A fully local web research and report writing assistant that protects your privacy. The AI models run locally using Ollama or LM Studio, and web searches are performed locally with SearXNG. By default, no API keys are required.

Note: Steps 1 and 2 are optional. You can instead run models and web searches in the cloud. To do so, add the required API keys to your .env file. Supported model providers include OpenRouter, Together AI, DeepInfra, and OpenAI. For web searches, Deep Researcher 2 supports Brave, DuckDuckGo, Perplexity, Serper and Tavily.

Getting started

1. Install Ollama and download a model.

   ollama pull qwen2.5:72b

2. Start SearXNG.

   git clone git@github.com:lars20070/searxng-docker.git
   cd searxng-docker
   sed -i '' "s|ultrasecretkey|$(openssl rand -hex 32)|g" searxng/settings.yml
   docker compose up -d
   # Check localhost:8080

3. Set up environment variables.
   Create a .env file and fill in the placeholders including the TOPIC.

   cp .env.example .env

4. Start up the workflow.

   uv run research

stateDiagram-v2
    WebSearch: Web Search
    SummarizeSearchResults: Summarize Search Results
    ReflectOnSummary: Reflect on Summary
    FinalizeSummary: Finalize Summary
    [*] --> WebSearch
    WebSearch --> SummarizeSearchResults: web search result
    SummarizeSearchResults --> ReflectOnSummary: web search summary
    ReflectOnSummary --> WebSearch: reflection
    ReflectOnSummary --> FinalizeSummary
    FinalizeSummary --> [*]

Loading

Deep Researcher 2 workflow