Step-by-step guide to building a local AI web research agent using Ollama, the Qwen3.5:4b model, and Python. The agent accepts a research prompt, fetches the top 5 web results via Ollama's web search API, extracts readable text with BeautifulSoup, and sends everything to a locally running Qwen model for summarization. The result is saved as a timestamped Markdown file. The setup runs entirely on-device with no API costs and preserves privacy. Full working code is provided along with setup instructions for macOS, Windows, and Linux.
Table of contents
BackgroundMotivation and ArchitectureStep 1: Install Ollama and Get an API KeyStep 2: Pull the Qwen ModelStep 3: Install Python DependenciesStep 4: Write the Agent CodeStep 5: Run the AgentConclusion2K Impressions