first personal push

Production-ready: Fix PDF extraction with pdfplumber, optimize objective length for <30s latency, and fix name extraction accuracy

Final production push: 0.5B speed optimization, structured JSON, and ALL CAPS name formatting

Robust JSON parsing: Increase tokens, improve cleaning, and forbid markdown/arrays

Optimize for speed: Switch to 0.5B model and reduce threads for <30s latency

Fix runaway generation by enforcing max_tokens and using create_chat_completion

Fix token truncation by increasing n_ctx to 8192

Trigger HF Space rebuild

Add comprehensive logging with timing for every CV processing step

Enhance startup sequence: model downloads on startup, not on first request

Add API documentation with Flasgger/Swagger

Add missing huggingface_hub dependency

Add resume parser microservice with local LLM

initial commit