Conversation Feedback API

Collect and manage user feedback on AI responses to improve your models and user experience.

Overview

The Conversation Feedback API allows you to collect user ratings (upvotes/downvotes) on AI responses in conversations. This feedback can be used for analytics, model training, and improving the quality of responses over time.

Rate a Conversation Message

Submit a rating for a specific assistant message in a conversation.

POST /messages/{conversation}/rate

Request Parameters

Parameter	Type	Required	Description
user	string	Yes	The user identifier that the conversation belongs to
message_id	string	Yes	The ID of the message to rate
rating	integer	Yes	The rating value: 1 (upvote) or -1 (downvote)

Path Parameters

Parameter	Description
conversation	The nanoid of the conversation containing the message

Response

On success, the server returns a 200 OK response with the following JSON:

{
  "message": "Message rated successfully.",
  "rating": 1
}

Error Codes

Code	Description
404	Conversation not found, message not found, or user mismatch
422	Validation error or attempt to rate a non-assistant message

Example Request

curl -X POST "https://api.vectorforgeai.com/v1/messages/abc123def456/rate" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "user": "user123",
  "message_id": "resp_assi_xyz789",
  "rating": 1
}'

Use Cases

Collect user feedback on AI responses to identify areas for improvement
Track user satisfaction with different types of responses
Build datasets for fine-tuning models based on user-preferred responses
Create analytics dashboards to understand user sentiment over time

Best Practices

Implement both upvote and downvote options in your UI for balanced feedback
Consider coupling ratings with optional text feedback for more detailed insights
Use the rating data to identify patterns in responses that users find helpful or unhelpful
Periodically review highly-rated and lowly-rated responses to improve your system prompts