jlightner fd1fd6c6f9 fix: Pipeline LLM audit — temperature=0, realistic token ratios, structured request_params ... Audit findings & fixes: - temperature was never set (API defaulted to 1.0) → now explicit 0.0 for deterministic JSON - llm_max_tokens=65536 exceeded hard_limit=32768 → aligned to 32768 - Output ratio estimates were 5-30x too high (based on actual pipeline data): stage2: 0.6→0.05, stage3: 2.0→0.3, stage4: 0.5→0.3, stage5: 2.5→0.8 - request_params now structured as api_params (what's sent to LLM) vs pipeline_config (internal estimator settings) — no more ambiguous 'hard_limit' in request params - temperature=0.0 sent on both primary and fallback endpoints		2026-04-01 07:20:09 +00:00
..
__init__.py	fix: restore complete project tree from ub01 canonical state	2026-03-31 02:10:41 +00:00
embedding_client.py	fix: restore complete project tree from ub01 canonical state	2026-03-31 02:10:41 +00:00
llm_client.py	fix: Pipeline LLM audit — temperature=0, realistic token ratios, structured request_params	2026-04-01 07:20:09 +00:00
qdrant_client.py	fix: Parallel search with match_context, deterministic Qdrant IDs, raised embedding timeout	2026-04-01 06:54:34 +00:00
schemas.py	fix: Moment-to-page linking via moment_indices in stage 5 synthesis	2026-04-01 00:34:37 -05:00
stages.py	fix: Pipeline LLM audit — temperature=0, realistic token ratios, structured request_params	2026-04-01 07:20:09 +00:00