Share your feedback

#1
by karenwky - opened
Agents-MCP-Hackathon org

Hello, everyone! 👋

First of all, thank you so much for choosing this project as Agent Track Honorable Mention! 🥹🙏❤️

Thank you to everyone who tried the chatbot and liked the space as well! ❤️

Any feedback is welcome! 🤗

Karen

Agents-MCP-Hackathon org
edited Jun 22, 2025

Some issues I've noticed from the trace:

  • "ValueError: Calculated available context size -383 was not non-negative" when usage increases, I restart the space for quick fix temporarily
  • "RecursionError: maximum recursion depth exceeded while calling a Python object" sometimes seeing this error which I have no idea about
  • stricter system prompt is needed for limiting the chatbot only answer questions related to the fictional company
  • response length restriction for fitting chatbot size can be added in system prompt for better UX
karenwky pinned discussion

Nice work on the honorable mention. A practical improvement I would make first is adding a visible “run health” / fallback behavior around the errors you listed.

For the context-size error, I would cap conversation history aggressively and summarize older turns before they hit the model. For recursion errors, I would add a max tool/agent step counter and log the last intent before stopping.

For the fictional-company constraint, I would test with a small eval set:

  • valid company question
  • unrelated general question
  • prompt asking it to ignore the company constraint
  • long multi-turn conversation that drifts off topic

That gives you a repeatable way to see whether prompt changes actually improve the bot instead of just feeling stricter.

Sign up or log in to comment