Commit graph

529 commits

Author SHA1 Message Date
trvon
94a49472f1
feat: fixing bug with multiple processes access gpu, read/write pools now added in metadata repository, and updated plugins for opencode and vscode 2026-02-14 18:16:57 -05:00
trvon
8f38ee76be
feat: updating metrics for search and retrieval for future learning system. Additional tuning for search performance 2026-02-13 21:09:26 -05:00
trvon
8256917e5e
feat: Hardening moodel logic for embedings. Adding trace-linked feedback event persistence checkpoint 2026-02-12 23:42:21 -05:00
trvon
11ba59f4bb
feat: updating embedding default checking to paragraph 2026-02-12 11:55:59 -05:00
trvon
b40b60f054
fix: bug fixes for resource limiting and onnxruntime model shapes for embeddings 2026-02-12 01:52:33 -05:00
trvon
a44c2fb4af
feat: implementing embedding chunking strategy and two staged ingestion that processes then distributes document data 2026-02-12 00:46:04 -05:00
trvon
44a94eabe1
feat: fuzzing and config expansion. Moving to optimizing ingestion. Adding tunable embedding (chunk size etc.) 2026-02-11 23:21:26 -05:00
trvon
4ecdbe7d73
fix: moved SocketServer read/write into dedicated IOCoordinator and fixed proxy socket paths resolution logic. Fixed TSAN race in SymSpell init 2026-02-11 14:40:23 -05:00
trvon
cc28b8a343
feat: instrumented embedding for deeper insight into hotspots. Fixed CoreML scaling that resulted in high ram usage due to VRAM calculation. Expanded testing for embedding and draining of work 2026-02-11 12:07:08 -05:00
trvon
a0fd82e25c fix: Fixing issues that break windows release build. This should allow new msvc version Working on debug builds. 2026-02-10 15:03:39 -08:00
trvon
04b55be44e
fix: apply tuning logic for profiles to GPU 2026-02-10 15:49:26 -05:00
trvon
ff77c6b304
fix: improving socket and chunking reliability and robustness. Added better regression tests and errror logging logic for state connections 2026-02-10 12:25:30 -05:00
trvon
2e966c4f1b
fix: windows build fixes 2026-02-09 23:26:33 -05:00
trvon
a177027320
fix: house cleaning 2026-02-09 21:12:26 -05:00
trvon
925426426a
feat: expanding mcp UI support, improvements for slots for component usage and other socket server stabilizations 2026-02-09 20:02:26 -05:00
trvon
cd7c698e07
feat: stability improvements for document ingestion 2026-02-09 19:14:45 -05:00
trvon
0c8a5ee451
fix: adding cap at onnx concurrency limit 2026-02-09 17:09:53 -05:00
trvon
1d255d59ce
feat: address bugs with scaling, batching in daemon processing 2026-02-09 17:07:00 -05:00
trvon
cc1212bb1f
fix: fixing pipeline ingestion of kg and embeddings. Still finding stalls and will continue to investigate 2026-02-09 10:39:48 -05:00
trvon
4d6c163c9d
feat: adding drain to constrexpr metrics 2026-02-09 01:45:38 -05:00
trvon
0c8ffea321
feat: improving ingestion speed with kg batch and updating metrics to use constexpr 2026-02-09 01:33:24 -05:00
trvon
c64a0b239d
feat: adding fuzzing updates and fixes in ipc 2026-02-08 22:20:47 -05:00
trvon
b2c6121990
Merge branch 'main' of github.com:trvon/yams 2026-02-08 22:19:55 -05:00
trvon
1e0ca1566c
fix: fixing bind issue when generating large batches in repair 2026-02-08 22:19:15 -05:00
trvon
fa3d15eedf
Merge branch 'main' of github.com:trvon/yams 2026-02-08 22:00:46 -05:00
trvon
f4f3c055fe
feat: benchmark expansion and normalization. Imrpoving metrics for repair and models 2026-02-08 20:35:39 -05:00
trvon
c83110319d
feat: benchmark, model testing / enhancements, insertion performance updates 2026-02-08 18:33:29 -05:00
trvon
1a0667d28f
fix: improving graph repair 2026-02-08 14:51:57 -05:00
trvon
ac843d1e93
Merge branch 'main' of github.com:trvon/yams 2026-02-08 14:06:09 -05:00
trvon
a3522b8b22
fix: expose gpu usage, add command imrpovments and other bug fixes 2026-02-08 14:04:57 -05:00
trvon
c4a7992122
fix: repairmanager, post ingest and pipeline fixes 2026-02-08 14:02:57 -05:00
trvon
430057f4bf
fix: SocketServer slots for proxy and cli requests, fixed tests and increase doctor request handling 2026-02-08 13:02:08 -05:00
trvon
3c09becde5
fix: repair improvements and socketserver hardening 2026-02-07 23:37:32 -05:00
trvon
448f5e47b4
fix: fixing repairmanager logic, windows build issues, linting, resource scaling 2026-02-07 22:07:56 -05:00
trvon
3fd243f687
fix: repair command improvements and optimizations 2026-02-06 10:15:50 -05:00
trvon
90d28d722e
fix: docker build fixes addition clamping logic for onbound memory group 2026-02-06 09:27:34 -05:00
trvon
df1348f3c7
fix: tuning improvements and better tests. Now working address memory issue of huge queue when ingesting allot of files 2026-02-06 04:31:09 -05:00
trvon
d960839e54
fix: embeddings optimizations 2026-02-05 21:27:12 -05:00
trvon
ab06614505
fix(repair): fix SIGSEGV crash, embedding failures, and runaway repair loops
- Fix writer_drain() SIGSEGV by setting connection_closing_ on write failure
- Add InternalError/ResourceExhausted to embedding retry list
- Fix document-level embedding chunk_id (0→999999) and level mismatch
- Add circuit breaker to bail out after consecutive/rate-threshold failures
- Add semaphore concurrency limiter (max 2) for repair embed requests
- Replace silent catch-all with error-logging catch blocks
2026-02-05 21:15:26 -05:00
trvon
fc4b5ceffb
fix: fixing metrics system and removing deadcode 2026-02-05 19:41:50 -05:00
trvon
e0612298d3
fix: fixing rocm builds, removing dead code and some optimizations 2026-02-05 18:33:54 -05:00
trvon
ded16c11d4
improvements of parallel ingestion 2026-02-05 11:51:24 -05:00
trvon
8a06575a0c
fix: reverted bad change. moved embedding generation to repair coordinator. Still tuning resource usage and governors 2026-02-05 10:21:25 -05:00
trvon
a9f8a3dd63
Revert "reverting resource changes"
This reverts commit b35c84cb2e.
2026-02-05 00:18:42 -05:00
trvon
1af0356419
adding back governor code 2026-02-05 00:16:01 -05:00
trvon
b35c84cb2e
reverting resource changes 2026-02-05 00:11:56 -05:00
trvon
f2a9adbd42
revert: test fixes 2026-02-05 00:09:31 -05:00
trvon
4b2185dbcf
feat(governance): implement Netflix Gradient2 adaptive concurrency limiters
Replace CPU-threshold-based scaling with latency-feedback-based adaptive
concurrency control using the Netflix Gradient2 algorithm.

Core Changes:
- Add GradientLimiter class implementing Netflix Gradient2 algorithm
- Integrate limiters into all 6 post-ingest stages (extraction, kg, symbol,
  entity, title, embed)
- Add TuneAdvisor configuration knobs for gradient limiter tuning
- Simplify ResourceGovernor to emergency-only backstop (CPU > 95%, mem > 95%)
- Update TuningManager to read dynamic limits from gradient limiters
2026-02-04 23:38:33 -05:00
trvon
b071112836
feat: test coverage improvements. Scaling bug addresses 2026-02-04 20:20:18 -05:00
trvon
14c35926f1
test improvements and bug fixes in parser, scaling logic, and mcp tools 2026-02-04 14:42:28 -05:00