"""Feedback tools for rating tool executions and improving RL learning."""
from typing import Any, Optional
from fccs_agent.config import config
from fccs_agent.services.feedback_service import get_feedback_service
async def submit_feedback(
execution_id: int,
rating: int,
feedback: Optional[str] = None
) -> dict[str, Any]:
"""Submit user feedback for a tool execution to improve RL learning.
This tool allows you to rate a tool execution (1-5 stars) and provide
optional feedback. This helps the RL system learn which tools work best
for your use cases and improves future recommendations.
Args:
execution_id: The ID of the tool execution to rate (from tool result)
rating: Rating from 1-5 stars (5 = excellent, 1 = poor)
feedback: Optional text feedback about the execution
Returns:
dict: Status of feedback submission
"""
if rating < 1 or rating > 5:
return {
"status": "error",
"error": f"Rating must be between 1 and 5, got {rating}"
}
feedback_service = get_feedback_service()
if not feedback_service:
return {
"status": "error",
"error": "Feedback service not available"
}
try:
feedback_service.add_user_feedback(
execution_id=execution_id,
rating=rating,
feedback=feedback
)
return {
"status": "success",
"message": f"Feedback submitted: {rating} stars",
"execution_id": execution_id,
"rating": rating,
"feedback": feedback
}
except Exception as e:
return {
"status": "error",
"error": f"Failed to submit feedback: {str(e)}"
}
async def get_recent_executions(
tool_name: Optional[str] = None,
limit: int = 10
) -> dict[str, Any]:
"""Get recent tool executions that can be rated.
Use this to find execution IDs for tools you want to provide feedback on.
Args:
tool_name: Optional filter by specific tool name
limit: Maximum number of executions to return (default: 10)
Returns:
dict: List of recent executions with their IDs and status
"""
feedback_service = get_feedback_service()
if not feedback_service:
return {
"status": "error",
"error": "Feedback service not available"
}
try:
executions = feedback_service.get_recent_executions(
tool_name=tool_name,
limit=limit
)
return {
"status": "success",
"executions": executions,
"count": len(executions)
}
except Exception as e:
return {
"status": "error",
"error": f"Failed to get executions: {str(e)}"
}
# Tool definitions for MCP server
TOOL_DEFINITIONS = [
{
"name": "submit_feedback",
"description": "Submit user feedback (rating 1-5 stars) for a tool execution to improve RL learning / Enviar feedback do usuario para melhorar aprendizado RL",
"inputSchema": {
"type": "object",
"properties": {
"execution_id": {
"type": "integer",
"description": "The ID of the tool execution to rate (found in tool result or from get_recent_executions)",
},
"rating": {
"type": "integer",
"description": "Rating from 1-5 stars (5 = excellent, 4 = good, 3 = average, 2 = poor, 1 = bad)",
"minimum": 1,
"maximum": 5,
},
"feedback": {
"type": "string",
"description": "Optional text feedback about the execution",
},
},
"required": ["execution_id", "rating"],
},
},
{
"name": "get_recent_executions",
"description": "Get recent tool executions that can be rated / Obter execucoes recentes que podem ser avaliadas",
"inputSchema": {
"type": "object",
"properties": {
"tool_name": {
"type": "string",
"description": "Optional filter by specific tool name",
},
"limit": {
"type": "integer",
"description": "Maximum number of executions to return (default: 10)",
"default": 10,
},
},
},
},
]
