pi-local-rag

Local BM25 RAG pipeline for the Pi coding agent. Index your local files and search them with keyword matching — zero cloud dependency, works fully offline.

Features

• Hybrid BM25 + vector search — TF-IDF scoring with exact phrase and filename boosts, combined with local ONNX embeddings
• Smart chunking — splits files into ~50-line blocks at natural blank-line boundaries
• Incremental indexing — skips unchanged files (SHA-256 hash check)
• Zero cloud dependency — uses only Node.js built-ins + local Transformers.js model
• 3 AI tools — rag_index, rag_query, rag_status for the agent to use directly

Install

pi install npm:pi-local-rag

Or via git:

pi install git:github.com/vahidkowsari/pi-local-rag

Commands

Command	Description
/rag index <path>	Index a file or directory
/rag search <query>	Search indexed content
/rag status	Show index stats (files, chunks, tokens)
/rag rebuild	Re-index changed files, prune deleted
/rag clear	Wipe the entire index
/rag on	Enable auto-injection
/rag off	Disable auto-injection

AI Tools

The extension registers three tools the agent can call directly:

• rag_index — Index a path into the pipeline
• rag_query — Hybrid BM25+vector search, returns file paths + line numbers + previews
• rag_status — Show index stats and RAG config

How It Works

1. Index — files are chunked (~50 lines each), embedded with Xenova/all-MiniLM-L6-v2 (384-dim), and stored locally at ~/.pi/rag/
2. Search — hybrid scoring: alpha × BM25 + (1-alpha) × cosine_similarity (default alpha=0.4)
3. Auto-inject — before every agent turn, the prompt is searched and relevant chunks are prepended to the system prompt

Storage

Index data is stored at ~/.pi/rag/. If you previously used an older version of this plugin (~/.pi/lens/), the directory is automatically migrated on first run.

Configuration

Auto-injection is on by default. Tune via /rag status:

Setting	Default	Description
ragEnabled	true	Auto-inject context before each turn
ragTopK	5	Max chunks to inject
ragScoreThreshold	0.1	Min hybrid score to include
ragAlpha	0.4	BM25/vector blend (0=pure vector, 1=pure BM25)