4 Commits

Author	SHA1	Message	Date
Ty	41401a7266	fix(inference): #87 redact image bytes from inference_log ... Log full text + image_ref (photo_id/seq/bytes) only; wire path still sends data-URLs. Call 1/2 SELECT id+seq for refs. Cap is safety net after redaction (METADATA-66).	2026-09-20 23:01:38 -07:00
Ty	7287a55707	fix(inference): #86 tighten Call 1 prompt and closed lists ... Ship locked expert Call 1 seed: order cats before edge, species=best confidence, mill-count edge rules, Wild for cookies/burl, multi figure, feature_tag_ids only (no free-text bags, no grade). LISTS keys match the seed; feat- prefix stripped for the model. Gate and review UI apply feature_tag_ids. Seed auto-activates on deploy when still seed-authored.	2026-09-20 22:54:36 -07:00
Ty	81be17d5cd	fix(inference): METADATA-35 fit Call 1/2 images to 1024 long edge ... Downscale processed PNGs only on the inference send path (keep alpha, never upscale). Stored inventory/Woo blobs unchanged. Adds Pillow + unit/e2e-style tests; cite tests on METADATA-35.	2026-09-20 22:32:14 -07:00
Ty	8a8a89ed3e	feat(inference): S7 Call 1/Call 2 + prompt revision REST + test-inference + Generate text UI ... - backend/inference.py: prompt seed (rev 0) + append-only revision REST (POST /prompts/{call}/revisions, /activate, GET; METADATA-105/106). Call 1 vision classification (1024px processed photos + synced taxonomy, 0.7 confidence gate, METADATA-35/40/45), Call 2 long description (user-triggered, METADATA-55/60/65), inference_log per call (METADATA-66), inference-off server_description (METADATA-70), test-inference probe (SETTINGS-45). Stdlib urllib, no new deps. - frontend: create/inference/GenerateText.svelte (control visible only when content enabled + endpoint reachable, METADATA-110; Call 2 + Call 1 buttons) + dev toggle. - tests: 6 inference tests (seed+revision, call1 gate, call2 stores desc, off->409, server desc, test-inference unconfigured). Verified: pytest 54 passed; gates exit 0; npm build ok.	2026-09-19 17:04:01 -07:00