AI Gateway

The AI Gateway provides a unified, vendor-agnostic API for accessing AI models across 12 providers. With 302 certified models (as of February 2026), you can access chat completions, multimodal (vision), embeddings, text-to-speech, speech-to-text, image generation, video generation, and music generation through a single consistent interface.

Key Features

• Unified API: Single interface for all providers (OpenAI, Anthropic, Google, Mistral, Cohere, Grok, DeepSeek, ElevenLabs, Stability AI, Black Forest Labs, Runway, Luma, plus custom/vLLM/Kubernetes for self-hosted)
• SDK Compatible: Drop-in replacement for OpenAI SDK - requests via the /v1/ prefix automatically enable OpenAI compatibility mode (strips provider and provider_response fields from responses)
• Multi-Modal Support: Text, vision (image input), audio (TTS/STT), video, and music generation. Vision Language Models (GPT-4o, GPT-5, Claude 3+, Gemini 2.0+, Pixtral, etc.) are classified as multimodal type with vision capability
• Certified Models: Models are loaded from static JSON certification data on gateway startup
• Guardrails: Input and output guardrails with content filtering, PII detection, and prompt injection protection (applied on completions endpoint)
• Consistent Error Handling: Standardized error responses across all providers
• Usage Tracking: Token counting, latency monitoring, and cost tracking
• Organization Scoping: Multi-tenant isolation via X-Organization-ID header on model management endpoints

Authentication

API requests require the X-User-Id header for user identification. Context headers (X-App-Id, X-Workflow-Id, or X-Workspace-Id) are required for TTS, STT, and generation endpoints but are optional for chat and completion endpoints (used for analytics tracking when provided).

Required Headers

Header	Required	Description
X-User-Id	Yes	Authenticated user's ID
X-Organization-ID	Recommended	Organization ID for multi-tenant resource isolation (required for model CRUD and fine-tuning)
X-App-Id	Conditional	App ID (required for TTS/STT/Generations, optional for chat/completions)
X-Workflow-Id	Conditional	Workflow ID (alternative to X-App-Id)
X-Workspace-Id	Conditional	Workspace ID (alternative to X-App-Id)
X-OpenAI-Compat	Optional	Set to true for strict OpenAI SDK compatibility (removes provider and provider_response fields). Automatically enabled when using /v1/ prefix
Content-Type	Yes	application/json for most endpoints

Note: For TTS, STT, and generation endpoints, at least one of X-App-Id, X-Workflow-Id, or X-Workspace-Id must be provided. For chat/completions/embeddings, these headers are optional but recommended for analytics.

OpenAI Compatibility Mode

The gateway supports two path prefixes:

• /v1/ - Automatically enables OpenAI SDK compatibility mode (removes provider and provider_response fields from responses)
• /api/v1/ - Standard gateway format with full response data including provider metadata

You can also enable compatibility mode explicitly on /api/v1/ routes by setting the X-OpenAI-Compat: true header.

Example Request

curl -X POST https://ai-gateway.strongly.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "X-User-Id: your-user-id" \
  -H "X-Organization-ID: your-org-id" \
  -d '{
    "model": "507f1f77bcf86cd799439011",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

The model field should contain the Strongly-generated model ID from your configured models.

Model ID Format

When calling the AI Gateway, you must use the Strongly-generated model ID - the unique identifier assigned when a model is added to your account. This is a MongoDB ObjectId (24-character hex string).

Why use Strongly IDs?

• The same vendor model (e.g., GPT-4o) can be added multiple times with different API keys
• Each instance has a unique Strongly ID for proper isolation and billing
• Enables multi-tenancy where users manage their own model configurations

Getting Model IDs:

• When you add a model via the platform, you receive its unique ID
• Use the /v1/models endpoint to list available models and their IDs
• Model IDs are returned in the _id field

Example model ID: 507f1f77bcf86cd799439011

Quick Start

Chat Completion

// modelId is the Strongly-generated ID from your configured models
const modelId = '507f1f77bcf86cd799439011';

const response = await fetch('https://ai-gateway.strongly.ai/v1/chat/completions', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'X-User-Id': userId,
    'X-App-Id': appId  // optional for chat, used for analytics
  },
  body: JSON.stringify({
    model: modelId,
    messages: [
      { role: 'system', content: 'You are a helpful assistant.' },
      { role: 'user', content: 'What is the capital of France?' }
    ],
    max_tokens: 1000,
    temperature: 0.7
  })
});

const data = await response.json();
console.log(data.choices[0].message.content);

Text-to-Speech

// ttsModelId is the Strongly-generated ID for your TTS model
const ttsModelId = '507f1f77bcf86cd799439012';

const response = await fetch('https://ai-gateway.strongly.ai/v1/audio/speech', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'X-User-Id': userId,
    'X-App-Id': appId  // required for TTS
  },
  body: JSON.stringify({
    model: ttsModelId,
    input: 'Hello, welcome to Strongly AI!',
    voice: 'alloy',
    response_format: 'mp3'
  })
});

// Response is streaming audio
const audioBlob = await response.blob();

Image Generation

// imageModelId is the Strongly-generated ID for your image model
const imageModelId = '507f1f77bcf86cd799439013';

const response = await fetch('https://ai-gateway.strongly.ai/v1/generations/images', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'X-User-Id': userId,
    'X-App-Id': appId  // required for generations
  },
  body: JSON.stringify({
    model: imageModelId,
    prompt: 'A beautiful sunset over mountains',
    size: '1024x1024',
    quality: 'hd'
  })
});

const data = await response.json();
console.log(data.data[0].url);

API Endpoints Overview

Endpoint	Method	Description
/v1/models	GET	List all available models
/v1/models/{model_id}	GET	Get model details
/v1/models	POST	Create a new model
/v1/models/{model_id}	PUT	Update a model
/v1/models/{model_id}	DELETE	Delete a model
/v1/models/providers	GET	List providers with capabilities
/v1/models/keys	POST	Create an API key
/v1/models/keys	GET	List API keys
/v1/models/keys/{key_id}	GET	Get API key details
/v1/models/keys/{key_id}	DELETE	Delete an API key
/v1/models/keys/{key_id}/deactivate	POST	Deactivate an API key
/v1/chat/completions	POST	Chat completion
/v1/completions	POST	Text completion
/v1/embeddings	POST	Generate embeddings
/v1/audio/speech	POST	Text-to-speech
/v1/audio/speech/voices	GET	List available TTS voices
/v1/audio/speech/models	GET	List TTS models
/v1/audio/transcriptions	POST	Speech-to-text
/v1/audio/translations	POST	Audio translation
/v1/audio/transcriptions/models	GET	List STT models
/v1/audio/transcriptions/languages	GET	List supported STT languages
/v1/generations	POST	Unified generation (image/video/music)
/v1/generations	GET	List generation jobs
/v1/generations/images	POST	Image generation
/v1/generations/videos	POST	Video generation
/v1/generations/music	POST	Music generation
/v1/generations/{job_id}	GET	Get generation job status
/v1/generations/{job_id}	DELETE	Cancel generation job
/v1/generations/models	GET	List generation models
/v1/generations/models/{model_id}/operations/{operation_id}	GET	Get provider operation status

See API Endpoints for complete documentation.

Supported Capabilities

Capability	Providers
Multimodal (Vision)	OpenAI (GPT-4o/4.1/5+, o1/o3/o4), Anthropic (Claude 3+), Google (Gemini 1.5+/2.0+/2.5), Mistral (Pixtral, Large, Medium, Small), Cohere (Aya Vision, Command-A Vision), Grok (Vision)
Chat/Completions	OpenAI, Google, Mistral, Cohere, Grok, DeepSeek
Embeddings	OpenAI, Google, Mistral, Cohere
Text-to-Speech	OpenAI, ElevenLabs, Google
Speech-to-Text	OpenAI, Mistral
Image Generation	OpenAI, Google, Stability AI, Black Forest Labs, Grok
Video Generation	Google, Runway, Luma
Music Generation	ElevenLabs
Reranking	Cohere

Error Handling

The gateway uses a standard error format. Errors are returned as:

{
  "detail": "The model '507f1f77bcf86cd799439011' not found. Please ensure the model is configured in your account."
}

For custom error types (authentication, validation), the format includes structured fields:

{
  "detail": {
    "message": "X-User-Id header is required",
    "error_code": "missing_user_id"
  }
}

Common error codes:

• missing_user_id - X-User-Id header not provided (HTTP 401)
• model_not_found - Requested model doesn't exist or is inactive (HTTP 404)
• validation_error - Invalid request parameters or missing context headers (HTTP 400)
• rate_limit_exceeded - Too many requests (HTTP 429, includes Retry-After header)
• provider_error - Upstream provider returned an error (HTTP 500/502)
• Guardrail violations - Request blocked by input/output guardrails (HTTP 400)

Next Steps

• View all API endpoints
• Browse certified models
• Learn about providers
• Deploy self-hosted models
• Fine-tune models on your custom data
• Configure autoscaling
• Set up monitoring
• Optimize costs