Add httpx dep for Groq/Gemini proxy support

Update inference proxy (Groq/Gemini/CPU priority chain)

Fix: quote uvicorn[standard], split torch pip install

Add inference server app.py

Switch to transformers fallback (TGI CPU unsupported)

Add --trust-remote-code for Qwen2.5

Deploy Qwen2.5-1.5B-Instruct TGI (persona synthesis backend)

Deploy Qwen2.5-1.5B-Instruct TGI (persona synthesis backend)

initial commit