karthink
karthink
Do you have this issue when you use non-streaming responses? `(setq gptel-stream nil)` If yes it will be much easier to figure out the reason.
This doesn't look like a gptel issue. What happens if you try starting the servers from `mcp-hub` instead? (Press `s` to start a server.)
The data appears to be encoded correctly, but there is some error in the encoding of the buffer text that follows. I'm not sure what's going on.
Thanks. I tried this out on Windows (native build of Emacs) and Linux, and files with spaces are not an issue when added to the context for Ollama. It's a...
This is a neat idea, but I think gptel can do better. For most models we know the context window size, so it should be possible to cut it off...
The short answer is that gptel's default result callback, which controls how `gptel-send` handles and displays responses and tool calls in buffers, does not offer this level of fine-grained control....
> For now, I'm going for the second option, specifically I'm changing gptel--display-tool-calls, my plan is to have "confirm" become "run" and "run without request" and add a "insert only"...
> As I understand for tool calling, the LLM will return tool call json fields in > their response, the usual sequence is confirm a tool call, the tool definition...
> Now about the other idea, however, in the normal tool call UI, I'm struggle to make a new shortcut option that invoke the tool call and get the result,...
Which backend and model are you using here?