Spaces:
Sleeping
Sleeping
Commit History
-
Embed extracted video frames in PDF report per room section
- caption_frames() now returns (path, caption) pairs - run_inventory_pipeline groups frames by room classification - build_pdf_report accepts room_frames dict and renders images after each room's item table with caption labels - Image scaling to 4.2x2.8 inches within page margins - Photo section heading per room (Photos - Living Room)20cb603 -
Fix empty report: redesign pipeline without LLM JSON
Core problem: flan-t5-small (80M params) cannot produce structured JSON. New approach: - Vision model captions each frame directly - Keyword-based room classification (no LLM needed) - Predefined room items (Door, Ceiling, Walls, etc.) matched to captions - LLM only generates executive summary text - All structured data built in Python This eliminates JSON parsing failures that caused empty reports.013d951 -
Redesign PDF report to match professional inventory report layout
- Cover page with property address, date, report type - Table of contents grouped by room - Per-room sections with dark blue header tables - Alternating row styling matching sample format - Declaration page with signature fields - Footer with page numbers on every page - Auto-assign room/section numbers from datadb35ed6 -
Remove CrewAI dependency, replace with sequential local LLM calls
6a6f2b4 -
Use CrewAI BaseLLM with local pipelines
08214f5 -
Use local open-source captioning + text models
226bc0b -
Resolve Groq vision model dynamically
38ba3ce
-
Add crewai-files dependency
16dd9a5 -
Switch to Groq + FFmpeg pipeline
e35c80f -
Resolve Gemini model aliases before CrewAI
77ba290 -
Use CrewAI LLM provider for Gemini
ab0f999 -
Fix google-genai upload/delete call
b2f4056 -
Fix CrewAI process and cleanup timing
9f50f55 -
Fix tenacity retry predicate
1bb2158 -
Pin tenacity to 8.4.2 for CrewAI deps
98e33b6 -
Align FastAPI with CrewAI deps
e320610 -
Initial AutoInventory MVP
5bd1bf0