docs(cookbook): add zero-cloud local hybrid RAG recipe with SQLite FTS5 and Sentence Transformers

[Cagrik34]

Summary

This recipe introduces a self-contained, zero-cloud reference implementation of Hybrid Retrieval-Augmented Generation (RAG) integrating SQLite FTS5 (BM25) and Dense Vector Cosine Similarity with open-source Sentence Transformers (sentence-transformers/all-MiniLM-L6-v2).

Key Technical Details

• Zero External Infrastructure: Operates entirely within an embedded SQLite database (:memory: or local .db), eliminating external vector database overhead.
• Dual-Engine Retrieval: Fuses 384-dimensional dense semantic vectors with native SQLite FTS5 full-text token matching.
• Reciprocal Rank Fusion (RRF, $k=60$): Resolves score calibration challenges and guarantees recall of exact numeric tokens, part codes, and domain acronyms.
• Strict Grounding: Demonstrates citation formatting ([1], [2]) ready for any open-source SLM/LLM pipeline.

Location

• notebooks/en/zero_cloud_hybrid_rag_sqlite_fts5.ipynb
• Added entry to notebooks/en/_toctree.yml under LLM Recipes.

Reviewers

cc @merveenoyan @stevhliu

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[Cagrik34]

Hi @huggingface team, friendly check-in. This is a zero-cloud Hybrid RAG cookbook recipe: sentence-transformers embeddings + SQLite FTS5 (BM25) + RRF (k=60), no third-party vector databases or paid APIs. Self-contained notebook with reproducible outputs. Ready for review whenever convenient. Thank you!