To report incorrect results or bugs in DeepResearch, users can submit detailed feedback through three primary channels: an in-app feedback form, a dedicated support email, or a public issue tracker on GitHub. The most direct method is the in-app feedback tool, accessible via a “Report Issue” button in the interface. This form allows users to describe the problem, attach screenshots or logs, and automatically includes metadata like the query timestamp and system version. For technical users, the GitHub issue tracker provides a transparent way to report bugs, propose fixes, or track resolutions. Alternatively, emailing support@deepresearch.ai with a subject line like “Bug Report: Incorrect API Response” ensures the team prioritizes the request.
When submitting a report, include specific details to help the team reproduce and diagnose the issue. For example, if an API endpoint returns a 500 error when processing a valid JSON payload, provide the exact input, headers, and response body. If a search query yields irrelevant results, share the exact query, expected output, and a screenshot of the incorrect result. For reproducibility, include environment details (e.g., browser version, OS, SDK version) and error logs. If the bug is intermittent, note the frequency and any patterns (e.g., “Fails 30% of the time with batch requests over 10MB”). Structured data like code snippets or HAR files can accelerate troubleshooting.
After submission, the DeepResearch team triages reports based on severity and impact. Critical bugs affecting core functionality (e.g., authentication failures) are typically addressed within 24-48 hours, while minor UI issues may take longer. Users receive a confirmation email with a tracking ID and can monitor status updates via GitHub or email. For open-source components, contributors can submit pull requests with test cases demonstrating the fix. The team also maintains a public changelog documenting resolved issues. For example, a recent patch (v2.1.3) fixed a query parser bug reported via GitHub, where nested AND/OR operators caused incorrect filtering. Transparency in the resolution process ensures developers understand how their feedback improves the platform.
Zilliz Cloud is a managed vector database built on Milvus perfect for building GenAI applications.
Try FreeLike the article? Spread the word