You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Many of the responses from the API are very large and eat valuable context window. We should think through strategies on how to give the model options for smaller results. That might be *_small tools that return trimmed results or possibly the option for the model to pass in a nested set of keys to select.
The text was updated successfully, but these errors were encountered:
This blocks a use case of using the GH mcp with GH copilot agent mode, which I'd expect to integrate well.
Example of what happens now:
I give GHC Agent mode in vscode the GH mcp server, ask it for summarizing issues with a label, it queries github, receives 200k tokens as a response, everything else falls out of the context window and it forgets what is it doing.
Many of the responses from the API are very large and eat valuable context window. We should think through strategies on how to give the model options for smaller results. That might be
*_small
tools that return trimmed results or possibly the option for the model to pass in a nested set of keys to select.The text was updated successfully, but these errors were encountered: