Conversation
|
@benbrandt I thought about this and I think we won't be able to abstract away the fact that agent execution doesn't happen on the current machine and that the changed code is not available right away. So I would prefer to keep it more explicit in the protocol |
|
|
||
| > How do things work today and what problems does this cause? Why would we change things? | ||
|
|
||
| Currently, ACP sessions assume the agent runs locally alongside the client, with direct filesystem access. The agent can read and write files in the user's working directory. |
There was a problem hiding this comment.
If I understand correctly: this is more about which code the agent should be working on?
Is the code in the cwd, or will it have to fetch it?
In this scenario, is it important that the agent knows what to do with this information? Or will there be something in between?
If the agent should, then I guess we need some capabilities. If not, then I wonder if this is somehow better handled by the proxy RFD where something gets a cwd ready for the agent session before calling session/new for the agent?
There was a problem hiding this comment.
Basically, I am trying to understand whose responsibility it is to set this thing up which will help me better grasp how this should fit in the protocol
There was a problem hiding this comment.
I.e. should this be here? Another method? somewhere else? etc.
2 participants
No description provided.