8 Commits

Include renames

Author	SHA1	Message	Date
woogi	d431a026ce	feat: treat disputed extracted values as unverified in constructability and critic prompts	2026-08-10 10:56:53 -05:00
woogi	8631a26006	feat: surface disputed extracted values to critic and specialist prompts	2026-08-10 10:47:34 -05:00
woogi	df4d15fd0c	feat: deterministic disputed-value detection for cluster assertions	2026-08-10 10:40:10 -05:00
woogi	3d7fce7bf9	Kill extract-wave truncation: 65k ceiling, hard thinking budget, reasoning-token telemetry ... Job 98194fa8d215 showed every extract call hitting the 32k cap with only ~20k chars visible despite reasoning effort=low - Gemini 2.5 Pro still burned ~25k thinking tokens per sheet. - EXTRACT_MAX_TOKENS default 32768 -> 65536 (model output ceiling) - new EXTRACT_REASONING_MAX_TOKENS (default 2048): OpenRouter reasoning max_tokens / Gemini thinking_budget; takes precedence over effort - log per-call reasoning token counts (usage.completion_tokens_details) and include thinking count in the finish_reason=length marker	2026-08-09 08:17:41 -05:00
woogi	76e0a52658	Fix sheet-extraction page loss: bare-list wrap, compact retry, reasoning cap ... - SheetExtractorAgent accepts top-level array responses as the objects array instead of discarding them (recovered the failure mode behind 16/38 failed sheets on job 475a6f184dd1) - Second-chance compact retry per page before declaring extraction failed - call_json: reasoning_effort param (cloud-only extra_body), finish_reason capture + explicit max_tokens log line, finish_reason in raw dumps, cache key covers reasoning_effort - EXTRACT_MAX_TOKENS default 16384 -> 32768 (Gemini thinking tokens count against the cap); new EXTRACT_REASONING_EFFORT=low default for extract - tests: 5 new fallback-ladder tests	2026-08-07 07:35:53 -05:00
woogi	7488cf68c5	Verbose per-call LLM logging, raw request/response dumps, and end-of-log cost summary. ... - [LLM] line per call: stage, model, prompt size, output size, cost, parsed item counts - LLM_RAW_DUMP: full prompt/response JSON per call under outputs/<job>/llm_raw/ - Cost block at tail of job.log (per-stage, per-model, cached vs live) - Agent mode: reset llm cost counters per job; review finalization now teed into job.log + dumps	2026-08-05 15:17:17 -05:00
John Wilganowski	1c1d2ff21b	Add required human review gate to the Agent pipeline. ... Agent web jobs now stop after Brain consolidation and enter needs_review with a persisted review queue (blocking: high-severity, low-confidence, sensitive-category findings; audit sample of clean clusters). Humans decide confirm/reject/unsure/needs_clarification via new review API and frontend queue; a finalizer applies decisions (rejections suppressed with reason codes), performs bounded targeted reruns for clarifications, drafts RFIs only for kept issues, and only then marks the job done and sends the final email. Two-phase email (review-required, then final report), per-decision feedback labels with redacted aggregate metrics, restart recovery from job artifacts, and CLI --no-review bypass. Classic pipeline unchanged. 65 non-LLM tests.	2026-07-28 19:23:57 +00:00
woogiandCursor	82a48d99cf	Add scoped Agent-mode pipeline as experimental Classic fork. ... Wire specialist waves, Brain consolidation, and Classic-compatible reports so Agent mode can run end-to-end via OpenRouter without changing the default Classic path. Co-authored-by: Cursor <cursoragent@cursor.com>	2026-07-18 14:31:13 +00:00