Commit graph

4 commits

Author SHA1 Message Date
Trevon
771deebc84
fix: harden CI and metadata stability (#48)
* feat: dead code removal, profling foudnation improvements, addressing code smells in reference db implementation

* fix: test fixes and optimmizations for benchmarks

* fix: harden audit findings

   Reject invalid tokenizer ids, fix move-order hazards, bound manifest
   and audio IO, and guard numeric/process edge cases found by Fold.

* fix: route SymSpell writes through coordinator

* test: fix batch content initializers

* fix(ci): restore cross-platform builds

* fix: address Copilot review findings
2026-06-11 16:42:54 -04:00
Trevon
6d8877a6d4
feat(search): add Simeon bandit tuning (#33)
* feat(search): per-profile simeon bandit with training-free self-tuning

* fix(zyp): free parallel buffers correctly

* fix: code cleanup and simeon version bump

* fix(search): guard Simeon bandit state
2026-05-17 04:08:50 -04:00
trvon
65564158a9
perf(zyp): true parallel PDF extraction + throughput benchmark
- Implement Document::extractAllParallelized() using std::async
  to distribute pages across threads with per-thread Document
  instances, achieving true parallelism (not the old fake alias)

- zyp_plugin.cpp now uses extractAllParallelized() by default
  for PDF extraction during ingestion

- Add zyp_extraction_bench standalone benchmark measuring
  sequential vs 2/4/N-thread extraction throughput (pages/s, MB/s)

Build: ninja bench_zyp
Run:   build/debug/plugins/zyp/zyp_extraction_bench <pdf>
2026-05-15 12:13:33 -04:00
trvon
bc71d36534
migrated pdf plugin, daemon shutdown logic fix 2025-12-31 14:46:07 -05:00