feat: enhance LLM handsoff tool execution with system prompt and run hooks
This commit is contained in:
@@ -79,8 +79,10 @@ class FunctionToolExecutor(BaseFunctionToolExecutor[AstrAgentContext]):
|
||||
event=event,
|
||||
chat_provider_id=prov_id,
|
||||
prompt=input_,
|
||||
system_prompt=tool.agent.instructions,
|
||||
tools=toolset,
|
||||
max_steps=30,
|
||||
run_hooks=tool.agent.run_hooks,
|
||||
)
|
||||
yield mcp.types.CallToolResult(
|
||||
content=[mcp.types.TextContent(type="text", text=llm_resp.completion_text)]
|
||||
|
||||
Reference in New Issue
Block a user