Spy Search is an agentic search framework designed to provide a faster, more cost-effective alternative to commercial LLM search solutions. It targets users seeking efficient daily web search capabilities, leveraging open-source models to deliver comprehensive reports with current information.

How It Works

The framework utilizes an agentic approach, allowing for flexible configuration of LLM providers (OpenAI, Claude, Gork, DeepSeek, Ollama) and models. Users can define specific agents, such as a "reporter," and configure local databases or API endpoints. The system aims to generate long, consistent reports by integrating with various LLM backends.

Quick Start & Requirements

• Installation: Clone the repository (git clone https://github.com/JasonHonKL/spy-search.git), navigate into the directory (cd spy-search), and run python setup.py.
• Configuration: Add API keys to a .env file and edit config.json for local setup. An example configuration for Ollama is provided.
• Docker: Recommended for ease of use. Build and run with docker-compose up --build or docker-compose up -d --build.
• Prerequisites: Docker and Docker Compose are required for the Docker method. Ollama and a compatible model are needed if using the Ollama provider.
• Access: Backend API at http://localhost:8000, Frontend at http://localhost:8080.
• Demo: A demo video is available on YouTube.

Highlighted Details

• Aims to outperform current web search agents in speed and efficiency.
• Supports multiple LLM providers including OpenAI, Claude, Gork, DeepSeek, and Ollama.
• Capable of generating long, consistent reports (around 2000 words) with current information.
• Dockerized deployment simplifies dependency management.

Maintenance & Community

The project welcomes community contributions via pull requests and issues. Links to a demo video and community Discord are provided.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README. Users should verify licensing for commercial or closed-source integration.

Limitations & Caveats

The project acknowledges that current searching speed is slow, with plans to address this in v1.0. The README does not specify the exact open-source models used or provide performance benchmarks against competitors like Perplexity.