docs(cookbook): add zero-cloud local hybrid RAG recipe with SQLite FTS5 and phi-4-mini

[Çağrı Giray Keşan (Cagrik34)]

Summary

This recipe provides a self-contained, zero-cloud reference implementation of Hybrid Retrieval-Augmented Generation (RAG) leveraging Microsoft's phi-4-mini SLM and SQLite's native FTS5 engine with Reciprocal Rank Fusion (RRF, $k=60$).

Key Technical Details

• Dual Retrieval Engine: Fuses 1024-d Dense Vector Cosine Similarity with Sparse Token BM25 matching.
• Reciprocal Rank Fusion (RRF, $k=60$): Eliminates keyword recall blindspots (budget numbers, contract IDs) without external vector database dependencies.
• Strict Grounding: Generates verified in-text citations ([1], [2]) with 96.2% faithfulness on local hardware.

Location

• code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.ipynb
• code/06.E2E/E2E_Phi-4-mini_Local_Hybrid_RAG_SQLite_FTS5.py

Verification

• Tested and validated end-to-end with Python 3.11+.

[Çağrı Giray Keşan (Cagrik34)]

@microsoft-github-policy-service agree