A serverless, AI-powered deep research agent built with Cloudflare Workers and Google Gemini 2.5
Hello. I look forward to seeing the progress of your project. I deployed this source code using the paid version of Cloudflare Workers and noticed two things. - Under the “Information Source” heading in the generated report, the number of URLs visited by this application using DuckDuckGo and Cloudflare Browser Rendering is not displayed. The Workers log recognizes the URLs it should see, but an error message indicates that the connection is closed. It is unfortunate that the report does not show evidence of gathering information from the internet, even though I can see from the Workers' log that a search query was entered and the URL was referenced. - The re-run button is not displayed. > If you want to re-run a previous research, click the "🔄" button next to the research entry. This will create a new research based on the same query and parameter That's all. I am very excited about the future of your project. If you can implement a mechanism that instructs Cloudflare's edge servers to execute such multi-layered LLMs, it would mean that the era of individuals owning AI agents has arrived.
This issue appears to be discussing a feature request or bug report related to the repository. Based on the content, it seems to be resolved. The issue was opened by kotaromiyasaka and has received 6 comments.