wrap_model_call(
self,
request: ModelRequest,
handler: Callable[[ModelRequest], ModelResponse]
) | Name | Type | Description |
|---|---|---|
request* | ModelRequest | |
handler* | Callable[[ModelRequest], ModelResponse] |
Inject a compact-tool usage nudge into the system prompt.
This only updates prompt text so the model can decide whether to call
compact_conversation earlier in long sessions. It does not execute the
tool automatically.
The model request to process.
The handler to call with the modified request.