Embed extracted video frames in PDF report per room section

Fix empty report: redesign pipeline without LLM JSON

Redesign PDF report to match professional inventory report layout

Remove CrewAI dependency, replace with sequential local LLM calls

Use CrewAI BaseLLM with local pipelines

Use local open-source captioning + text models

Resolve Groq vision model dynamically

Add crewai-files dependency

Switch to Groq + FFmpeg pipeline

Resolve Gemini model aliases before CrewAI

Use CrewAI LLM provider for Gemini

Fix google-genai upload/delete call

Fix CrewAI process and cleanup timing

Fix tenacity retry predicate

Pin tenacity to 8.4.2 for CrewAI deps

Align FastAPI with CrewAI deps

Initial AutoInventory MVP