsm);
configSvc.__test_setModelProvider(provider);
configSvc.__test_setContentStore(std::make_shared<StubContentStore>());
configSvc.__test_setMetadataRepo(std::make_shared<StubPruneMetadataRepository>());
RequestDispatcher configDispatcher(&lifecycle, &configSvc, &state);
EmbedDocumentsRequest req;
req.modelName.clear();
req.batchSize = 4;
req.skipExisting = true;
auto resp = dispatchRequest(configDispatcher, Request{req});
REQUIRE(std::holds_alternative<EmbedDocumentsResponse>(resp));
CHECK(provider->isModelLoaded("config-repair-model"));
}
}
TEST_CASE("RequestDispatcher: graph maintenance handlers report missing graph component",
"[daemon][graph][dispatcher]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_graph_dispatcher_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
SECTION("graph repair returns internal error when graph component is unavailable") {
GraphRepairRequest req;
req.dryRun = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("GraphComponent not available") != std::string::npos);
}
SECTION("graph validate returns internal error when graph component is unavailable") {
auto resp = dispatchRequest(dispatcher, Request{GraphValidateRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("GraphComponent not available") != std::string::npos);
}
}
TEST_CASE("RequestDispatcher: graph query and ingest handlers cover dispatcher branches",
"[daemon][graph][dispatcher][query]") {
GraphDispatcherFixture fixture;
SECTION("graph query reports unavailable metadata repository") {
auto resp = dispatchRequest(*fixture.dispatcher, Request{GraphQueryRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Metadata repository unavailable");
}
SECTION("graph query reports unavailable knowledge graph store") {
fixture.initMetadata(false);
auto resp = dispatchRequest(*fixture.dispatcher, Request{GraphQueryRequest{}});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK_FALSE(graphResp.kgAvailable);
CHECK(graphResp.warning == "Knowledge graph not available");
}
SECTION("graph path history reports unavailable metadata repository") {
GraphPathHistoryRequest req;
req.path = "/repo/file.cpp";
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Metadata repository unavailable");
}
SECTION("graph explore reports unavailable metadata repository") {
auto resp =
dispatchRequest(*fixture.dispatcher, Request{GraphExploreRequest{.query = "entry"}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Metadata repository unavailable");
}
SECTION("graph explore reports unavailable knowledge graph store") {
fixture.initMetadata(false);
auto resp =
dispatchRequest(*fixture.dispatcher, Request{GraphExploreRequest{.query = "entry"}});
REQUIRE(std::holds_alternative<GraphExploreResponse>(resp));
const auto& graphResp = std::get<GraphExploreResponse>(resp);
CHECK(graphResp.query == "entry");
CHECK_FALSE(graphResp.kgAvailable);
REQUIRE(graphResp.warnings.size() == 1);
CHECK(graphResp.warnings.front() == "Knowledge graph not available");
}
SECTION("graph query traversal requires an origin") {
fixture.initMetadata();
auto resp = dispatchRequest(*fixture.dispatcher, Request{GraphQueryRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message.find("No valid origin specified") != std::string::npos);
}
SECTION("graph explore returns full and omitted snippet modes through dispatcher") {
fixture.initMetadata();
const auto sourcePath = fixture.testDir / "src" / "fallback.cpp";
std::filesystem::create_directories(sourcePath.parent_path());
{
std::ofstream out(sourcePath);
REQUIRE(out.good());
out << "int fallbackEntry() {\n return 7;\n}\n";
}
REQUIRE(
fixture.repo->insertDocument(makeGraphDispatcherDocument(sourcePath, "hash-fallback"))
.has_value());
metadata::KGNode node;
node.nodeKey = "function:demo::fallbackEntry@" + sourcePath.string();
node.label = "fallbackEntry";
node.type = "function";
REQUIRE(fixture.kgStore->upsertNode(node).has_value());
GraphExploreRequest fullReq;
fullReq.query = "fallbackEntry";
auto fullRespRaw = dispatchRequest(*fixture.dispatcher, Request{fullReq});
REQUIRE(std::holds_alternative<GraphExploreResponse>(fullRespRaw));
const auto& fullResp = std::get<GraphExploreResponse>(fullRespRaw);
REQUIRE(fullResp.files.size() == 1);
CHECK(fullResp.files.front().mode == "full");
CHECK(fullResp.files.front().content.find("fallbackEntry") != std::string::npos);
CHECK(fullResp.entrySymbols.size() == 1);
CHECK(fullResp.entrySymbols.front().label == "fallbackEntry");
CHECK(fullResp.snippetsRendered == 1);
GraphExploreRequest omittedReq;
omittedReq.query = "fallbackEntry";
omittedReq.includeCode = false;
auto omittedRespRaw = dispatchRequest(*fixture.dispatcher, Request{omittedReq});
REQUIRE(std::holds_alternative<GraphExploreResponse>(omittedRespRaw));
const auto& omittedResp = std::get<GraphExploreResponse>(omittedRespRaw);
REQUIRE(omittedResp.files.size() == 1);
CHECK(omittedResp.files.front().mode == "omitted");
CHECK(omittedResp.files.front().content.empty());
CHECK(omittedResp.snippetRenderMicros == 0);
CHECK(omittedResp.snippetsRendered == 0);
}
SECTION("graph query lists node types") {
fixture.initMetadata();
fixture.upsertNode("fn:list:1", "alpha", "function");
fixture.upsertNode("fn:list:2", "beta", "function");
fixture.upsertNode("path:list:1", "file.cpp", "file");
GraphQueryRequest req;
req.listTypes = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.kgAvailable);
CHECK(graphResp.originNode.label == "listTypes");
REQUIRE(graphResp.nodeTypeCounts.size() == 2);
CHECK(graphResp.nodeTypeCounts[0] == std::pair<std::string, uint64_t>{"function", 2});
CHECK(graphResp.nodeTypeCounts[1] == std::pair<std::string, uint64_t>{"file", 1});
}
SECTION("graph query list types reports store errors") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_nodes RENAME TO kg_nodes_broken");
GraphQueryRequest req;
req.listTypes = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code != ErrorCode::Success);
CHECK_FALSE(err.message.empty());
}
SECTION("graph query lists relation counts") {
fixture.initMetadata();
auto callerId = fixture.upsertNode("fn:rel:caller", "caller", "function");
auto calleeId = fixture.upsertNode("fn:rel:callee", "callee", "function");
auto fileId = fixture.upsertNode("path:rel:file", "file.cpp", "file");
fixture.addEdge(callerId, calleeId, "calls");
fixture.addEdge(callerId, fileId, "includes");
GraphQueryRequest req;
req.listRelations = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.originNode.label == "listRelations");
REQUIRE(graphResp.relationTypeCounts.size() == 2);
CHECK(graphResp.relationTypeCounts[0] == std::pair<std::string, uint64_t>{"calls", 1});
CHECK(graphResp.relationTypeCounts[1] == std::pair<std::string, uint64_t>{"includes", 1});
}
SECTION("graph query list relations reports store errors") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_edges RENAME TO kg_edges_broken");
GraphQueryRequest req;
req.listRelations = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code != ErrorCode::Success);
CHECK_FALSE(err.message.empty());
}
SECTION("graph query search mode requires a pattern") {
fixture.initMetadata();
GraphQueryRequest req;
req.searchMode = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message == "searchPattern is required for search mode");
}
SECTION("graph query search mode returns matching nodes with properties") {
fixture.initMetadata();
fixture.upsertNode("fn:search:alpha", "AlphaNode", "function", R"({"rank":1})");
fixture.upsertNode("fn:search:beta", "BetaNode", "function");
GraphQueryRequest req;
req.searchMode = true;
req.searchPattern = "%Alpha%";
req.includeNodeProperties = true;
req.limit = 10;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.originNode.label == "search:%Alpha%");
REQUIRE(graphResp.connectedNodes.size() == 1);
CHECK(graphResp.connectedNodes[0].nodeKey == "fn:search:alpha");
CHECK(graphResp.connectedNodes[0].properties == R"({"rank":1})");
}
SECTION("graph query search mode reports store errors") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_nodes RENAME TO kg_nodes_broken");
GraphQueryRequest req;
req.searchMode = true;
req.searchPattern = "%Alpha%";
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code != ErrorCode::Success);
CHECK_FALSE(err.message.empty());
}
SECTION("graph query list by type requires node type") {
fixture.initMetadata();
GraphQueryRequest req;
req.listByType = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message == "nodeType is required for listByType mode");
}
SECTION("graph query list by type returns paginated nodes") {
fixture.initMetadata();
fixture.upsertNode("fn:type:1", "TypeOne", "function", R"({"kind":"a"})");
fixture.upsertNode("fn:type:2", "TypeTwo", "function", R"({"kind":"b"})");
fixture.upsertNode("path:type:1", "file", "file");
GraphQueryRequest req;
req.listByType = true;
req.nodeType = "function";
req.limit = 1;
req.offset = 0;
req.includeNodeProperties = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.originNode.label == "listByType:function");
CHECK(graphResp.totalNodesFound == 2);
CHECK(graphResp.truncated);
REQUIRE(graphResp.connectedNodes.size() == 1);
CHECK(graphResp.connectedNodes[0].type == "function");
CHECK_FALSE(graphResp.connectedNodes[0].properties.empty());
}
SECTION("graph query list by type scopes and paginates within current code roots") {
fixture.initMetadata();
const auto sourcePath = fixture.testDir / "src" / "inside.cpp";
const auto includePath = fixture.testDir / "include" / "inside.hpp";
const auto testPath = fixture.testDir / "tests" / "outside.cpp";
for (const auto& path : {sourcePath, includePath, testPath}) {
std::filesystem::create_directories(path.parent_path());
std::ofstream out(path);
REQUIRE(out.good());
out << "int value;\n";
}
REQUIRE(fixture.repo->insertDocument(makeGraphDispatcherDocument(sourcePath, "hash-src"))
.has_value());
REQUIRE(
fixture.repo->insertDocument(makeGraphDispatcherDocument(includePath, "hash-include"))
.has_value());
REQUIRE(fixture.repo->insertDocument(makeGraphDispatcherDocument(testPath, "hash-test"))
.has_value());
fixture.upsertNode("function:insideSource@" + sourcePath.generic_string(), "insideSource",
"function");
fixture.upsertNode("function:insideHeader@" + includePath.generic_string(), "insideHeader",
"function");
fixture.upsertNode("function:outsideTest@" + testPath.generic_string(), "outsideTest",
"function");
GraphQueryRequest firstPage;
firstPage.listByType = true;
firstPage.nodeType = "function";
firstPage.scopePathPrefix = fixture.testDir.string();
firstPage.limit = 1;
auto firstRaw = dispatchRequest(*fixture.dispatcher, Request{firstPage});
REQUIRE(std::holds_alternative<GraphQueryResponse>(firstRaw));
const auto& first = std::get<GraphQueryResponse>(firstRaw);
CHECK(first.totalNodesFound == 2);
CHECK(first.truncated);
REQUIRE(first.connectedNodes.size() == 1);
CHECK(first.connectedNodes.front().label != "outsideTest");
firstPage.offset = 1;
auto secondRaw = dispatchRequest(*fixture.dispatcher, Request{firstPage});
REQUIRE(std::holds_alternative<GraphQueryResponse>(secondRaw));
const auto& second = std::get<GraphQueryResponse>(secondRaw);
CHECK(second.totalNodesFound == 2);
CHECK_FALSE(second.truncated);
REQUIRE(second.connectedNodes.size() == 1);
CHECK(second.connectedNodes.front().label != "outsideTest");
CHECK(second.connectedNodes.front().nodeId != first.connectedNodes.front().nodeId);
}
SECTION("graph query list by type reports store errors") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_nodes RENAME TO kg_nodes_broken");
GraphQueryRequest req;
req.listByType = true;
req.nodeType = "function";
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code != ErrorCode::Success);
CHECK_FALSE(err.message.empty());
}
SECTION("graph query isolated mode canonicalizes relation name") {
fixture.initMetadata();
auto calledId = fixture.upsertNode("fn:isolated:called", "called", "function");
fixture.upsertNode("fn:isolated:one", "isolated1", "function");
fixture.upsertNode("fn:isolated:two", "isolated2", "function");
auto callerId = fixture.upsertNode("fn:isolated:caller", "caller", "function");
fixture.addEdge(callerId, calledId, "calls");
GraphQueryRequest req;
req.isolatedMode = true;
req.nodeType = "function";
req.isolatedRelation = " CALL ";
req.limit = 10;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.originNode.label == "isolated:function:calls");
CHECK(graphResp.totalNodesFound == 3);
for (const auto& node : graphResp.connectedNodes) {
CHECK(node.nodeKey != "fn:isolated:called");
}
}
SECTION("graph query isolated mode uses default node type and relation") {
fixture.initMetadata();
auto calledId = fixture.upsertNode("fn:isolated:default:called", "called", "function");
fixture.upsertNode("fn:isolated:default:free", "free", "function");
auto callerId = fixture.upsertNode("fn:isolated:default:caller", "caller", "function");
fixture.addEdge(callerId, calledId, "calls");
GraphQueryRequest req;
req.isolatedMode = true;
req.limit = 10;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.originNode.label == "isolated:function:calls");
REQUIRE(graphResp.connectedNodes.size() == 2);
}
SECTION("graph query isolated mode reports store errors") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_nodes RENAME TO kg_nodes_broken");
GraphQueryRequest req;
req.isolatedMode = true;
req.limit = 10;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code != ErrorCode::Success);
CHECK_FALSE(err.message.empty());
}
SECTION("graph query node key lookup returns not found") {
fixture.initMetadata();
GraphQueryRequest req;
req.nodeKey = "missing-node";
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "Node not found: missing-node");
}
SECTION("graph query traverses from node key with canonical relation filters") {
fixture.initMetadata();
auto callerId = fixture.upsertNode("fn:key:caller", "caller", "function");
auto calleeId = fixture.upsertNode("fn:key:callee", "callee", "function");
auto fileId = fixture.upsertNode("path:key:file", "header", "file");
fixture.addEdge(callerId, calleeId, "calls", 0.8f, R"({"source":"unit"})");
fixture.addEdge(callerId, fileId, "includes");
GraphQueryRequest req;
req.nodeKey = "fn:key:caller";
req.maxDepth = 1;
req.maxResults = 10;
req.relationFilters = {" Call "};
req.includeEdgeProperties = true;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
CHECK(graphResp.kgAvailable);
CHECK(graphResp.originNode.nodeKey == "fn:key:caller");
REQUIRE(graphResp.connectedNodes.size() == 1);
CHECK(graphResp.connectedNodes[0].nodeKey == "fn:key:callee");
REQUIRE(graphResp.edges.size() == 1);
CHECK(graphResp.edges[0].relation == "calls");
CHECK(graphResp.edges[0].properties == R"({"source":"unit"})");
}
SECTION("graph query canonicalizes relation aliases and skips blank filters") {
fixture.initMetadata();
auto originId = fixture.upsertNode("fn:key:origin", "origin", "function");
auto includeId = fixture.upsertNode("path:key:include", "header", "file");
auto inheritId = fixture.upsertNode("cls:key:inherit", "Base", "class");
auto implementId = fixture.upsertNode("iface:key:impl", "Iface", "interface");
auto referenceId = fixture.upsertNode("sym:key:ref", "Ref", "symbol");
auto renameToId = fixture.upsertNode("path:key:rename_to", "new.cpp", "file");
auto renameFromId = fixture.upsertNode("path:key:rename_from", "old.cpp", "file");
auto customId = fixture.upsertNode("sym:key:custom", "Custom", "symbol");
fixture.addEdge(originId, includeId, "includes");
fixture.addEdge(originId, inheritId, "inherits");
fixture.addEdge(originId, implementId, "implements");
fixture.addEdge(originId, referenceId, "references");
fixture.addEdge(originId, renameToId, "renamed_to");
fixture.addEdge(originId, renameFromId, "renamed_from");
fixture.addEdge(originId, customId, "custom_rel");
GraphQueryRequest req;
req.nodeKey = "fn:key:origin";
req.maxDepth = 1;
req.maxResults = 20;
req.relationFilters = {" ", " Include ", "inherit", "implement",
"reference", "rename_to", "rename_from", "custom-rel"};
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphQueryResponse>(resp));
const auto& graphResp = std::get<GraphQueryResponse>(resp);
REQUIRE(graphResp.edges.size() == 7);
std::unordered_set<std::string> relations;
for (const auto& edge : graphResp.edges) {
relations.insert(edge.relation);
}
CHECK(relations.count("includes") == 1);
CHECK(relations.count("inherits") == 1);
CHECK(relations.count("implements") == 1);
CHECK(relations.count("references") == 1);
CHECK(relations.count("renamed_to") == 1);
CHECK(relations.count("renamed_from") == 1);
CHECK(relations.count("custom_rel") == 1);
}
SECTION("graph path history requires path") {
fixture.initMetadata();
auto resp = dispatchRequest(*fixture.dispatcher, Request{GraphPathHistoryRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidArgument);
CHECK(err.message == "Path is required for path history query");
}
SECTION("graph path history returns empty response when kg store is unavailable") {
fixture.initMetadata(false);
GraphPathHistoryRequest req;
req.path = "/repo/file.cpp";
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphPathHistoryResponse>(resp));
const auto& historyResp = std::get<GraphPathHistoryResponse>(resp);
CHECK(historyResp.queryPath == "/repo/file.cpp");
CHECK(historyResp.history.empty());
CHECK_FALSE(historyResp.hasMore);
}
SECTION("graph path history returns snapshot entries") {
fixture.initMetadata();
fixture.upsertNode("path:snap1:/repo/file.cpp", "/repo/file.cpp", "path",
R"({"snapshot_id":"snap1","path":"/repo/file.cpp"})");
fixture.upsertNode("path:snap2:/repo/file.cpp", "/repo/file.cpp", "path",
R"({"snapshot_id":"snap2","path":"/repo/file.cpp"})");
GraphPathHistoryRequest req;
req.path = "/repo/file.cpp";
req.limit = 10;
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<GraphPathHistoryResponse>(resp));
const auto& historyResp = std::get<GraphPathHistoryResponse>(resp);
CHECK(historyResp.queryPath == "/repo/file.cpp");
REQUIRE(historyResp.history.size() == 2);
CHECK(historyResp.history[0].snapshotId == "snap2");
CHECK(historyResp.history[0].blobHash.empty());
CHECK(historyResp.history[0].changeType == "unknown");
CHECK(historyResp.history[1].snapshotId == "snap1");
CHECK_FALSE(historyResp.hasMore);
}
SECTION("graph path history reports store errors") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_nodes RENAME TO kg_nodes_broken");
GraphPathHistoryRequest req;
req.path = "/repo/file.cpp";
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code != ErrorCode::Success);
CHECK_FALSE(err.message.empty());
}
SECTION("kg ingest reports unavailable metadata repository") {
auto resp = dispatchRequest(*fixture.dispatcher, Request{KgIngestRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Metadata repository unavailable");
}
SECTION("kg ingest reports unavailable knowledge graph store") {
fixture.initMetadata(false);
auto resp = dispatchRequest(*fixture.dispatcher, Request{KgIngestRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Knowledge graph store unavailable");
}
SECTION("kg ingest records node edge and alias resolution errors") {
fixture.initMetadata();
auto existingId = fixture.upsertNode("fn:existing", "existing", "function");
(void)existingId;
KgIngestRequest req;
req.nodes.push_back(KgIngestNode{"fn:existing", "existing updated", "function", ""});
req.edges.push_back(KgIngestEdge{"fn:existing", "fn:missing", "calls", 1.0f, ""});
req.aliases.push_back(KgIngestAlias{"fn:missing", "missing-alias", "unit", 0.5f});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK(ingestResp.success);
CHECK(ingestResp.nodesInserted == 1);
CHECK(ingestResp.edgesSkipped == 1);
CHECK(ingestResp.aliasesSkipped == 1);
REQUIRE(ingestResp.errors.size() == 2);
CHECK(ingestResp.errors[0] == "Edge destination node not found: fn:missing");
CHECK(ingestResp.errors[1] == "Alias node not found: fn:missing");
}
SECTION("kg ingest inserts nodes edges and aliases") {
fixture.initMetadata();
KgIngestRequest req;
req.nodes.push_back(
KgIngestNode{"fn:ingest:src", "Source", "function", R"({"role":"src"})"});
req.nodes.push_back(KgIngestNode{"fn:ingest:dst", "Dest", "function", ""});
req.edges.push_back(KgIngestEdge{"fn:ingest:src", "fn:ingest:dst", "calls", 0.7f,
R"({"origin":"dispatcher"})"});
req.aliases.push_back(KgIngestAlias{"fn:ingest:src", "sourceAlias", "unit", 0.9f});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK(ingestResp.success);
CHECK(ingestResp.nodesInserted == 2);
CHECK(ingestResp.edgesInserted == 1);
CHECK(ingestResp.aliasesInserted == 1);
CHECK(ingestResp.errors.empty());
auto sourceNode = fixture.kgStore->getNodeByKey("fn:ingest:src");
REQUIRE(sourceNode.has_value());
REQUIRE(sourceNode.value().has_value());
auto edges =
fixture.kgStore->getEdgesFrom(sourceNode.value()->id, std::string_view("calls"), 10, 0);
REQUIRE(edges.has_value());
REQUIRE(edges.value().size() == 1);
auto aliases = fixture.kgStore->resolveAliasExact("sourceAlias", 10);
REQUIRE(aliases.has_value());
REQUIRE(aliases.value().size() == 1);
}
SECTION("kg ingest reports node upsert failures") {
fixture.initMetadata();
fixture.execSql("ALTER TABLE kg_nodes RENAME TO kg_nodes_broken");
KgIngestRequest req;
req.nodes.push_back(KgIngestNode{"fn:broken:node", "Broken", "function", ""});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK_FALSE(ingestResp.success);
CHECK(ingestResp.nodesInserted == 0);
REQUIRE(ingestResp.errors.size() == 1);
CHECK(ingestResp.errors[0].find("Node upsert failed:") == 0);
}
SECTION("kg ingest inserts edges without de-duplication when requested") {
fixture.initMetadata();
fixture.upsertNode("fn:addedges:src", "Source", "function");
fixture.upsertNode("fn:addedges:dst", "Dest", "function");
KgIngestRequest req;
req.skipExistingEdges = false;
req.edges.push_back(KgIngestEdge{"fn:addedges:src", "fn:addedges:dst", "calls", 1.0f, ""});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK(ingestResp.success);
CHECK(ingestResp.edgesInserted == 1);
CHECK(ingestResp.errors.empty());
}
SECTION("kg ingest reports edge insert failures") {
fixture.initMetadata();
fixture.upsertNode("fn:edgefail:src", "Source", "function");
fixture.upsertNode("fn:edgefail:dst", "Dest", "function");
fixture.execSql("ALTER TABLE kg_edges RENAME TO kg_edges_broken");
KgIngestRequest req;
req.skipExistingEdges = false;
req.edges.push_back(KgIngestEdge{"fn:edgefail:src", "fn:edgefail:dst", "calls", 1.0f, ""});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK_FALSE(ingestResp.success);
CHECK(ingestResp.edgesInserted == 0);
REQUIRE(ingestResp.errors.size() == 1);
CHECK(ingestResp.errors[0].find("Edge insert failed:") == 0);
}
SECTION("kg ingest reports alias insert failures") {
fixture.initMetadata();
fixture.upsertNode("fn:aliasfail:node", "Node", "function");
fixture.execSql("ALTER TABLE kg_aliases RENAME TO kg_aliases_broken");
KgIngestRequest req;
req.aliases.push_back(KgIngestAlias{"fn:aliasfail:node", "brokenAlias", "unit", 0.4f});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK(ingestResp.success);
CHECK(ingestResp.aliasesInserted == 0);
REQUIRE(ingestResp.errors.size() == 1);
CHECK(ingestResp.errors[0].find("Alias insert failed:") == 0);
}
SECTION("kg ingest resolves existing nodes for edges and aliases") {
fixture.initMetadata();
auto srcId = fixture.upsertNode("fn:existing:src", "Source", "function");
auto dstId = fixture.upsertNode("fn:existing:dst", "Dest", "function");
auto aliasNodeId = fixture.upsertNode("fn:existing:alias", "AliasTarget", "function");
(void)srcId;
(void)dstId;
(void)aliasNodeId;
KgIngestRequest req;
req.edges.push_back(KgIngestEdge{"fn:existing:src", "fn:existing:dst", "calls", 1.0f, ""});
req.aliases.push_back(KgIngestAlias{"fn:existing:alias", "existingAlias", "", 0.6f});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK(ingestResp.success);
CHECK(ingestResp.nodesInserted == 0);
CHECK(ingestResp.edgesInserted == 1);
CHECK(ingestResp.aliasesInserted == 1);
CHECK(ingestResp.errors.empty());
auto aliases = fixture.kgStore->resolveAliasExact("existingAlias", 10);
REQUIRE(aliases.has_value());
REQUIRE(aliases.value().size() == 1);
}
SECTION("kg ingest reports missing edge source nodes") {
fixture.initMetadata();
fixture.upsertNode("fn:existing:dst_only", "Dest", "function");
KgIngestRequest req;
req.edges.push_back(
KgIngestEdge{"fn:missing:src", "fn:existing:dst_only", "calls", 1.0f, ""});
auto resp = dispatchRequest(*fixture.dispatcher, Request{req});
REQUIRE(std::holds_alternative<KgIngestResponse>(resp));
const auto& ingestResp = std::get<KgIngestResponse>(resp);
CHECK(ingestResp.success);
CHECK(ingestResp.edgesInserted == 0);
CHECK(ingestResp.edgesSkipped == 1);
REQUIRE(ingestResp.errors.size() == 1);
CHECK(ingestResp.errors[0] == "Edge source node not found: fn:missing:src");
}
}
TEST_CASE("RequestDispatcher: plugin handlers cover readiness and error branches",
"[daemon][plugin][dispatcher]") {
struct ReadyServiceFixture {
// ServiceManager keeps non-owning references to both dependencies. Member teardown is
// reverse declaration order, so the service shuts down before either dependency.
std::unique_ptr<StateComponent> state;
std::unique_ptr<DaemonLifecycleFsm> lifecycleFsm;
std::unique_ptr<ServiceManager> service;
};
auto makeReadyService = [](bool configureExternalPlugin = false) {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_plugin_dispatcher_");
if (configureExternalPlugin) {
cfg.pluginConfigs["dispatcher_external_plugin"] = R"({"mode":"trusted"})";
}
auto state = std::make_unique<StateComponent>();
state->readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state->readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
auto lifecycleFsm = std::make_unique<DaemonLifecycleFsm>();
auto svc = std::make_unique<ServiceManager>(cfg, *state, *lifecycleFsm);
svc->__test_pluginScanComplete(0);
REQUIRE(svc->getPluginHostFsmSnapshot().state == PluginHostState::Ready);
return ReadyServiceFixture{std::move(state), std::move(lifecycleFsm), std::move(svc)};
};
SECTION("plugin handlers reject non-ready host state") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
svc->__test_pluginLoadFailed("simulated plugin init failure");
auto resp = dispatchRequest(dispatcher, Request{PluginTrustListRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "Plugin subsystem not ready (state=failed)");
}
SECTION("plugin handlers reject missing service manager and non-ready FSM states") {
auto state = std::make_unique<StateComponent>();
state->readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state->readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
StubLifecycle lifecycle;
{
RequestDispatcher dispatcher(&lifecycle, nullptr, state.get());
auto resp = dispatchRequest(dispatcher, Request{PluginTrustListRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "Plugin host unavailable");
}
{
auto [readyState, lifecycleFsm, svc] = makeReadyService();
RequestDispatcher dispatcher(&lifecycle, svc.get(), readyState.get());
svc->__test_pluginScanStarted(1);
auto resp = dispatchRequest(dispatcher, Request{PluginTrustListRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "Plugin subsystem not ready (state=scanning)");
}
{
auto [readyState, lifecycleFsm, svc] = makeReadyService();
RequestDispatcher dispatcher(&lifecycle, svc.get(), readyState.get());
svc->__test_pluginLoaded("mock-loaded-plugin");
svc->__test_pluginScanComplete(1);
svc->__test_pluginLoaded("mock-loaded-plugin");
auto resp = dispatchRequest(dispatcher, Request{PluginTrustListRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InvalidState);
CHECK(err.message == "Plugin subsystem not ready (state=loading)");
}
}
SECTION("plugin handlers report missing hosts and trust failures") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
svc->__test_setAbiHost({});
svc->__test_setExternalPluginHost({});
auto scanResp = dispatchRequest(dispatcher, Request{PluginScanRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(scanResp));
const auto& scanErr = std::get<ErrorResponse>(scanResp);
CHECK(scanErr.code == ErrorCode::NotImplemented);
CHECK(scanErr.message == "No plugin host available");
PluginLoadRequest loadReq;
loadReq.pathOrName = "missing-plugin";
auto loadResp = dispatchRequest(dispatcher, Request{loadReq});
REQUIRE(std::holds_alternative<ErrorResponse>(loadResp));
const auto& loadErr = std::get<ErrorResponse>(loadResp);
CHECK(loadErr.code == ErrorCode::NotImplemented);
CHECK(loadErr.message == "No plugin host available");
auto unloadResp =
dispatchRequest(dispatcher, Request{PluginUnloadRequest{"missing-plugin"}});
REQUIRE(std::holds_alternative<ErrorResponse>(unloadResp));
const auto& unloadErr = std::get<ErrorResponse>(unloadResp);
CHECK(unloadErr.code == ErrorCode::NotFound);
CHECK(unloadErr.message == "Plugin not found or unload failed");
auto trustAddResp =
dispatchRequest(dispatcher, Request{PluginTrustAddRequest{"/tmp/missing-host"}});
REQUIRE(std::holds_alternative<ErrorResponse>(trustAddResp));
const auto& trustAddErr = std::get<ErrorResponse>(trustAddResp);
CHECK(trustAddErr.code == ErrorCode::Unknown);
CHECK(trustAddErr.message == "Trust add failed");
auto trustRemoveResp =
dispatchRequest(dispatcher, Request{PluginTrustRemoveRequest{"/tmp/missing-host"}});
REQUIRE(std::holds_alternative<ErrorResponse>(trustRemoveResp));
const auto& trustRemoveErr = std::get<ErrorResponse>(trustRemoveResp);
CHECK(trustRemoveErr.code == ErrorCode::Unknown);
CHECK(trustRemoveErr.message == "Trust remove failed");
}
SECTION("plugin load dry run reports missing plugin") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
PluginLoadRequest req;
req.pathOrName = (svc->getResolvedDataDir() / "missing_plugin.py").string();
req.dryRun = true;
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::NotFound);
CHECK(err.message == "Plugin not found");
}
SECTION("plugin dry run does not execute an untrusted standalone script") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
auto shimPath = makePythonShimPath();
ScopedEnvVar pythonGuard("PATH", shimPath.c_str());
const auto scanDir = makeTempDir("yams_untrusted_plugin_scan_");
const auto marker = scanDir / "scan-executed";
const auto script = scanDir / "untrusted.py";
{
std::ofstream out(script);
CHECK(out.good());
out << "from pathlib import Path\nPath(" << nlohmann::json(marker.string()).dump()
<< ").write_text('executed')\n";
}
PluginLoadRequest request;
request.pathOrName = script.string();
request.dryRun = true;
auto response = dispatchRequest(dispatcher, Request{request});
CHECK(std::holds_alternative<ErrorResponse>(response));
CHECK_FALSE(std::filesystem::exists(marker));
std::filesystem::remove_all(scanDir);
}
SECTION("plugin scan and load search default directories") {
auto homeDir = makeTempDir("yams_plugin_home_");
auto homeText = homeDir.string();
ScopedEnvVar homeGuard("HOME", homeText.c_str());
#ifdef _WIN32
ScopedEnvVar localAppDataGuard("LOCALAPPDATA", homeText.c_str());
#endif
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
#ifdef _WIN32
auto abiDir = homeDir / "yams" / "plugins";
#else
auto abiDir = homeDir / ".local" / "lib" / "yams" / "plugins";
#endif
std::filesystem::create_directories(abiDir);
auto fakePlugin = abiDir /
#ifdef _WIN32
"yams_default_scan.dll";
#else
"yams_default_scan.so";
#endif
{
std::ofstream out(fakePlugin, std::ios::binary);
out << "not a valid plugin";
}
#ifdef _WIN32
auto externalDir = homeDir / "yams" / "external-plugins";
#else
auto externalDir = homeDir / ".local" / "lib" / "yams" / "external-plugins";
#endif
std::filesystem::create_directories(externalDir);
auto externalPluginDir =
createMockExternalPlugin(externalDir, "dispatcher_external_default");
auto missingResp = dispatchRequest(
dispatcher, Request{PluginLoadRequest{"missing_default_plugin.so", "", false}});
REQUIRE(std::holds_alternative<ErrorResponse>(missingResp));
const auto& missingErr = std::get<ErrorResponse>(missingResp);
CHECK(missingErr.code == ErrorCode::NotFound);
CHECK(missingErr.message == "Plugin not found");
auto scanResp = dispatchRequest(dispatcher, Request{PluginScanRequest{}});
REQUIRE(std::holds_alternative<PluginScanResponse>(scanResp));
const auto& scan = std::get<PluginScanResponse>(scanResp);
bool foundDefaultPlugin = false;
bool foundDefaultExternalPlugin = false;
for (const auto& record : scan.plugins) {
if (record.path == fakePlugin.string()) {
foundDefaultPlugin = true;
CHECK(record.name == "yams_default_scan");
}
if (record.path == externalPluginDir.string()) {
foundDefaultExternalPlugin = true;
CHECK(record.name == "dispatcher_external_plugin");
}
}
CHECK(foundDefaultPlugin);
CHECK(foundDefaultExternalPlugin);
auto trustResp =
dispatchRequest(dispatcher, Request{PluginTrustAddRequest{abiDir.string()}});
REQUIRE(std::holds_alternative<SuccessResponse>(trustResp));
auto* external = svc->getExternalPluginHost();
REQUIRE(external != nullptr);
REQUIRE(external->trustAdd(externalDir).has_value());
PluginLoadRequest req;
req.pathOrName = fakePlugin.filename().string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Plugin load failed for: " + fakePlugin.string());
PluginLoadRequest externalReq;
externalReq.pathOrName = externalPluginDir.filename().string();
auto externalResp = dispatchRequest(dispatcher, Request{externalReq});
REQUIRE(std::holds_alternative<PluginLoadResponse>(externalResp));
const auto& externalLoad = std::get<PluginLoadResponse>(externalResp);
CHECK(externalLoad.loaded);
CHECK(externalLoad.record.name == "dispatcher_external_plugin");
CHECK(externalLoad.record.path == externalPluginDir.string());
auto externalUnloadResp =
dispatchRequest(dispatcher, Request{PluginUnloadRequest{"dispatcher_external_plugin"}});
REQUIRE(std::holds_alternative<ErrorResponse>(externalUnloadResp));
CHECK((std::get<ErrorResponse>(externalUnloadResp).code == ErrorCode::InvalidState));
CHECK((external->listLoaded().size() == 1));
}
SECTION("plugin handlers route external plugins when ABI host is absent") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
svc->__test_setAbiHost({});
auto pluginDir = createMockExternalPlugin(svc->getResolvedDataDir(), "dispatcher_external");
PluginScanRequest scanReq;
scanReq.target = pluginDir.string();
auto scanResp = dispatchRequest(dispatcher, Request{scanReq});
REQUIRE(std::holds_alternative<PluginScanResponse>(scanResp));
const auto& scan = std::get<PluginScanResponse>(scanResp);
REQUIRE(scan.plugins.size() == 1);
CHECK(scan.plugins.front().name == "dispatcher_external_plugin");
PluginLoadRequest dryRunReq;
dryRunReq.pathOrName = pluginDir.string();
dryRunReq.dryRun = true;
auto dryRunResp = dispatchRequest(dispatcher, Request{dryRunReq});
REQUIRE(std::holds_alternative<PluginLoadResponse>(dryRunResp));
const auto& dryRun = std::get<PluginLoadResponse>(dryRunResp);
CHECK_FALSE(dryRun.loaded);
CHECK(dryRun.message == "dry-run");
CHECK(dryRun.record.name == "dispatcher_external_plugin");
auto* external = svc->getExternalPluginHost();
REQUIRE(external != nullptr);
REQUIRE(external->trustAdd(pluginDir).has_value());
PluginLoadRequest loadReq;
loadReq.pathOrName = pluginDir.string();
auto loadResp = dispatchRequest(dispatcher, Request{loadReq});
REQUIRE(std::holds_alternative<PluginLoadResponse>(loadResp));
const auto& load = std::get<PluginLoadResponse>(loadResp);
CHECK(load.loaded);
CHECK(load.message == "loaded");
CHECK(load.record.name == "dispatcher_external_plugin");
CHECK(external->listLoaded().size() == 1);
auto unloadResp =
dispatchRequest(dispatcher, Request{PluginUnloadRequest{"dispatcher_external_plugin"}});
REQUIRE(std::holds_alternative<ErrorResponse>(unloadResp));
CHECK((std::get<ErrorResponse>(unloadResp).code == ErrorCode::InvalidState));
CHECK((external->listLoaded().size() == 1));
}
SECTION("plugin load routes direct external files before ABI fallback") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
auto shimPath = makePythonShimPath();
ScopedEnvVar pythonGuard("PATH", shimPath.c_str());
auto pluginDir =
createMockExternalPlugin(svc->getResolvedDataDir(), "dispatcher_external_py");
auto pluginFile = pluginDir / "plugin.py";
auto* external = svc->getExternalPluginHost();
REQUIRE(external != nullptr);
REQUIRE(external->trustAdd(pluginFile).has_value());
PluginLoadRequest req;
req.pathOrName = pluginFile.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PluginLoadResponse>(resp));
const auto& load = std::get<PluginLoadResponse>(resp);
CHECK(load.loaded);
CHECK(load.record.name == "dispatcher_external_plugin");
CHECK(load.record.path == pluginFile.string());
auto unloadResp =
dispatchRequest(dispatcher, Request{PluginUnloadRequest{"dispatcher_external_plugin"}});
REQUIRE(std::holds_alternative<ErrorResponse>(unloadResp));
CHECK((std::get<ErrorResponse>(unloadResp).code == ErrorCode::InvalidState));
CHECK((external->listLoaded().size() == 1));
}
SECTION("plugin load recognizes manifest-adjacent executable files") {
#ifdef _WIN32
SKIP("manifest-adjacent executable file semantics are POSIX-specific");
#else
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
auto pluginDir =
createMockExternalPlugin(svc->getResolvedDataDir(), "dispatcher_external_exec");
auto executable = pluginDir / "plugin_runner";
std::filesystem::copy_file(pluginDir / "plugin.py", executable,
std::filesystem::copy_options::overwrite_existing);
std::filesystem::permissions(executable,
std::filesystem::perms::owner_read |
std::filesystem::perms::owner_write |
std::filesystem::perms::owner_exec,
std::filesystem::perm_options::replace);
auto* external = svc->getExternalPluginHost();
REQUIRE(external != nullptr);
REQUIRE(external->trustAdd(executable).has_value());
PluginLoadRequest req;
req.pathOrName = executable.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<PluginLoadResponse>(resp));
const auto& load = std::get<PluginLoadResponse>(resp);
CHECK(load.loaded);
CHECK(load.record.name == "dispatcher_external_plugin");
CHECK(load.record.path == executable.string());
auto unloadResp =
dispatchRequest(dispatcher, Request{PluginUnloadRequest{"dispatcher_external_plugin"}});
REQUIRE(std::holds_alternative<ErrorResponse>(unloadResp));
CHECK((std::get<ErrorResponse>(unloadResp).code == ErrorCode::InvalidState));
CHECK((external->listLoaded().size() == 1));
#endif
}
SECTION("plugin load falls back after untrusted external file fails") {
auto [state, lifecycleFsm, svc] = makeReadyService();
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
auto pluginDir =
createMockExternalPlugin(svc->getResolvedDataDir(), "dispatcher_external_untrusted");
auto pluginFile = pluginDir / "plugin.py";
PluginLoadRequest req;
req.pathOrName = pluginFile.string();
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message == "Plugin load failed for: " + pluginFile.string());
}
SECTION("plugin trust add loads regular files with configured plugin settings") {
auto [state, lifecycleFsm, svc] = makeReadyService(true);
StubLifecycle lifecycle;
RequestDispatcher dispatcher(&lifecycle, svc.get(), state.get());
auto shimPath = makePythonShimPath();
ScopedEnvVar pythonGuard("PATH", shimPath.c_str());
auto pluginDir = createMockExternalPlugin(svc->getResolvedDataDir(),
"dispatcher_external_trust_file", true);
auto pluginFile = pluginDir / "plugin.py";
auto trustAddResp =
dispatchRequest(dispatcher, Request{PluginTrustAddRequest{pluginFile.string()}});
REQUIRE(std::holds_alternative<SuccessResponse>(trustAddResp));
auto* external = svc->getExternalPluginHost();
REQUIRE(external != nullptr);
REQUIRE(waitForCondition([&] { return external->listLoaded().size() == 1; }));
CHECK(external->listLoaded().front().path == pluginFile);
auto secondTrustResp =
dispatchRequest(dispatcher, Request{PluginTrustAddRequest{pluginDir.string()}});
REQUIRE(std::holds_alternative<SuccessResponse>(secondTrustResp));
CHECK(waitForCondition([&] { return external->listLoaded().size() == 1; }));
auto unloadResp =
dispatchRequest(dispatcher, Request{PluginUnloadRequest{"dispatcher_external_plugin"}});
REQUIRE(std::holds_alternative<ErrorResponse>(unloadResp));
CHECK((std::get<ErrorResponse>(unloadResp).code == ErrorCode::InvalidState));
CHECK((external->listLoaded().size() == 1));
}
}
TEST_CASE("RequestDispatcher: collection handlers report missing dependencies and batch defaults",
"[daemon][collections][dispatcher]") {
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_collections_dispatcher_");
StubLifecycle lifecycle;
StateComponent state;
state.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
state.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
ServiceManager svc(cfg, state, lifecycleFsm);
RequestDispatcher dispatcher(&lifecycle, &svc, &state);
struct SeededDocument {
int64_t id{0};
std::string hash;
std::string relativePath;
};
struct ReadyCollectionService {
std::shared_ptr<metadata::ConnectionPool> pool;
std::shared_ptr<metadata::MetadataRepository> metadata;
std::shared_ptr<StubContentStore> content;
std::shared_ptr<ServiceManager> manager;
};
auto makeReadyService = [&](StateComponent& readyState, DaemonLifecycleFsm& readyLifecycleFsm,
const std::string& prefix) {
DaemonConfig readyCfg = cfg;
readyCfg.dataDir = makeTempDir(prefix);
metadata::ConnectionPoolConfig poolConfig{};
auto pool = std::make_shared<metadata::ConnectionPool>(
makeTempMetadataDbPath(prefix + "db_").string(), poolConfig);
REQUIRE(pool->initialize().has_value());
ReadyCollectionService ready{
.pool = pool,
.metadata = makeReadyMetadataRepo(pool),
.content = std::make_shared<StubContentStore>(),
.manager = std::make_shared<ServiceManager>(readyCfg, readyState, readyLifecycleFsm),
};
ready.manager->__test_setMetadataRepo(ready.metadata);
ready.manager->__test_setContentStore(ready.content);
readyState.readiness.metadataRepoReady.store(true, std::memory_order_relaxed);
readyState.readiness.contentStoreReady.store(true, std::memory_order_relaxed);
return ready;
};
auto seedDocument = [&](const ReadyCollectionService& ready, const std::string& relativePath,
const std::string& text, const std::string& collectionName,
const std::string& snapshotId) {
const auto textBytes = std::as_bytes(std::span<const char>(text.data(), text.size()));
auto storeRes = ready.content->storeBytes(textBytes, {});
REQUIRE(storeRes.has_value());
std::filesystem::path relativeDocPath(relativePath);
std::string extension = relativeDocPath.extension().string();
if (!extension.empty() && extension.front() == '.') {
extension.erase(0, 1);
}
metadata::DocumentInfo doc{};
doc.fileName = relativeDocPath.filename().string();
doc.filePath = relativePath;
doc.fileExtension = extension;
doc.fileSize = static_cast<int64_t>(text.size());
doc.sha256Hash = storeRes.value().contentHash;
doc.mimeType = "text/plain";
const auto now =
std::chrono::time_point_cast<std::chrono::seconds>(std::chrono::system_clock::now());
doc.createdTime = now;
doc.modifiedTime = now;
doc.indexedTime = now;
auto idRes = ready.metadata->insertDocument(doc);
REQUIRE(idRes.has_value());
REQUIRE(
ready.metadata
->setMetadata(idRes.value(), "collection", metadata::MetadataValue(collectionName))
.has_value());
REQUIRE(ready.metadata
->setMetadata(idRes.value(), "snapshot_id", metadata::MetadataValue(snapshotId))
.has_value());
return SeededDocument{idRes.value(), storeRes.value().contentHash, relativePath};
};
SECTION("list snapshots returns internal error when metadata repo is unavailable") {
auto resp = dispatchRequest(dispatcher, Request{ListSnapshotsRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Metadata repository unavailable") != std::string::npos);
}
SECTION("list snapshots forwards repository errors") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
repo->setSnapshotError(Error{ErrorCode::DatabaseError, "snapshot lookup failed"});
svc.__test_setMetadataRepo(repo);
auto resp = dispatchRequest(dispatcher, Request{ListSnapshotsRequest{}});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::DatabaseError);
CHECK(err.message == "snapshot lookup failed");
}
SECTION("restore collection returns internal error before dependencies are initialized") {
RestoreCollectionRequest req;
req.collection = "dispatcher-collection";
req.outputDirectory = "out";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Metadata repository unavailable") != std::string::npos);
}
SECTION("restore collection returns internal error when content store is unavailable") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
svc.__test_setMetadataRepo(repo);
RestoreCollectionRequest req;
req.collection = "dispatcher-collection";
req.outputDirectory = "out";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Content store unavailable") != std::string::npos);
}
SECTION("restore collection forwards repository query errors") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
repo->setQueryError(Error{ErrorCode::DatabaseError, "collection query failed"});
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
RestoreCollectionRequest req;
req.collection = "dispatcher-collection";
req.outputDirectory = "out";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::DatabaseError);
CHECK(err.message == "Failed to find collection documents: collection query failed");
}
SECTION("restore snapshot returns internal error when metadata repo is unavailable") {
RestoreSnapshotRequest req;
req.snapshotId = "dispatcher-snapshot";
req.outputDirectory = "out";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Metadata repository unavailable") != std::string::npos);
}
SECTION("restore snapshot returns internal error when content store is unavailable") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
svc.__test_setMetadataRepo(repo);
RestoreSnapshotRequest req;
req.snapshotId = "dispatcher-snapshot";
req.outputDirectory = "out";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Content store unavailable") != std::string::npos);
}
SECTION("restore snapshot forwards repository query errors") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
repo->setSnapshotDocumentsError(Error{ErrorCode::DatabaseError, "snapshot query failed"});
svc.__test_setMetadataRepo(repo);
svc.__test_setContentStore(std::make_shared<StubContentStore>());
RestoreSnapshotRequest req;
req.snapshotId = "dispatcher-snapshot";
req.outputDirectory = "out";
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::DatabaseError);
CHECK(err.message == "Failed to find snapshot documents: snapshot query failed");
}
SECTION("metadata value counts returns internal error when metadata repo is unavailable") {
MetadataValueCountsRequest req;
req.keys = {"collection"};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::InternalError);
CHECK(err.message.find("Metadata repository unavailable") != std::string::npos);
}
SECTION("metadata value counts forwards repository errors") {
auto repo = std::make_shared<StubPruneMetadataRepository>();
repo->setMetadataValueCountsError(Error{ErrorCode::DatabaseError, "counts failed"});
svc.__test_setMetadataRepo(repo);
MetadataValueCountsRequest req;
req.keys = {"collection"};
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<ErrorResponse>(resp));
const auto& err = std::get<ErrorResponse>(resp);
CHECK(err.code == ErrorCode::DatabaseError);
CHECK(err.message == "counts failed");
}
SECTION("batch request returns not implemented items for each sequence") {
BatchRequest req;
req.items.push_back(BatchItem{.sequenceId = 7, .request = SearchRequest{.query = "one"}});
req.items.push_back(BatchItem{.sequenceId = 8, .request = ListRequest{}});
auto resp = dispatchRequest(dispatcher, Request{req});
REQUIRE(std::holds_alternative<BatchResponse>(resp));
const auto& batchResp = std::get<BatchResponse>(resp);
CHECK(batchResp.totalCount == 2);
CHECK(batchResp.successCount == 0);
CHECK(batchResp.errorCount == 2);
REQUIRE(batchResp.items.size() == 2);
CHECK(batchResp.items[0].sequenceId == 7);
CHECK_FALSE(batchResp.items[0].success);
REQUIRE(std::holds_alternative<ErrorResponse>(batchResp.items[0].response));
CHECK(std::get<ErrorResponse>(batchResp.items[0].response).code ==
ErrorCode::NotImplemented);
CHECK(batchResp.items[1].sequenceId == 8);
CHECK_FALSE(batchResp.items[1].success);
}
SECTION("restore collection exercises regex, layout, and io branches") {
StubLifecycle readyLifecycle;
StateComponent readyState;
DaemonLifecycleFsm readyLifecycleFsm;
auto ready =
makeReadyService(readyState, readyLifecycleFsm, "yams_collections_live_dispatcher_");
RequestDispatcher readyDispatcher(&readyLifecycle, ready.manager.get(), &readyState);
const std::string collectionName = "dispatcher-live-collection";
const std::string snapshotId = "dispatcher-live-snapshot";
const auto seeded = seedDocument(ready, "nested/alpha.txt", "alpha collection text\n",
collectionName, snapshotId);
auto includeOut = makeTempDir("yams_collection_include_");
RestoreCollectionRequest includeReq;
includeReq.collection = collectionName;
includeReq.outputDirectory = includeOut.string();
includeReq.includePatterns = {"*alpha.txt"};
includeReq.dryRun = true;
auto includeResp = dispatchRequest(readyDispatcher, Request{includeReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(includeResp));
const auto& included = std::get<RestoreCollectionResponse>(includeResp);
REQUIRE(included.files.size() == 1);
CHECK(included.filesRestored == 1);
CHECK(included.files.front().path == (includeOut / seeded.relativePath).string());
RestoreCollectionRequest invalidPatternReq;
invalidPatternReq.collection = collectionName;
invalidPatternReq.outputDirectory = includeOut.string();
invalidPatternReq.includePatterns = {"["};
invalidPatternReq.dryRun = true;
auto invalidPatternResp = dispatchRequest(readyDispatcher, Request{invalidPatternReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(invalidPatternResp));
const auto& invalidPattern = std::get<RestoreCollectionResponse>(invalidPatternResp);
CHECK(invalidPattern.filesRestored == 0);
CHECK(invalidPattern.files.empty());
RestoreCollectionRequest excludeReq;
excludeReq.collection = collectionName;
excludeReq.outputDirectory = includeOut.string();
excludeReq.excludePatterns = {"*alpha.txt"};
excludeReq.dryRun = true;
auto excludeResp = dispatchRequest(readyDispatcher, Request{excludeReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(excludeResp));
const auto& excluded = std::get<RestoreCollectionResponse>(excludeResp);
CHECK(excluded.filesRestored == 0);
CHECK(excluded.files.empty());
auto writeOut = makeTempDir("yams_collection_write_");
RestoreCollectionRequest writeReq;
writeReq.collection = collectionName;
writeReq.outputDirectory = writeOut.string();
writeReq.layoutTemplate = "deep/{hash}{ext}";
writeReq.overwrite = true;
writeReq.dryRun = false;
auto writeResp = dispatchRequest(readyDispatcher, Request{writeReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(writeResp));
const auto& written = std::get<RestoreCollectionResponse>(writeResp);
REQUIRE(written.files.size() == 1);
CHECK(written.filesRestored == 1);
const auto writtenPath = writeOut / "deep" / (seeded.hash + ".txt");
CHECK(std::filesystem::path(written.files.front().path).lexically_normal() ==
writtenPath.lexically_normal());
REQUIRE(std::filesystem::exists(writtenPath));
std::ifstream writtenFile(writtenPath, std::ios::binary);
REQUIRE(writtenFile.good());
std::string writtenText((std::istreambuf_iterator<char>(writtenFile)),
std::istreambuf_iterator<char>());
CHECK(writtenText == "alpha collection text\n");
auto createDirBlocker = makeTempDir("yams_collection_create_dir_blocker_") / "blocked";
{
std::ofstream blocker(createDirBlocker);
REQUIRE(blocker.good());
blocker << "blocker";
}
RestoreCollectionRequest createDirFailReq;
createDirFailReq.collection = collectionName;
createDirFailReq.outputDirectory = createDirBlocker.string();
createDirFailReq.createDirs = true;
createDirFailReq.overwrite = true;
createDirFailReq.dryRun = false;
auto createDirFailResp = dispatchRequest(readyDispatcher, Request{createDirFailReq});
REQUIRE(std::holds_alternative<ErrorResponse>(createDirFailResp));
CHECK(std::get<ErrorResponse>(createDirFailResp).code == ErrorCode::IOError);
CHECK(std::get<ErrorResponse>(createDirFailResp)
.message.find("Failed to create output directory") != std::string::npos);
auto parentFailOut = makeTempDir("yams_collection_parent_fail_");
auto occupiedParent = parentFailOut / "occupied";
{
std::ofstream blocker(occupiedParent);
REQUIRE(blocker.good());
blocker << "occupied";
}
RestoreCollectionRequest parentFailReq;
parentFailReq.collection = collectionName;
parentFailReq.outputDirectory = parentFailOut.string();
parentFailReq.layoutTemplate = "occupied/restored.txt";
parentFailReq.overwrite = true;
parentFailReq.dryRun = false;
auto parentFailResp = dispatchRequest(readyDispatcher, Request{parentFailReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(parentFailResp));
const auto& parentFailed = std::get<RestoreCollectionResponse>(parentFailResp);
REQUIRE(parentFailed.files.size() == 1);
CHECK(parentFailed.filesRestored == 0);
CHECK(parentFailed.files.front().skipped);
CHECK(parentFailed.files.front().skipReason.find("Failed to create parent directory") !=
std::string::npos);
auto openFailOut = makeTempDir("yams_collection_open_fail_");
RestoreCollectionRequest openFailReq;
openFailReq.collection = collectionName;
openFailReq.outputDirectory = openFailOut.string();
openFailReq.layoutTemplate = ".";
openFailReq.overwrite = true;
openFailReq.dryRun = false;
auto openFailResp = dispatchRequest(readyDispatcher, Request{openFailReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(openFailResp));
const auto& openFailed = std::get<RestoreCollectionResponse>(openFailResp);
REQUIRE(openFailed.files.size() == 1);
CHECK(openFailed.filesRestored == 0);
CHECK(openFailed.files.front().skipped);
CHECK(openFailed.files.front().skipReason == "Failed to open output file");
auto writeFailOut = makeTempDir("yams_collection_write_fail_");
RestoreCollectionRequest writeFailReq;
writeFailReq.collection = collectionName;
writeFailReq.outputDirectory = writeFailOut.string();
writeFailReq.layoutTemplate = "write-fail/{name}{ext}";
writeFailReq.overwrite = true;
writeFailReq.dryRun = false;
RequestDispatcher::__test_forceRestoreWriteFailureOnce();
auto writeFailResp = dispatchRequest(readyDispatcher, Request{writeFailReq});
REQUIRE(std::holds_alternative<RestoreCollectionResponse>(writeFailResp));
const auto& writeFailed = std::get<RestoreCollectionResponse>(writeFailResp);
REQUIRE(writeFailed.files.size() == 1);
CHECK(writeFailed.filesRestored == 0);
CHECK(writeFailed.files.front().skipped);
CHECK(writeFailed.files.front().skipReason == "Failed to write file content");
}
SECTION("restore snapshot exercises regex, layout, and io branches") {
StubLifecycle readyLifecycle;
StateComponent readyState;
DaemonLifecycleFsm readyLifecycleFsm;
auto ready =
makeReadyService(readyState, readyLifecycleFsm, "yams_snapshots_live_dispatcher_");
RequestDispatcher readyDispatcher(&readyLifecycle, ready.manager.get(), &readyState);
const std::string collectionName = "dispatcher-live-collection";
const std::string snapshotId = "dispatcher-live-snapshot";
const auto seeded = seedDocument(ready, "nested/alpha.txt", "alpha snapshot text\n",
collectionName, snapshotId);
auto defaultOut = makeTempDir("yams_snapshot_default_");
RestoreSnapshotRequest defaultReq;
defaultReq.snapshotId = snapshotId;
defaultReq.outputDirectory = defaultOut.string();
defaultReq.includePatterns = {"*alpha.txt"};
defaultReq.dryRun = true;
auto defaultResp = dispatchRequest(readyDispatcher, Request{defaultReq});
REQUIRE(std::holds_alternative<RestoreSnapshotResponse>(defaultResp));
const auto& defaultRestore = std::get<RestoreSnapshotResponse>(defaultResp);
REQUIRE(defaultRestore.files.size() == 1);
CHECK(defaultRestore.filesRestored == 1);
CHECK(defaultRestore.files.front().path == (defaultOut / seeded.relativePath).string());
RestoreSnapshotRequest invalidPatternReq;
invalidPatternReq.snapshotId = snapshotId;
invalidPatternReq.outputDirectory = defaultOut.string();
invalidPatternReq.includePatterns = {"["};
invalidPatternReq.dryRun = true;
auto invalidPatternResp = dispatchRequest(readyDispatcher, Request{invalidPatternReq});
REQUIRE(std::holds_alternative<RestoreSnapshotResponse>(invalidPatternResp));
const auto& invalidPattern = std::get<RestoreSnapshotResponse>(invalidPatternResp);
CHECK(invalidPattern.filesRestored == 0);
CHECK(invalidPattern.files.empty());
auto writeOut = makeTempDir("yams_snapshot_write_");
RestoreSnapshotRequest writeReq;
writeReq.snapshotId = snapshotId;
writeReq.outputDirectory = writeOut.string();
writeReq.layoutTemplate = "deep/{name}{ext}";
writeReq.overwrite = true;
writeReq.dryRun = false;
auto writeResp = dispatchRequest(readyDispatcher, Request{writeReq});
REQUIRE(std::holds_alternative<RestoreSnapshotResponse>(writeResp));
const auto& written = std::get<RestoreSnapshotResponse>(writeResp);
REQUIRE(written.files.size() == 1);
CHECK(written.filesRestored == 1);
const auto writtenPath = writeOut / "deep" / "alpha.txt";
CHECK(std::filesystem::path(written.files.front().path).lexically_normal() ==
writtenPath.lexically_normal());
REQUIRE(std::filesystem::exists(writtenPath));
std::ifstream writtenFile(writtenPath, std::ios::binary);
REQUIRE(writtenFile.good());
std::string writtenText((std::istreambuf_iterator<char>(writtenFile)),
std::istreambuf_iterator<char>());
CHECK(writtenText == "alpha snapshot text\n");
auto createDirBlocker = makeTempDir("yams_snapshot_create_dir_blocker_") / "blocked";
{
std::ofstream blocker(createDirBlocker);
REQUIRE(blocker.good());
blocker << "blocker";
}
RestoreSnapshotRequest createDirFailReq;
createDirFailReq.snapshotId = snapshotId;
createDirFailReq.outputDirectory = createDirBlocker.string();
createDirFailReq.createDirs = true;
createDirFailReq.overwrite = true;
createDirFailReq.dryRun = false;
auto createDirFailResp = dispatchRequest(readyDispatcher, Request{createDirFailReq});
REQUIRE(std::holds_alternative<ErrorResponse>(createDirFailResp));
CHECK(std::get<ErrorResponse>(createDirFailResp).code == ErrorCode::IOError);
CHECK(std::get<ErrorResponse>(createDirFailResp)
.message.find("Failed to create output directory") != std::string::npos);
auto parentFailOut = makeTempDir("yams_snapshot_parent_fail_");
auto occupiedParent = parentFailOut / "occupied";
{
std::ofstream blocker(occupiedParent);
REQUIRE(blocker.good());
blocker << "occupied";
}
RestoreSnapshotRequest parentFailReq;
parentFailReq.snapshotId = snapshotId;
parentFailReq.outputDirectory = parentFailOut.string();
parentFailReq.layoutTemplate = "occupied/restored.txt";
parentFailReq.overwrite = true;
parentFailReq.dryRun = false;
auto parentFailResp = dispatchRequest(readyDispatcher, Request{parentFailReq});
REQUIRE(std::holds_alternative<RestoreSnapshotResponse>(parentFailResp));
const auto& parentFailed = std::get<RestoreSnapshotResponse>(parentFailResp);
REQUIRE(parentFailed.files.size() == 1);
CHECK(parentFailed.filesRestored == 0);
CHECK(parentFailed.files.front().skipped);
CHECK(parentFailed.files.front().skipReason.find("Failed to create parent directory") !=
std::string::npos);
auto openFailOut = makeTempDir("yams_snapshot_open_fail_");
RestoreSnapshotRequest openFailReq;
openFailReq.snapshotId = snapshotId;
openFailReq.outputDirectory = openFailOut.string();
openFailReq.layoutTemplate = ".";
openFailReq.overwrite = true;
openFailReq.dryRun = false;
auto openFailResp = dispatchRequest(readyDispatcher, Request{openFailReq});
REQUIRE(std::holds_alternative<RestoreSnapshotResponse>(openFailResp));
const auto& openFailed = std::get<RestoreSnapshotResponse>(openFailResp);
REQUIRE(openFailed.files.size() == 1);
CHECK(openFailed.filesRestored == 0);
CHECK(openFailed.files.front().skipped);
CHECK(openFailed.files.front().skipReason == "Failed to open output file");
auto writeFailOut = makeTempDir("yams_snapshot_write_fail_");
RestoreSnapshotRequest writeFailReq;
writeFailReq.snapshotId = snapshotId;
writeFailReq.outputDirectory = writeFailOut.string();
writeFailReq.layoutTemplate = "write-fail/{hash}{ext}";
writeFailReq.overwrite = true;
writeFailReq.dryRun = false;
RequestDispatcher::__test_forceRestoreWriteFailureOnce();
auto writeFailResp = dispatchRequest(readyDispatcher, Request{writeFailReq});
REQUIRE(std::holds_alternative<RestoreSnapshotResponse>(writeFailResp));
const auto& writeFailed = std::get<RestoreSnapshotResponse>(writeFailResp);
REQUIRE(writeFailed.files.size() == 1);
CHECK(writeFailed.filesRestored == 0);
CHECK(writeFailed.files.front().skipped);
CHECK(writeFailed.files.front().skipReason == "Failed to write file content");
}
}
TEST_CASE("DaemonMetrics: snapshot includes canonical readiness flags",
"[daemon][metrics][readiness]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_metrics_readiness_flags_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
const std::array<std::string_view, 12> requiredCoreReadinessKeys = {
readiness::kIpcServer, readiness::kContentStore,
readiness::kDatabase, readiness::kMetadataRepo,
readiness::kSearchEngine, readiness::kModelProvider,
readiness::kVectorIndex, readiness::kVectorDb,
readiness::kPlugins, readiness::kVectorDbInitAttempted,
readiness::kVectorDbReady, readiness::kVectorDbDim};
for (const auto key : requiredCoreReadinessKeys) {
INFO("missing readiness key: " << key);
REQUIRE(snap->readinessStates.count(std::string(key)) > 0);
}
}
TEST_CASE("DaemonMetrics: snapshot exposes metadata integrity fast-path use",
"[daemon][metrics][readiness]") {
StateComponent state;
state.readiness.databaseIntegrityFastPath.store(true, std::memory_order_release);
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_metrics_integrity_fast_path_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
REQUIRE((snap->diagnosticCounters.count("database_integrity_fast_path") == 1));
CHECK((snap->diagnosticCounters.at("database_integrity_fast_path") == 1));
}
TEST_CASE("DaemonMetrics: snapshot reports generated build version", "[daemon][metrics][version]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_metrics_version_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
CHECK(snap->version == yams::version::string_v);
}
TEST_CASE("DaemonMetrics: cold snapshot refreshes synchronously", "[daemon][metrics][freshness]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_metrics_freshness_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
CHECK_FALSE(snap->statusSnapshotStale);
CHECK(snap->statusSnapshotAgeMs == 0);
CHECK(snap->statusDetailAgeMs == 0);
CHECK(snap->statusDetailStale);
CHECK_FALSE(snap->statusDetailAvailable);
}
TEST_CASE("RequestDispatcher: status refreshes cold snapshot without retry hint",
"[daemon][status][freshness]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_busy_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
RequestDispatcher dispatcher(nullptr, &svc, &state, &metrics);
auto resp = dispatchRequest(dispatcher, Request{StatusRequest{}});
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& status = std::get<StatusResponse>(resp);
CHECK(status.running);
CHECK(status.retryAfterMs == 0);
}
TEST_CASE("RequestDispatcher: status reports effective search maintenance provenance",
"[daemon][status][search][config]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig config;
config.dataDir = makeTempDir("yams_status_search_maintenance_");
config.searchMaintenance.automaticRebuildsEnabled = false;
config.searchMaintenance.automaticRebuildsSource = "typed:test";
ServiceManager services(config, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &services, services.getWorkCoordinator());
RequestDispatcher dispatcher(nullptr, &services, &state, &metrics);
const auto response = dispatchRequest(dispatcher, Request{StatusRequest{}});
REQUIRE(std::holds_alternative<StatusResponse>(response));
const auto& status = std::get<StatusResponse>(response);
CHECK_FALSE(status.searchAutomaticRebuildsEnabled);
CHECK((status.searchAutomaticRebuildsSource == "typed:test"));
CHECK(status.runtimeTuning.contains("ipc.timeout_ms"));
CHECK(status.runtimeTuning.contains("ipc.timeout_ms.source"));
CHECK(status.runtimeTuning.contains("post_ingest.total_concurrent"));
CHECK(status.runtimeTuning.contains("resource.memory_budget_bytes"));
}
TEST_CASE("RequestDispatcher: cold status still reports active repair work",
"[daemon][status][freshness]") {
StateComponent state;
state.stats.repairRunning.store(true, std::memory_order_relaxed);
state.stats.repairQueueDepth.store(7, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_busy_repair_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
RequestDispatcher dispatcher(nullptr, &svc, &state, &metrics);
auto resp = dispatchRequest(dispatcher, Request{StatusRequest{}});
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& status = std::get<StatusResponse>(resp);
CHECK(status.retryAfterMs == 0);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairRunning)) == 1);
REQUIRE(status.requestCounts.count(std::string(metrics::kRepairQueueDepth)) == 1);
CHECK(status.requestCounts.at(std::string(metrics::kRepairRunning)) == 1);
CHECK(status.requestCounts.at(std::string(metrics::kRepairQueueDepth)) == 7);
}
TEST_CASE("RequestDispatcher: status mirrors memory breakdown into requestCounts",
"[daemon][status][memory]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_status_mem_breakdown_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
RequestDispatcher dispatcher(nullptr, &svc, &state, &metrics);
auto resp = dispatchRequest(dispatcher, Request{StatusRequest{}});
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& status = std::get<StatusResponse>(resp);
REQUIRE(status.requestCounts.count("status_mem_rss_bytes") == 1);
CHECK(status.requestCounts.at("status_mem_rss_bytes") > 0);
#if defined(__APPLE__)
REQUIRE(status.requestCounts.count("status_mem_phys_footprint_bytes") == 1);
const auto footprintBytes = status.requestCounts.at("status_mem_phys_footprint_bytes");
CHECK(footprintBytes > 0);
const auto expectedMb = static_cast<double>(footprintBytes) / (1024.0 * 1024.0);
CHECK(std::abs(status.memoryUsageMb - expectedMb) < std::max(1.0, expectedMb * 0.05));
#endif
}
TEST_CASE("DaemonMetrics: repair idle and busy counters export from state",
"[daemon][metrics][repair]") {
StateComponent state;
state.stats.repairIdleTicks.store(3, std::memory_order_relaxed);
state.stats.repairBusyTicks.store(9, std::memory_order_relaxed);
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_repair_counter_export_");
ServiceManager svc(cfg, state, lifecycleFsm);
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
CHECK(snap->repairIdleTicks == 3);
CHECK(snap->repairBusyTicks == 9);
}
TEST_CASE("StatusResponse: post_ingest_rpc requestCounts keys round-trip",
"[daemon][status][protocol][post_ingest]") {
StatusResponse s{};
s.requestCounts["post_ingest_rpc_queued"] = 3;
s.requestCounts["post_ingest_rpc_capacity"] = 64;
s.requestCounts["post_ingest_rpc_max_per_batch"] = 8;
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const auto& resp = std::get<Response>(dec.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& decoded = std::get<StatusResponse>(resp);
REQUIRE(decoded.requestCounts.at("post_ingest_rpc_queued") == 3);
REQUIRE(decoded.requestCounts.at("post_ingest_rpc_capacity") == 64);
REQUIRE(decoded.requestCounts.at("post_ingest_rpc_max_per_batch") == 8);
}
TEST_CASE("StatusResponse: maintenance phase fields round-trip",
"[daemon][status][protocol][maintenance]") {
StatusResponse s{};
s.databasePhase = "ready";
s.databasePhaseElapsedMs = 11;
s.maintenancePhase = "salvaging";
s.maintenancePhaseElapsedMs = 222;
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const auto& resp = std::get<Response>(dec.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& decoded = std::get<StatusResponse>(resp);
CHECK(decoded.databasePhase == "ready");
CHECK(decoded.databasePhaseElapsedMs == 11);
CHECK(decoded.maintenancePhase == "salvaging");
CHECK(decoded.maintenancePhaseElapsedMs == 222);
}
TEST_CASE("StatusResponse: search maintenance provenance round-trips",
"[daemon][status][protocol][search][config]") {
StatusResponse status{};
status.searchAutomaticRebuildsEnabled = false;
status.searchAutomaticRebuildsSource = "compatibility-environment:YAMS_DISABLE_SEARCH_REBUILDS";
status.runtimeTuning = {{"ipc.timeout_ms", "4321"},
{"ipc.timeout_ms.source", "config:tuning.ipc.timeout_ms"}};
Message message{};
message.payload = Response{std::in_place_type<StatusResponse>, status};
const auto encoded = ProtoSerializer::encode_payload(message);
REQUIRE(encoded.has_value());
const auto decodedMessage = ProtoSerializer::decode_payload(encoded.value());
REQUIRE(decodedMessage.has_value());
const auto& response = std::get<Response>(decodedMessage.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(response));
const auto& decoded = std::get<StatusResponse>(response);
CHECK_FALSE(decoded.searchAutomaticRebuildsEnabled);
CHECK((decoded.searchAutomaticRebuildsSource ==
"compatibility-environment:YAMS_DISABLE_SEARCH_REBUILDS"));
CHECK((decoded.runtimeTuning.at("ipc.timeout_ms") == "4321"));
CHECK((decoded.runtimeTuning.at("ipc.timeout_ms.source") == "config:tuning.ipc.timeout_ms"));
}
TEST_CASE("StatusResponse: freshness requestCounts keys round-trip",
"[daemon][status][protocol][freshness]") {
StatusResponse s{};
s.requestCounts["status_snapshot_age_ms"] = 42;
s.requestCounts["status_snapshot_stale"] = 0;
s.requestCounts["status_detail_age_ms"] = 17;
s.requestCounts["status_detail_stale"] = 1;
s.requestCounts["status_detail_available"] = 1;
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const auto& resp = std::get<Response>(dec.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& decoded = std::get<StatusResponse>(resp);
REQUIRE(decoded.requestCounts.at("status_snapshot_age_ms") == 42);
REQUIRE(decoded.requestCounts.at("status_snapshot_stale") == 0);
REQUIRE(decoded.requestCounts.at("status_detail_age_ms") == 17);
REQUIRE(decoded.requestCounts.at("status_detail_stale") == 1);
REQUIRE(decoded.requestCounts.at("status_detail_available") == 1);
}
TEST_CASE("StatusResponse: backpressure requestCounts keys round-trip",
"[daemon][status][protocol][post_ingest]") {
StatusResponse s{};
s.requestCounts["post_ingest_backpressure_rejects"] = 123;
s.requestCounts["kg_jobs_depth"] = 3880;
s.requestCounts["kg_jobs_capacity"] = 4096;
s.requestCounts["kg_jobs_fill_pct"] = 95;
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const auto& resp = std::get<Response>(dec.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& decoded = std::get<StatusResponse>(resp);
REQUIRE(decoded.requestCounts.at("post_ingest_backpressure_rejects") == 123);
REQUIRE(decoded.requestCounts.at("kg_jobs_depth") == 3880);
REQUIRE(decoded.requestCounts.at("kg_jobs_capacity") == 4096);
REQUIRE(decoded.requestCounts.at("kg_jobs_fill_pct") == 95);
}
TEST_CASE("StatusResponse: memory breakdown requestCounts keys round-trip",
"[daemon][status][protocol][memory]") {
StatusResponse s{};
s.requestCounts["status_mem_rss_bytes"] = 123456;
s.requestCounts["status_mem_phys_footprint_bytes"] = 654321;
s.requestCounts["status_mem_malloc_allocated_bytes"] = 777777;
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const auto& resp = std::get<Response>(dec.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& decoded = std::get<StatusResponse>(resp);
REQUIRE(decoded.requestCounts.at("status_mem_rss_bytes") == 123456);
REQUIRE(decoded.requestCounts.at("status_mem_phys_footprint_bytes") == 654321);
REQUIRE(decoded.requestCounts.at("status_mem_malloc_allocated_bytes") == 777777);
}
TEST_CASE("AbiPluginHost: failed plugin loads are retained in last scan skips",
"[daemon][plugin][status][load-failure]") {
#if defined(__APPLE__)
const std::filesystem::path pluginPath{"/opt/homebrew/lib/yams/plugins/yams_glint.dylib"};
if (!std::filesystem::exists(pluginPath)) {
SKIP("Installed yams_glint plugin not present");
}
AbiPluginHost host;
const auto trustFile = makeTempDir("yams_glint_skip_status_") / "plugins.trust";
host.setTrustFile(trustFile);
REQUIRE(host.trustAdd(pluginPath.parent_path()));
const auto loadResult = host.load(pluginPath, "{}");
if (loadResult) {
SUCCEED("yams_glint loaded successfully on this machine");
REQUIRE(host.unload(loadResult.value().name));
return;
}
const auto skips = host.getLastScanSkips();
REQUIRE_FALSE(skips.empty());
const auto it = std::find_if(skips.begin(), skips.end(),
[&](const auto& entry) { return entry.first == pluginPath; });
REQUIRE(it != skips.end());
CHECK((it->second.find("dlopen failed") != std::string::npos ||
it->second.find("preflight failed") != std::string::npos ||
it->second.find("Plugin init failed") != std::string::npos));
#else
SUCCEED("macOS-specific plugin load retention test");
#endif
}
// =============================================================================
// StatusResponse Protocol Serialization Tests
// =============================================================================
TEST_CASE("StatusResponse: Protocol serialization", "[daemon][status][protocol]") {
SECTION("Round-trip with lifecycle state and error") {
StatusResponse s{};
s.running = true;
s.ready = false;
s.uptimeSeconds = 42;
s.requestsProcessed = 7;
s.activeConnections = 1;
s.memoryUsageMb = 12.5;
s.cpuUsagePercent = 3.2;
s.version = "test";
s.overallStatus = "failed";
s.lifecycleState = "starting";
s.lastError = "boom";
s.dataDir = "/tmp/daemon-data-dir";
s.metadataDbPath = "/tmp/daemon-data-dir/yams.db";
s.vectorDbPath = "/tmp/daemon-data-dir/vectors.db";
s.requestCounts["worker_threads"] = 4;
// FSM-exported fields
s.requestCounts["service_fsm_state"] = 3;
s.requestCounts["embedding_state"] = 2;
s.requestCounts["plugin_host_state"] = 1;
// Readiness states
s.readinessStates["embedding_ready"] = false;
s.readinessStates["plugins_ready"] = true;
s.readinessStates["plugins_degraded"] = true;
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
// Encode
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
// Decode
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const Message& out = dec.value();
REQUIRE(std::holds_alternative<Response>(out.payload));
const auto& resp = std::get<Response>(out.payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
const auto& decoded = std::get<StatusResponse>(resp);
// Verify key fields survived round-trip
REQUIRE(decoded.running == true);
REQUIRE(decoded.ready == false);
REQUIRE(decoded.uptimeSeconds == 42);
REQUIRE(decoded.version == "test");
REQUIRE(decoded.overallStatus == "failed");
REQUIRE(decoded.lastError == "boom");
REQUIRE(decoded.dataDir == "/tmp/daemon-data-dir");
REQUIRE(decoded.metadataDbPath == "/tmp/daemon-data-dir/yams.db");
REQUIRE(decoded.vectorDbPath == "/tmp/daemon-data-dir/vectors.db");
// Verify FSM state counts
REQUIRE(decoded.requestCounts.at("worker_threads") == 4);
REQUIRE(decoded.requestCounts.at("service_fsm_state") == 3);
// Verify readiness flags
REQUIRE(decoded.readinessStates.at("embedding_ready") == false);
REQUIRE(decoded.readinessStates.at("plugins_ready") == true);
REQUIRE(decoded.readinessStates.at("plugins_degraded") == true);
}
SECTION("Empty status response serializes correctly") {
StatusResponse s{};
Message m{};
m.payload = Response{std::in_place_type<StatusResponse>, s};
auto enc = ProtoSerializer::encode_payload(m);
REQUIRE(enc.has_value());
auto dec = ProtoSerializer::decode_payload(enc.value());
REQUIRE(dec.has_value());
const auto& resp = std::get<Response>(dec.value().payload);
REQUIRE(std::holds_alternative<StatusResponse>(resp));
}
}
// =============================================================================
// Plugin Degradation Tests
// =============================================================================
// Note: Full integration tests (DISABLED in original) are skipped here
// These would test daemon start with trust failures and recovery hooks
// They belong in integration tests due to full daemon lifecycle requirements
TEST_CASE("ServiceManager: Plugin degradation tracking", "[daemon][status][plugins]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_plugin_degrade_");
SECTION("Plugin degradation flag is tracked in FSM") {
// This is a unit test of the FSM tracking mechanism
// Full daemon integration tests are separate
ServiceManager svc(cfg, state, lifecycleFsm);
// ServiceManager should track plugin states via FSM
// Actual degradation testing requires daemon start, which is integration level
// Basic verification: ServiceManager exists and has FSM snapshot capability
auto snapshot = svc.getServiceManagerFsmSnapshot();
// ServiceManager starts uninitialized until explicitly initialized
REQUIRE(snapshot.state == ServiceManagerState::Uninitialized);
}
}
// =============================================================================
// FSM State Metrics
// =============================================================================
TEST_CASE("DaemonMetrics: FSM state export", "[daemon][metrics][fsm]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_fsm_metrics_");
ServiceManager svc(cfg, state, lifecycleFsm);
SECTION("Lifecycle FSM state is tracked") {
lifecycleFsm.dispatch(BootstrappedEvent{});
auto snapshot = lifecycleFsm.snapshot();
REQUIRE(snapshot.state != LifecycleState::Unknown);
}
SECTION("Service manager FSM state is accessible") {
auto snapshot = svc.getServiceManagerFsmSnapshot();
// ServiceManager starts in Uninitialized state until initialized
REQUIRE(snapshot.state == ServiceManagerState::Uninitialized);
}
SECTION("Embedding provider FSM state is accessible") {
auto snapshot = svc.getEmbeddingProviderFsmSnapshot();
// Embedding provider starts in Unavailable state until plugin loaded
REQUIRE(snapshot.state == EmbeddingProviderState::Unavailable);
}
SECTION("Plugin host FSM state is accessible") {
auto snapshot = svc.getPluginHostFsmSnapshot();
// When ABI plugins are disabled, PluginManager initializes successfully with
// an empty plugin list, transitioning to Ready state. When plugins are enabled
// but PluginManager hasn't been created yet, the fallback is NotInitialized.
REQUIRE((snapshot.state == PluginHostState::NotInitialized ||
snapshot.state == PluginHostState::Ready));
}
}
// =============================================================================
// DatabaseManager Metrics Tests
// =============================================================================
TEST_CASE("DatabaseManager: Metrics tracking", "[daemon][metrics][database]") {
StateComponent state;
DatabaseManager::Dependencies deps{.state = &state};
DatabaseManager dbm(deps);
SECTION("Stats start at zero") {
const auto& stats = dbm.getStats();
REQUIRE(stats.openDurationMs.load() == 0);
REQUIRE(stats.migrationDurationMs.load() == 0);
REQUIRE(stats.openErrors.load() == 0);
REQUIRE(stats.migrationErrors.load() == 0);
REQUIRE(stats.repositoryInitErrors.load() == 0);
}
SECTION("Stats are returned by reference") {
// Verify we can access stats multiple times
const auto& stats1 = dbm.getStats();
const auto& stats2 = dbm.getStats();
REQUIRE(stats1.openErrors.load() == stats2.openErrors.load());
}
}
TEST_CASE("DaemonMetrics: DatabaseManager metrics export", "[daemon][metrics][database]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_db_metrics_");
ServiceManager svc(cfg, state, lifecycleFsm);
SECTION("DatabaseManager metrics are included in snapshot") {
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
// DatabaseManager metrics should be present (may be 0 if not initialized)
REQUIRE(snap->dbOpenErrors == 0); // Default value when no errors
}
}
// =============================================================================
// WorkCoordinator Metrics Tests
// =============================================================================
TEST_CASE("WorkCoordinator: Metrics tracking", "[daemon][metrics][work]") {
WorkCoordinator wc;
SECTION("Stats before start") {
auto stats = wc.getStats();
REQUIRE(stats.workerCount == 0);
REQUIRE(stats.activeWorkers == 0);
REQUIRE(stats.isRunning == false);
}
SECTION("Stats accessors don't crash") {
REQUIRE(wc.getWorkerCount() == 0);
REQUIRE(wc.getActiveWorkerCount() == 0);
REQUIRE(wc.isRunning() == false);
}
}
TEST_CASE("DaemonMetrics: WorkCoordinator metrics export", "[daemon][metrics][work]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_wc_metrics_");
ServiceManager svc(cfg, state, lifecycleFsm);
SECTION("WorkCoordinator metrics are included in snapshot") {
DaemonMetrics metrics(nullptr, &state, &svc, svc.getWorkCoordinator());
auto snap = metrics.getSnapshot();
REQUIRE(snap != nullptr);
// WorkCoordinator metrics should be present
// ServiceManager creates WorkCoordinator with hardware_concurrency() workers
REQUIRE(snap->workerThreads > 0); // Should have workers (worker_threads)
REQUIRE(snap->workCoordinatorRunning == true); // Should be running
REQUIRE(snap->workCoordinatorProgressProbesPosted > 0);
CHECK(snap->workCoordinatorProgressProbesCompleted <=
snap->workCoordinatorProgressProbesPosted);
CHECK(snap->vectorCheckpointPhase == 0); // VectorCheckpointPhase::Idle
}
}
// =============================================================================
// Search Load Metrics Tests
// =============================================================================
TEST_CASE("ServiceManager: Search load metrics", "[daemon][metrics][search]") {
StateComponent state;
DaemonLifecycleFsm lifecycleFsm;
DaemonConfig cfg;
cfg.dataDir = makeTempDir("yams_search_metrics_");
ServiceManager svc(cfg, state, lifecycleFsm);
SECTION("Search load metrics return valid structure") {
auto metrics = svc.getSearchLoadMetrics();
REQUIRE(metrics.active == 0);
REQUIRE(metrics.queued == 0);
REQUIRE(metrics.concurrencyLimit > 0);
}
SECTION("Search lifecycle counters update load metrics") {
svc.onSearchRequestQueued();
auto queued = svc.getSearchLoadMetrics();
REQUIRE(queued.queued == 1);
REQUIRE(queued.active == 0);
const auto cap = queued.concurrencyLimit;
REQUIRE(svc.tryStartSearchRequest(cap));
auto started = svc.getSearchLoadMetrics();
REQUIRE(started.queued == 0);
REQUIRE(started.active == 1);
svc.onSearchRequestFinished();
auto finished = svc.getSearchLoadMetrics();
REQUIRE(finished.active == 0);
REQUIRE(finished.queued == 0);
}
SECTION("Failed search start keeps queued count until rejection") {
// Pin cap to 1 so the second start is guaranteed to fail regardless of the
// machine-derived concurrencyLimit (varies with CPU cores).
constexpr std::uint32_t cap = 1;
svc.onSearchRequestQueued();
REQUIRE(svc.tryStartSearchRequest(cap));
svc.onSearchRequestQueued();
REQUIRE_FALSE(svc.tryStartSearchRequest(cap));
auto waiting = svc.getSearchLoadMetrics();
REQUIRE(waiting.active == 1);
REQUIRE(waiting.queued == 1);
svc.onSearchRequestRejected();
auto rejected = svc.getSearchLoadMetrics();
REQUIRE(rejected.active == 1);
REQUIRE(rejected.queued == 0);
svc.onSearchRequestFinished();
auto finished = svc.getSearchLoadMetrics();
REQUIRE(finished.active == 0);
REQUIRE(finished.queued == 0);
}
}
// =============================================================================
// FsmMetricsRegistry Integration Tests
// =============================================================================
TEST_CASE("FsmMetricsRegistry: Metrics collection", "[daemon][metrics][fsm]") {
// Reset registry to known state
FsmMetricsRegistry::instance().reset();
SECTION("Registry starts with zeros") {
auto snap = FsmMetricsRegistry::instance().snapshot();
REQUIRE(snap.transitions == 0);
REQUIRE(snap.headerReads == 0);
REQUIRE(snap.payloadReads == 0);
REQUIRE(snap.payloadWrites == 0);
REQUIRE(snap.bytesSent == 0);
REQUIRE(snap.bytesReceived == 0);
}
SECTION("Registry increments correctly") {
FsmMetricsRegistry::instance().incrementTransitions(5);
FsmMetricsRegistry::instance().incrementHeaderReads(3);
FsmMetricsRegistry::instance().incrementPayloadReads(2);
FsmMetricsRegistry::instance().addBytesSent(100);
FsmMetricsRegistry::instance().addBytesReceived(200);
auto snap = FsmMetricsRegistry::instance().snapshot();
REQUIRE(snap.transitions == 5);
REQUIRE(snap.headerReads == 3);
REQUIRE(snap.payloadReads == 2);
REQUIRE(snap.bytesSent == 100);
REQUIRE(snap.bytesReceived == 200);
}
SECTION("Registry reset works") {
FsmMetricsRegistry::instance().incrementTransitions(10);
FsmMetricsRegistry::instance().reset();
auto snap = FsmMetricsRegistry::instance().snapshot();
REQUIRE(snap.transitions == 0);
}
}
Reference in a new issue View git blame Copy permalink