update entrypoint

update dockerfile and requirements in root

Switch to Qwen-1.5-1.8B-Chat - verified multilingual model with good Indic support

Switch to AI4Bharat IndicLLM - better support for 11 Indic languages

Force rebuild with label

Use bitsandbytes 4-bit quantization instead of AirLLM (more stable)

Install airllm explicitly in Dockerfile

Fix cache directory permissions for non-root user

Use AirLLM 4-bit quantization for Sarvam-1 (uses ~1.5GB RAM)

Switch to TinyLlama-1.1B with float16 for lower memory

Fix torch_dtype deprecation in Dockerfile

Add HF_TOKEN build arg for model downloads

Fix Dockerfile: single-line model download

Pre-download full Sarvam-1 model in Dockerfile

Pre-download models in Dockerfile, use cache at runtime

Fix: Run backend on port 7860 for HF Spaces

Minimal backend to test container startup

Add detailed logging for debugging startup

Add healthcheck and better startup logging

Simplify: backend-only for HF Spaces, frontend on Netlify

Add root Dockerfile combining backend and frontend