API Key

Overview

The Model Hub API enables developers to interact with Sahara AI's platform for creating, deploying, and managing AI pipelines. Designed to align with the OpenAI Assistants API, it provides familiar endpoints and workflows for seamless integration. This document outlines how to authenticate, generate API keys, and use the API to develop and deploy pipelines.

1. API Authentication

Authentication Mechanisms

Sahara supports two authentication methods:

1. Bearer Token: Aligns with OpenAI's token-based authentication standard.

2. API Key: Provides more granular control with configurable permissions.

Steps to Generate an API Key

1. Access the Developer Portal:

• Navigate to the API Key Management under your User Profile icon dropdown section.

1. Generate API Key:

• Click Generate New API Key.

1. Assign permissions:

• Read: Fetch model details.

• Write: Create and deploy pipelines.

• Admin: Full access.

• Optionally, set an expiration date.

1. Secure Your API Key:

• Copy the key immediately; it will not be shown again.

• Store it securely.

Authentication Headers

Include the generated API key or bearer token in the header of your HTTP requests:

Authorization: Bearer YOUR_API_KEY

Error Handling

Common authentication errors:

• 401 Unauthorized: Invalid or missing token.

• 403 Forbidden: Insufficient permissions.

• 429 Too Many Requests: Exceeded rate limits.

2. API Endpoints

Endpoint: Create a Pipeline

URL: /v1/pipelines/create

Method: POST

Description: Creates a new AI pipeline.

Request Body:

{
  "name": "Pipeline Name",
  "type": "RAG",  // or "Prompt"
  "description": "Description of the pipeline",
  "instructions": "You are a helpful assistant.",
  "conversation_starter": "How can I assist you today?",
  "model": "model-id",
  "vaults": ["vault-id1", "vault-id2"]  // Required for RAG pipelines
}

Response:

/{
  "status": "success",
  "pipeline_id": "pipeline-12345",
  "endpoint": "https://api.sahara.ai/v1/pipelines/pipeline-12345"
}

Endpoint: Test Pipeline

URL: /v1/pipelines/{pipeline_id}/test

Method: POST

Description: Allows developers to test their pipeline before deployment.

Request Body:

{
  "input": "What is the weather today?"
}

Response:

{
  "status": "success",
  "response": {is
    "assistant": {
      "role": "assistant",
      "content": "The weather today is sunny."
    }
  },
  "usage": {
    "prompt_tokens": 12,
    "completion_tokens": 7,
    "total_tokens": 19
  }
}

Endpoint: Deploy Pipeline

URL: /v1/pipelines/{pipeline_id}/deploy

Method: POST

Description: Deploys the pipeline to Sahara’s compute network.

Request Body:

{
  "compute_provider": "provider-id",
  "config": {
    "temperature": 0.7,
    "max_tokens": 100,
    "top_p": 0.9
  }
}

Response:

{
  "status": "success",
  "api_endpoint": "https://api.sahara.ai/v1/inference/pipeline-12345"
}

Endpoint: Monitor Usage

URL: /v1/pipelines/{pipeline_id}/metrics

Method: GET

Description: Retrieves metrics such as token usage and API call statistics.

Response:

{
  "status": "success",
  "metrics": {
    "token_usage": {
      "prompt_tokens": 1200,
      "completion_tokens": 800,
      "total_tokens": 2000
    },
    "api_calls": 150
  }
}

3. Using the API in a Workflow

Step 1: Authenticate

Use your API key or bearer token in the header of all requests.

Step 2: Create a Pipeline

Send a POST request to /v1/pipelines/create with the required parameters. Ensure all fields are valid.

Step 3: Test the Pipeline

Test the functionality of your pipeline using /v1/pipelines/{pipeline_id}/test.

Step 4: Deploy the Pipeline

Deploy the tested pipeline via /v1/pipelines/{pipeline_id}/deploy and receive the unique inference endpoint.

Step 5: Monitor Metrics

Track usage metrics for optimization and scaling using /v1/pipelines/{pipeline_id}/metrics.

4. Error Handling and Best Practices

Error Codes

• 400 Bad Request: Check request formatting.

• 404 Not Found: Verify pipeline or model IDs.

• 500 Internal Server Error: Retry or contact support.

Best Practices

1. Secure Keys:

• Use environment variables to store API keys securely.

1. Monitor Usage:

• Regularly review metrics to optimize performance.

1. Retry Logic:

• Implement retry logic for transient errors (e.g., 500 Internal Server Error).