[myai] feat(rag): add FAISS semantic search for CVE similarity

Commit 8542833

Description

• Add sentence-transformers/all-MiniLM-L6-v2 embedder (384 dims, CPU)
• Add FAISS IndexFlatIP in-memory vector index (cosine similarity)
• Store embeddings as BYTEA in PostgreSQL (migration: ADD COLUMN embedding)
• Rebuild FAISS index at startup from stored embeddings
• embed_unindexed() background task indexes CVEs without embeddings
• POST /cve/embed — trigger manual embedding of unindexed CVEs
• GET /cve/similar/{cve_id}?k=5 — semantic similarity search via FAISS
• cve_rag: use FAISS for similar CVEs, fallback to CWE-based similarity
• NVD sync triggers embed_unindexed() after completion
• cve/stats now includes indexed count and faiss_index_size

Co-Authored-By: Claude Sonnet 4.6 noreply@anthropic.com

Files Changed

Diff Summary