Own-Dark14 highlights a limitation in GitHub Copilot’s session memory, where the agent fails to retain context consistently during code execution.

Summary

Own-Dark14 discusses a user experience issue encountered with GitHub Copilot’s agent during code execution within a single session. Initially, when prompted, the agent correctly executed code using docker-compose. Upon updating the code and requesting another run, the agent unexpectedly switched to running it with plain python, seemingly forgetting the previously used method.

Observations

  • Inconsistency Noted: The agent failed to remember its earlier context (execution via docker-compose) and changed behavior mid-session.
  • Expected Behavior: Ideally, the agent should recall the context of execution style previously used in the session for consistency.
  • Comparison: The author references Claude, noting its stronger session memory. Claude informs users not to re-upload files as it remembers within a session, benefiting user consistency.
  • User Recommendation: The post advocates for improved contextual consistency in GitHub Copilot, suggesting that session memory enhancement would make for a more user-friendly experience.

Conclusion

Retaining execution context, such as the method used to run code, is valuable for maintaining a seamless and predictable workflow in tools like GitHub Copilot.

This post appeared first on “Reddit Github Copilot”. Read the entire article here