Haiku SQLite RAG

A SQLite-based Retrieval-Augmented Generation (RAG) system built for efficient document storage, chunking, and hybrid search capabilities.

Features

Local SQLite: No need to run additional servers
Support for various embedding providers: You can use Ollama, VoyageAI, OpenAI or add your own
Vector Embeddings: Uses sqlite-vec for efficient similarity search
Hybrid Search: Full-text search (FTS5) combined with vector embeddings using Reciprocal Rank Fusion
Multi-format Support: Parse 40+ file formats including PDF, DOCX, HTML, Markdown, audio and more
Web Content: Direct URL ingestion with automatic content type detection

Installation

uv pip install haiku.rag

By default Ollama (with the mxbai-embed-large model) is used for the embeddings. For other providers use:

VoyageAI: uv pip install haiku.rag --extra voyageai

Configuration

If you want to use an alternative embeddings provider (Ollama being the default) you will need to set the provider details through environment variables:

By default:

EMBEDDING_PROVIDER="ollama"
EMBEDDING_MODEL="mxbai-embed-large" # or any other model
EMBEDDING_VECTOR_DIM=1024

For VoyageAI:

EMBEDDING_PROVIDER="voyageai"
EMBEDDING_MODEL="voyage-3.5" # or any other model
EMBEDDING_VECTOR_DIM=1024

Quick Start

from pathlib import Path
from haiku.rag.client import HaikuRAG

# Use as async context manager (recommended)
async with HaikuRAG("path/to/database.db") as client:
    # Create document from text
    doc = await client.create_document(
        content="Your document content here",
        uri="doc://example",
        metadata={"source": "manual", "topic": "example"}
    )

    # Create document from file (auto-parses content)
    doc = await client.create_document_from_source("path/to/document.pdf")

    # Create document from URL
    doc = await client.create_document_from_source("https://example.com/article.html")

    # Retrieve documents
    doc = await client.get_document_by_id(1)
    doc = await client.get_document_by_uri("file:///path/to/document.pdf")

    # List all documents with pagination
    docs = await client.list_documents(limit=10, offset=0)

    # Update document content
    doc.content = "Updated content"
    await client.update_document(doc)

    # Delete document
    await client.delete_document(doc.id)

    # Search documents using hybrid search (vector + full-text)
    results = await client.search("machine learning algorithms", limit=5)
    for chunk, score in results:
        print(f"Score: {score:.3f}")
        print(f"Content: {chunk.content}")
        print(f"Document ID: {chunk.document_id}")
        print("---")


# Or use without the context manager.
client = HaikuRAG(":memory:")
try:
    # ... operations ...
finally:
    client.close()

Search Functionality

haiku.rag provides hybrid search combining vector similarity and full-text search:

Vector Search: Uses embeddings to find semantically similar content
Full-text Search: Uses SQLite FTS5 for exact keyword matching
Hybrid Ranking: Combines both using Reciprocal Rank Fusion (RRF)
Chunked Results: Returns relevant document chunks with scores

async with HaikuRAG("database.db") as client:

    results = await client.search(
        query="machine learning",
        limit=5,  # Maximum results to return, defaults to 5
        k=60      # RRF parameter for reciprocal rank fusion, defaults to 60
    )

    # Process results
    for chunk, relevance_score in results:
        print(f"Relevance: {relevance_score:.3f}")
        print(f"Content: {chunk.content}")
        print(f"From document: {chunk.document_id}")

Supported File Formats

haiku.rag supports 40+ file formats through MarkItDown:

Documents: PDF, DOCX, PPTX, XLSX
Web: HTML, XML
Text: TXT, MD, CSV, JSON, YAML
Code: PY, JS, TS, C, CPP, JAVA, GO, RS, and more
Media: MP3, WAV (transcription)

Contributing

Fork the repository
Create a feature branch
Add tests for new functionality
Ensure all tests pass: pytest
Run type checking & linting with pyright & ruff check
Submit a pull request

4.2 KiB Raw Blame History