Add Feedback

Overview

Add feedback ratings to requests to track user satisfaction and quality of LLM responses. Feedback can be used to identify problematic requests, improve prompts, and train evaluation models.

Endpoint

string

required

POST

string

required

/v1/request//feedback

Authentication

Requires API key authentication via the Authorization header:

Authorization: Bearer YOUR_API_KEY

Path Parameters

string

required

The unique identifier of the request to add feedback to

Request Body

boolean

required

The feedback rating for the request

true - Positive feedback (thumbs up)
false - Negative feedback (thumbs down)

Response

Returns a success or error response.

null

Always null on success

string

Error message if the request failed, otherwise null

Examples

Positive Feedback

Add a positive (thumbs up) rating:

curl -X POST "https://api.helicone.ai/v1/request/req_123abc/feedback" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "rating": true
  }'

Negative Feedback

Add a negative (thumbs down) rating:

curl -X POST "https://api.helicone.ai/v1/request/req_123abc/feedback" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "rating": false
  }'

Using in Code

async function addFeedback(requestId: string, isPositive: boolean) {
  const response = await fetch(
    `https://api.helicone.ai/v1/request/${requestId}/feedback`,
    {
      method: 'POST',
      headers: {
        'Authorization': 'Bearer YOUR_API_KEY',
        'Content-Type': 'application/json'
      },
      body: JSON.stringify({
        rating: isPositive
      })
    }
  );
  
  const result = await response.json();
  
  if (result.error) {
    console.error('Failed to add feedback:', result.error);
  } else {
    console.log('Feedback added successfully');
  }
}

// Add positive feedback
await addFeedback('req_123abc', true);

// Add negative feedback
await addFeedback('req_456def', false);

import requests

def add_feedback(request_id: str, is_positive: bool):
    response = requests.post(
        f'https://api.helicone.ai/v1/request/{request_id}/feedback',
        headers={
            'Authorization': 'Bearer YOUR_API_KEY',
            'Content-Type': 'application/json'
        },
        json={'rating': is_positive}
    )
    
    result = response.json()
    
    if result.get('error'):
        print(f"Error: {result['error']}")
    else:
        print('Feedback added successfully')

# Add positive feedback
add_feedback('req_123abc', True)

# Add negative feedback
add_feedback('req_456def', False)

const addFeedback = async (requestId, isPositive) => {
  const response = await fetch(
    `https://api.helicone.ai/v1/request/${requestId}/feedback`,
    {
      method: 'POST',
      headers: {
        'Authorization': 'Bearer YOUR_API_KEY',
        'Content-Type': 'application/json'
      },
      body: JSON.stringify({
        rating: isPositive
      })
    }
  );
  
  const result = await response.json();
  return result;
};

// Usage
await addFeedback('req_123abc', true);  // Positive
await addFeedback('req_456def', false); // Negative

Response Examples

Success Response

{
  "data": null,
  "error": null
}

Error Response

{
  "data": null,
  "error": "Request not found"
}

Integration Examples

React Component

Implement feedback buttons in your UI:

import { useState } from 'react';

function FeedbackButtons({ requestId }: { requestId: string }) {
  const [feedback, setFeedback] = useState<boolean | null>(null);
  const [loading, setLoading] = useState(false);

  const handleFeedback = async (rating: boolean) => {
    setLoading(true);
    
    try {
      const response = await fetch(
        `https://api.helicone.ai/v1/request/${requestId}/feedback`,
        {
          method: 'POST',
          headers: {
            'Authorization': `Bearer ${process.env.HELICONE_API_KEY}`,
            'Content-Type': 'application/json'
          },
          body: JSON.stringify({ rating })
        }
      );
      
      const result = await response.json();
      
      if (!result.error) {
        setFeedback(rating);
      }
    } catch (error) {
      console.error('Failed to submit feedback:', error);
    } finally {
      setLoading(false);
    }
  };

  return (
    <div className="feedback-buttons">
      <button
        onClick={() => handleFeedback(true)}
        disabled={loading || feedback !== null}
        className={feedback === true ? 'active' : ''}
      >
        👍 Helpful
      </button>
      <button
        onClick={() => handleFeedback(false)}
        disabled={loading || feedback !== null}
        className={feedback === false ? 'active' : ''}
      >
        👎 Not Helpful
      </button>
    </div>
  );
}

CLI Tool

Add feedback from the command line:

#!/bin/bash

# feedback.sh - Add feedback to a request

REQUEST_ID=$1
RATING=$2  # "positive" or "negative"

if [ "$RATING" = "positive" ]; then
  RATING_VALUE="true"
elif [ "$RATING" = "negative" ]; then
  RATING_VALUE="false"
else
  echo "Usage: $0 <request_id> <positive|negative>"
  exit 1
fi

curl -X POST "https://api.helicone.ai/v1/request/$REQUEST_ID/feedback" \
  -H "Authorization: Bearer $HELICONE_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{\"rating\": $RATING_VALUE}"

Use Cases

User Feedback Collection

Collect end-user feedback on AI responses:

// After showing AI response to user
function showResponseWithFeedback(response: string, requestId: string) {
  // Display response
  displayResponse(response);
  
  // Add feedback buttons
  addFeedbackButtons({
    onThumbsUp: () => addFeedback(requestId, true),
    onThumbsDown: () => addFeedback(requestId, false)
  });
}

A/B Testing

Compare different prompts or models:

async function runABTest() {
  // Test variant A
  const responseA = await callLLM(promptA);
  const feedbackA = await collectUserFeedback();
  await addFeedback(responseA.requestId, feedbackA);
  
  // Test variant B
  const responseB = await callLLM(promptB);
  const feedbackB = await collectUserFeedback();
  await addFeedback(responseB.requestId, feedbackB);
  
  // Analyze feedback rates for each variant
  const resultsA = await queryRequests({ promptId: promptA.id });
  const resultsB = await queryRequests({ promptId: promptB.id });
  
  const positiveRateA = calculatePositiveRate(resultsA);
  const positiveRateB = calculatePositiveRate(resultsB);
}

Quality Monitoring

Automatically flag low-quality responses:

async function monitorQuality() {
  // Query recent requests with negative feedback
  const problematicRequests = await queryRequests({
    filter: {
      feedback: {
        rating: { equals: false }
      }
    },
    sort: { created_at: 'desc' },
    limit: 100
  });
  
  // Alert team if negative feedback rate is high
  const negativeRate = problematicRequests.length / totalRequests;
  if (negativeRate > 0.2) {
    await sendAlert('High negative feedback rate detected!');
  }
}

Training Data Generation

Use feedback to create training datasets:

async function exportTrainingData() {
  // Get requests with positive feedback
  const positiveExamples = await queryRequests({
    filter: {
      feedback: { rating: { equals: true } }
    },
    includeInputs: true,
    limit: 1000
  });
  
  // Get requests with negative feedback
  const negativeExamples = await queryRequests({
    filter: {
      feedback: { rating: { equals: false } }
    },
    includeInputs: true,
    limit: 1000
  });
  
  // Export for fine-tuning or evaluation
  await exportToFile(positiveExamples, 'positive_examples.jsonl');
  await exportToFile(negativeExamples, 'negative_examples.jsonl');
}

Query Requests by Feedback

You can filter requests by feedback using the query endpoint:

# Get all requests with positive feedback
curl -X POST "https://api.helicone.ai/v1/request/query" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "filter": {
      "feedback": {
        "rating": {
          "equals": true
        }
      }
    },
    "limit": 100
  }'

# Get all requests with negative feedback
curl -X POST "https://api.helicone.ai/v1/request/query" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "filter": {
      "feedback": {
        "rating": {
          "equals": false
        }
      }
    },
    "limit": 100
  }'

Notes

Feedback can be updated by sending another request with a different rating
Only one feedback rating per request is stored (most recent wins)
Feedback is immediately reflected in query results and analytics
Use feedback in combination with custom properties for more detailed tracking
Feedback data is available in exports and webhooks
Consider implementing a feedback collection UI that’s easy for users to access

Best Practices

Make feedback easy: Place feedback buttons prominently in your UI
Explain context: Tell users what the feedback is for and how it helps
Don’t overdo it: Only ask for feedback on important interactions
Act on feedback: Regularly review negative feedback and improve prompts
Track trends: Monitor feedback rates over time to measure improvements
Combine signals: Use feedback alongside evaluation scores for better insights

Gateway

Requests

Sessions

Prompts

Evaluations

Overview

Endpoint

Authentication

Path Parameters

Request Body

Response

Examples

Positive Feedback

Negative Feedback

Using in Code

Response Examples

Success Response

Error Response

Integration Examples

React Component

CLI Tool

Use Cases

User Feedback Collection

A/B Testing

Quality Monitoring

Training Data Generation

Query Requests by Feedback

Notes

Best Practices

​Overview

​Endpoint

​Authentication

​Path Parameters

​Request Body

​Response

​Examples

​Positive Feedback

​Negative Feedback

​Using in Code

​Response Examples

​Success Response

​Error Response

​Integration Examples

​React Component

​CLI Tool

​Use Cases

​User Feedback Collection

​A/B Testing

​Quality Monitoring

​Training Data Generation

​Query Requests by Feedback

​Notes

​Best Practices

Overview

Endpoint

Authentication

Path Parameters

Request Body

Response

Examples

Positive Feedback

Negative Feedback

Using in Code

Response Examples

Success Response

Error Response

Integration Examples

React Component

CLI Tool

Use Cases

User Feedback Collection

A/B Testing

Quality Monitoring

Training Data Generation

Query Requests by Feedback

Notes

Best Practices