Update engine.py

Update engine.py

Update app.py

Update engine.py

Update Dockerfile

Update engine.py

Update engine.py

Update Dockerfile

Update engine.py

Update app.py

Update Dockerfile

Update app.py

Update engine.py

Update Dockerfile

Update app.py

Update engine.py

Update Dockerfile

Update app.py

Update engine.py

Update engine.py

Update engine.py

Update engine.py

Update engine.py

Update Dockerfile

Update engine.py

Update Dockerfile

Update engine.py

Update engine.py

Update app.py

Update engine.py

Update app.py

Update engine.py

Improve app: query expansion before encoding, /debug endpoint, better error handling

Improve RAG: query expansion, confidence scoring, graceful 3-tier fallback, better prompting

Clean Dockerfile: use openai client (no version conflicts)"

Fix: use OpenAI client with HF router URL (no provider param needed, works with any version)"

Fix: add httpcore/httpx deps for huggingface_hub 1.x"

Fix: 2-stage install — compatible base then force-upgrade huggingface_hub"

Fix version conflict: transformers==4.48.0 + huggingface_hub>=1.0"

Fix: bump huggingface_hub to 0.26+ for provider support"

HARD FIX: pin exact compatible versions (transformers 4.44, sentence-transformers 3.0, torch 2.3)"

Fix: pin transformers<5.0.0 to avoid nn import bug"

Fix: use torch 2.4 CPU for compat with newer transformers"

Fix deps: let pip resolve versions, just pin what matters"

Dockerfile: install sentencepiece binary first"

Fix: pin sentence-transformers==2.7.0, pre-install sentencepiece wheel"

Fix: create model cache dir with proper permissions"

Fix build: add sentencepiece deps (pkg-config, protobuf)"

Pin tokenizers in requirements"

Pin tokenizers version with pre-built wheel, remove Rust dep"