fix(agent): stabilize live e2e tool execution and loop isolation
This commit is contained in:
@@ -1,22 +0,0 @@
|
||||
intent:
|
||||
role: Intent Agent
|
||||
goal: Classify user intent and decide execution strategy
|
||||
backstory: >
|
||||
You are an expert intent classifier with deep understanding
|
||||
of user query patterns and dialogue acts. Your role is to
|
||||
analyze user input and determine the appropriate action.
|
||||
|
||||
execution:
|
||||
role: Execution Agent
|
||||
goal: Execute tasks with available tools
|
||||
backstory: >
|
||||
You are a skilled task executor with expertise in tool calling,
|
||||
API interactions, and result verification. You work systematically
|
||||
to complete user requests.
|
||||
|
||||
organization:
|
||||
role: Organization Agent
|
||||
goal: Organize output for user-friendly response
|
||||
backstory: >
|
||||
You specialize in presenting results in a clear, user-friendly manner.
|
||||
You ensure responses are well-structured and actionable.
|
||||
@@ -1,16 +0,0 @@
|
||||
intent:
|
||||
description: Identify user intent and required capabilities
|
||||
expected_output: >
|
||||
Structured intent classification with intent type, confidence score,
|
||||
and recommended action plan
|
||||
|
||||
execution:
|
||||
description: Execute intent with tools and model calls
|
||||
expected_output: >
|
||||
Verified execution results with tool outputs, status, and any errors
|
||||
|
||||
organization:
|
||||
description: Format final response and references
|
||||
expected_output: >
|
||||
User-friendly response with structured output, citations, and
|
||||
clear next steps if applicable
|
||||
@@ -1,6 +0,0 @@
|
||||
intent: []
|
||||
|
||||
execution:
|
||||
- back.create_calendar_event
|
||||
|
||||
organization: []
|
||||
Reference in New Issue
Block a user