Spaces:
Paused
Paused
Commit History
-
feat: help, document viewer, knowledge graph; empty start + language-matched demo
DocUAcc02af3 -
Real OCR on image upload via cloud ZeroGPU Lab (text/seed stay deterministic)
- LMC_IMAGE_OCR decouples photo OCR from the text extractor: /ingest/image routes to the ZeroGPU NuExtract3 Lab (real image->markdown + KVT4 facts), while seeded card + text paste + safety stay on deterministic stub (instant load, zero cloud unless a photo is explicitly uploaded) - pipeline.ingest gains a per-doc extractor override (no change to card default) - normalize fact timestamps to tz-aware UTC in observation_to_kvt (fixes naive-vs-aware crash when cloud-OCR facts merge into stub-seeded series) - Dockerfile: add gradio_client, set LMC_IMAGE_OCR=ocrlab Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>97a786e
-
Deploy LiveMedCard demo (Docker, stub, seeded bilingual card)
Verified signature38622c4