LeetTools is an AI search assistant designed for customizable search workflows, enabling users to generate reports and analyze information from both web and local knowledge bases. It targets developers and researchers who need flexible data ingestion, indexing, and retrieval pipelines, offering a command-line interface with configurable LLM and search providers.

How It Works

LeetTools employs an automated document pipeline for data ingestion, indexing, and storage, abstracting infrastructure concerns. It supports various backends like DuckDB for local operations and can integrate with MongoDB, Milvus, and Neo4j for different data management needs. Users can configure distinct LLM providers and models for specific workflow components, allowing for highly tailored search and generation tasks.

Quick Start & Requirements

• Installation: pip install leettools (Conda/Venv recommended).
• Prerequisites: Python 3.11+, OpenAI-compatible LLM endpoint (e.g., local Ollama, DeepSeek API), API keys for services.
• Configuration: Set environment variables like EDS_LLM_API_KEY, EDS_DEFAULT_LLM_BASE_URL, EDS_WEB_RETRIEVER (e.g., google, firecrawl).
• Docker: docker/start.sh to run the web UI (under development).
• Source: Clone repo, install requirements, and run from source.
• Docs: https://github.com/leettools-dev/leettools

Highlighted Details

• Supports multiple LLM providers (OpenAI-compatible, Ollama, DeepSeek) and embedding services.
• Configurable web retrievers including Google Search and FireCrawl.
• Offers pre-defined workflows: answer, digest, search, news, extract, opinions.
• Enables building local knowledge bases from web PDFs and querying them.

Maintenance & Community

• Actively uses libraries like DuckDB, Ollama, FastAPI.
• Community support via GitHub discussions.
• Contribution guidelines available.

Licensing & Compatibility

• Licensed under Apache License, Version 2.0.
• Permissive license suitable for commercial use and integration with closed-source projects.

Limitations & Caveats

The web UI is currently under development and not yet open-sourced. Production use of the default shared proxy search service is not recommended; users should configure their own search API keys.