Fix empty source_type from LLM crashing Pydantic Source validation

Fix vectorstore: sentence-transformers 2.7.0, BM25 fallback

Fix ChromaDB tenant error; add real CRPF tenders; update Tab 2

Fix ChromaDB tenant error on stale .chroma directory

Add Interpretability tab, fix mandatory verdict logic, demo reset, UI polish

Step 9: evaluator — per-criterion verdict with threshold safety rules

Step 8: vector store and bidder processor — ChromaDB indexing and retrieval

Step 7: OCR pipeline — 3-tier extraction with caching

Step 6: criteria extractor, audit, fallback, and Tab 2 wiring

Step 5: pdf_utils and chunker — PyMuPDF extraction and text chunking

Step 4: LLM client — DeepSeek wrapper with retry and LLMUnavailable

Step 1: project skeleton — all stubs, directory structure, and entry point