rapid-rag 0.2.0

Fast local RAG - search your documents with AI, no cloud needed

rapid-rag

Fast local RAG - search your documents with AI, no cloud needed.

Installation

pip install rapid-rag

For PDF support:

pip install rapid-rag[pdf]

Quick Start

from rapid_rag import RapidRAG

# Create a RAG instance
rag = RapidRAG("my_documents")

# Add documents
rag.add("doc1", "The quick brown fox jumps over the lazy dog.")
rag.add_file("report.pdf")
rag.add_directory("./docs/")

# Semantic search
results = rag.search("fox jumping")
for r in results:
    print(f"{r['score']:.3f}: {r['content'][:100]}")

# RAG query with LLM (requires Ollama)
answer = rag.query("What does the fox do?", model="qwen2.5:7b")
print(answer["answer"])

CLI Usage

# Initialize a collection
rapid-rag init my_docs

# Add documents
rapid-rag add ./documents/ -c my_docs -r

# Search
rapid-rag search "query here" -c my_docs

# RAG query (requires Ollama)
rapid-rag query "What is X?" -c my_docs -m qwen2.5:7b

# Info
rapid-rag info -c my_docs

TIBET Provenance

Track every operation with cryptographic provenance:

from rapid_rag import RapidRAG, TIBETProvider

# Enable TIBET tracking
tibet = TIBETProvider(actor="my_app")
rag = RapidRAG("docs", tibet=tibet)

# All operations now create provenance tokens
rag.add_file("report.pdf")
results = rag.search("query")
answer = rag.query("Question?")

# Get provenance chain
tokens = tibet.get_tokens()
for t in tokens:
    print(f"{t.token_type}: {t.erachter}")
    print(f"  ERIN: {t.erin}")  # What happened
    print(f"  ERACHTER: {t.erachter}")  # Why

TIBET uses Dutch provenance semantics:

• ERIN: What's IN the action (content)
• ERAAN: What's attached (references)
• EROMHEEN: Context around it
• ERACHTER: Intent behind it

Features

• Local-first: Everything runs on your machine
• Fast: ChromaDB + sentence-transformers
• Simple API: Add, search, query in 3 lines
• File support: .txt, .md, .pdf
• Chunking: Automatic with overlap
• LLM integration: Works with Ollama
• TIBET: Cryptographic provenance for all operations

Requirements

• Python 3.10+
• For LLM queries: Ollama running locally

License

MIT - Humotica