LLM search engine faster than perplexity!
Spy Search is an agentic search framework designed to outperform current web search agents with a focus on faster, more efficient daily usage.
While commercial solutions like Manus charge $200 per month, Spy Search leverages open-source models to provide a cost-effective alternative without sacrificing performance. Currently our searching-speed is quite slow yet we can generate a long length consistent report (around 2000 words) with latest current information! This problem will be tackle after the release of v1.0.
News: 2025-06-10 Spy-searcher has just released v0.3 !
First you have to clone the repo
git clone https://github.com/JasonHonKL/spy-search.git
cd spy-searchTo set up just run
python setup.pyAdd your API key in the .env file if you want to use API. Currently we support openAI, Claude, Gork & DeepSeek.
Configure the config.json file for your local setup. You can copy the example configuration:
cp config.example.json config.jsonThen edit config.json to match your environment.
If you use Ollama, here's an example of config.json:
{
"provider": "ollama",
"model": "qwen3:8b",
"agents": [
"reporter"
],
"db": "./local_files/test",
"base_url": "http://host.docker.internal:11434",
"language": "en"
}Build and run the application using Docker. This method is recommended for most users as it handles all dependencies automatically:
# Build and start the container
docker-compose up --build
# Or run in background
docker-compose up -d --buildPrerequisites:
# Ubuntu/Debian
sudo apt-get install docker-compose-plugin
# Or check if installed
docker-compose --versionconfig.json available in Ollama (e.g., qwen3:8b)Changes to config.json are automatically synced and take effect immediately - no restart needed.
Now you can access:
Watch the demo video on YouTube:
https://github.com/user-attachments/assets/3e6ef332-d055-421c-bf0a-5f866ba52b11
We welcome contributions from the community! Here’s how you can contribute:
main branch.Once submitted, maintainers will review your PR and provide feedback or merge it if it meets the project standards.
Maintainers will acknowledge your issue, label it appropriately, and work on resolving it or discuss it with you.
Thank you for helping improve this project! Your contributions make a difference.
No open issues yet, or sync has not completed.