AI CuratorBeta

Live Capture API

The most powerful feature in AI Curator. Stream data in real-time from any source — for fine-tuning, RAG, or any LLM workflow — via simple HTTP POST requests.

Overview

The Live Capture API turns any tool into a data source. If your tool can make an HTTP POST request with JSON, it can stream data to AI Curator in real-time. Use captured data for fine-tuning or feed it into your RAG pipeline — same capture, different destinations. No batch processing required — data flows as it's generated.

Endpoint
POST http://localhost:3333/api/capture
Real-time
Stream data as it's generated. No delays, no batching.
Simple
Single endpoint, JSON payload. One request to capture anything.
Local
Runs on your machine. No external APIs or cloud services.

POST /api/capture

PropertyTypeRequiredDescription
sourcestringyesIdentifier for the data source (e.g., "slack", "vscode", "my-script")
recordsarrayyesArray of data records — instruction-output pairs for fine-tuning, or content chunks for RAG indexing
records[].instructionstringyesThe instruction or question
records[].outputstringyesThe expected response or answer
records[].systemPromptstringnoSystem prompt for the sample
records[].categorystringnoCategory tag (e.g., "coding", "support", "writing")
records[].qualityRatingnumbernoQuality rating 1–5
records[].contextobjectnoAdditional metadata about the sample

Request Body

JSON
 { "source": "vscode-extension", "records": [ { "instruction": "Explain this function", "output": "This function processes incoming data...", "systemPrompt": "You are a coding assistant", "category": "coding", "qualityRating": 5, "context": { "language": "typescript", "file": "utils.ts" } } ] }

Integration Examples

cURL

Terminal
 curl -X POST http://localhost:3333/api/capture \ -H "Content-Type: application/json" \ -d '{ "source": "slack", "records": [{ "instruction": "How do I reset my password?", "output": "Go to Settings → Account → Reset Password...", "category": "support", "qualityRating": 4 }] }'

JavaScript / Fetch

JavaScript
 await fetch("http://localhost:3333/api/capture", { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ source: "vscode-extension", records: [{ instruction: "Explain this function", output: explanation, context: { code, language }, category: "coding", }], }), });

Python

Python
 import requests requests.post('http://localhost:3333/api/capture', json={ 'source': 'chat-interface', 'records': [{ 'instruction': user_message, 'output': assistant_response, 'qualityRating': 4 if user_thumbs_up else 3 }] })

Shell Pipe

Terminal
 # Extract error-resolution pairs from logs tail -f /var/log/app.log | grep "ERROR" | while read line; do curl -X POST localhost:3333/api/capture \ -d "{\"source\":\"logs\",\"records\":[{\"instruction\":\"$line\",\"output\":\"...\"}]}" done

Response Format

200 OKSamples captured successfully
 { "success": true, "captured": 3, "datasetId": 1 }
400 Bad RequestInvalid request body
 { "success": false, "error": "Missing required field: source" }
Live Capture must be enabled
The Live Capture API is disabled by default. Enable it in Settings → Live Capture in the Web UI before sending requests.

Known Integrations

VS Code Extension

Capture code explanations, refactoring decisions, and debug sessions directly from your IDE.

VS Code
 fetch("http://localhost:3333/api/capture", { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ source: "vscode-extension", records: [{ instruction, output, context }] }) })

OpenWebUI Plugin

Official plugin available in the OpenWebUI community. Automatically captures conversations from your self-hosted AI chat interface.

Available in OpenWebUI Community Plugins