Skip to main content

Batch Predictions

Make predictions on multiple samples in a single API call.

Input Format​

To make batch predictions, provide arrays with multiple values for each feature:

response = requests.post(
    f"{API_URL}/projects/{PROJECT_ID}/predict",
    headers={"X-API-Key": API_KEY, "Content-Type": "application/json"},
    json={
        "input": {
            "date": ["2024-01-15", "2024-01-16", "2024-01-17"],
            "category": ["A", "B", "A"],
            "value": [100, 200, 150]
        }
    }
)
important

All feature arrays must have the same length. This length determines the number of predictions returned.

Response Format​

The response contains an array of predictions matching the input order:

{
  "success": true,
  "data": {
    "predictions": ["low", "high", "medium"]
  }
}

Batch Size Limits​

ParameterLimit
Maximum samples per request1,000
Maximum payload size10 MB

For larger batches, split your data into multiple requests.

Performance Considerations​

  • Async recommended: Large batches may take longer than the sync timeout
  • Parallel requests: For very large datasets, send multiple batch requests in parallel
  • Webhooks: Use webhooks to get notified when large batch jobs complete

Example: Processing a CSV File​

import pandas as pd
import requests

# Load your data
df = pd.read_csv("data.csv")

# Convert to API format
input_data = {
    col: df[col].tolist()
    for col in df.columns
}

# Make batch prediction
response = requests.post(
    f"{API_URL}/projects/{PROJECT_ID}/predict",
    headers={"X-API-Key": API_KEY, "Content-Type": "application/json"},
    json={"input": input_data}
)

# Add predictions to dataframe
df["prediction"] = response.json()["data"]["predictions"]