All issues

Extract `BaseAPIDescriptionScorer` shared by `LLMDescriptionScorer` and `TypeSafeDescriptionScorer`

#357

· voorhs opened

on Sep 19, 2026

Advisor treats `description_llm` / `description_typesafe` as embedder consumers and estimates no API cost for them

#356

· voorhs opened

on Sep 19, 2026

`LLMDescriptionScorer` prompt puts the utterance before the descriptions, defeating provider prefix caching

#355

· voorhs opened

on Sep 19, 2026

`LLMDescriptionScorer` runs cache hits through the `max_per_second` limiter — 10 s per 100 cached utterances

#354

· voorhs opened

on Sep 19, 2026

API scorers: `predict()` fails inside a running event loop; `LLMDescriptionScorer` leaks an unclosed loop per instance on Python 3.14

#353

· voorhs opened

on Sep 19, 2026

`OptimizationConfig.seed` is `PositiveInt` — `seed=0` is rejected while `Pipeline(seed=0)` accepts it

good first issue

#352

· voorhs opened

on Sep 19, 2026

`Generator` lets `openai.APIStatusError` escape → one HTTP 4xx/5xx call aborts the whole `description_llm` run

#351

· voorhs opened

on Sep 19, 2026

Cache writes are not atomic (embeddings + structured outputs) → an interrupted/concurrent write poisons future reads

#335

· voorhs opened

on Jun 22, 2026

BertScorer multilabel fine-tuning collapses on sparse data (no pos_weight/focal, no warmup, brittle default LR)

#324

· voorhs opened

on Jun 20, 2026

BertScorer early-stopping default (scoring_f1@0.5) collapses on sparse multilabel

#323

· voorhs opened

on Jun 20, 2026

Upgrade transformers to 5.x (drops conftest mistral monkey-patch)

#295

· voorhs opened

on Jun 5, 2026

Устройство dump и load у Embedder и VectorIndex

#265

· voorhs opened

on Sep 20, 2025