Skip to main content

​
Getting Started with Eden AI V3 API

Welcome to the Eden AI V3 API! This guide will help you understand V3’s unified architecture and make your first API calls.

​
Overview

Eden AI V3 introduces a revolutionary approach to AI API integration with:
  • Universal AI Endpoint - Single endpoint for all non-LLM features
  • OpenAI-Compatible Format - Drop-in replacement for OpenAI’s API
  • Persistent File Storage - Upload once, use in multiple requests
  • Built-in API Discovery - Explore features and schemas programmatically
  • Mandatory Streaming - All LLM responses use Server-Sent Events (SSE)
If you were an user before 2026/01/05, you still have access to the previous version: https://old-app.edenai.run/. We’ll continue supporting the old version until the end of 2026. It your’re looking for the documentation, you can find it here

​
V3 Architecture

V3 uses a model string format instead of separate provider parameters: 
feature/subfeature/provider[/model]
Examples:
  • text/ai_detection/openai/gpt-4
  • ocr/financial_parser/google
  • image/generation/openai/dall-e-3
  • openai/gpt-4 (for LLM endpoints)
This unified format allows a single endpoint to handle all features intelligently.

​
Prerequisites

Before you start, you’ll need:
  1. API Token - Get your token from the Eden AI dashboard
  2. HTTP Client - Use cURL, Python requests, or any HTTP client
  3. Credits - Ensure your account has sufficient credits

​
Base URL

All V3 API endpoints are available at: 
https://api.edenai.run/v3

​
Authentication

All requests must include your API token in the Authorization header:

​
Your First API Call: Universal AI

The Universal AI endpoint handles all non-LLM features through a single endpoint. Let’s analyze some text:

​
Response Format

All V3 responses follow a consistent structure: 
{
  "status": "success",
  "cost": 0.0001,
  "provider": "openai",
  "feature": "text",
  "subfeature": "ai_detection",
  "output": {
    "ai_score": 0.85,
    "is_ai_generated": true
  }
}

​
Your First LLM Call: OpenAI-Compatible

The LLM endpoint provides OpenAI-compatible chat completions with mandatory streaming:

​
Streaming Response Format

LLM responses use Server-Sent Events (SSE): 
data: {"id":"chatcmpl-123","object":"chat.completion.chunk","choices":[{"delta":{"content":"Hello"},"index":0}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","choices":[{"delta":{"content":"!"},"index":0}]}

data: [DONE]

​
Working with Files

V3 introduces persistent file storage. Upload files once, then reference them by ID:

​
Step 1: Upload a File

​
Step 2: Use the File in Requests

​
Understanding Model Strings

The model string is the key to V3’s unified architecture. Here’s the format: 
feature/subfeature/provider[/model]

​
Breaking It Down

ComponentDescriptionExample
featureCategory of AI capabilitytext, ocr, image, translation
subfeatureSpecific functionalityai_detection, moderation, ocr, generation
providerAI provideropenai, google, amazon, anthropic
modelSpecific model (optional)gpt-4, claude-3-5-sonnet, gemini-pro

​
Examples

Text Analysis:
  • text/ai_detection/openai/gpt-4 - Detect AI-generated text with GPT-4
  • text/moderation/openai - Content moderation with OpenAI’s default model
  • text/embeddings/cohere/embed-english-v3.0 - Generate embeddings with Cohere
OCR:
  • ocr/financial_parser/google - Extract financial information with Google DocumentAI
  • ocr/identity_parser/amazon - Parse ID documents with Amazon Textract
  • ocr/invoice_parser/microsoft - Extract invoice data with Azure
Image:
  • image/generation/openai/dall-e-3 - Generate images with DALL-E 3
  • image/object_detection/google - Detect objects with Google Vision
  • image/face_detection/amazon - Detect faces with AWS Rekognition
LLM (simplified format):
  • openai/gpt-4 - Chat with GPT-4
  • anthropic/claude-3-5-sonnet-20241022 - Chat with Claude
  • google/gemini-pro - Chat with Gemini

​
API Discovery

V3 includes built-in endpoints to explore available features programmatically:

​
Error Handling

V3 uses standard HTTP status codes with detailed error messages: 
{
  "status": "error",
  "error": {
    "code": "invalid_model_string",
    "message": "Model string format must be feature/subfeature/provider[/model]",
    "details": {
      "provided": "invalid/model",
      "expected": "feature/subfeature/provider[/model]"
    }
  }
}

​
Common Status Codes

  • 200 - Success
  • 400 - Bad Request (invalid model string or input)
  • 401 - Unauthorized (invalid API token)
  • 402 - Payment Required (insufficient credits)
  • 404 - Not Found (feature/provider not available)
  • 422 - Validation Error (invalid request body)
  • 429 - Rate Limit Exceeded
  • 500 - Internal Server Error

​
V3 vs V2: Key Differences

AspectV2V3
EndpointsFeature-specific (/v2/text/moderation)Universal (/v3/universal-ai) + LLM
Provider Formatproviders parameterModel string (text/moderation/openai)
File HandlingPer-request uploadsPersistent storage with file IDs
LLM StreamingOptionalMandatory (always SSE)
API DiscoveryDocumentation onlyBuilt-in /v3/info endpoints
OpenAI CompatibilityCustom formatNative OpenAI format

​
Next Steps

Now that you understand V3 basics, explore specific features:

​
Universal AI

​
OpenAI-Compatible LLM

​
File Management

​
API Discovery

​
Need Help?