Automated AI Web Researcher
A local AI research assistant uses Ollama to investigate web sources, save retrieved content with links, and answer questions about its findings.
Automated AI Web Researcher investigates a topic using a locally running LLM. It's for people who want to delegate web searches and keep the collected material for review, rather than gather each page themselves.
- Research planning: It breaks a question into prioritized areas and develops further lines of inquiry from what it finds.
- Web investigation: The assistant searches the web, selects relevant pages, and extracts content using searches through DuckDuckGo.
- Saved research: It writes retrieved content, source URLs, investigated areas, and the final summary to a local text file.
- Summaries and follow-up questions: You can pause research for an assessment of the evidence, stop it for a summary, then discuss the findings with the model.
The Python application runs in a console on macOS and Linux; Windows support uses a separate branch. Model inference runs locally through Ollama, while web searches and page retrieval require an internet connection and contact external websites. It isn't an offline research tool.
The project is open source under the MIT license and is a prototype. It recommends long-context Phi-3 models, including phi3:3.8b-mini-128k-instruct and phi3:14b-medium-128k-instruct, so the model can review the material collected during a research session.
Logotyper
