Multi-Provider LLM API — Azure Function

A Python Azure Function that provides a unified REST API for interacting with multiple Large Language Model (LLM) providers and user management. Built natively on the Azure Functions Python v2 programming model with Azure Table Storage for user data persistence.

Supported Providers

Provider	SDK	Models (examples)
OpenAI	openai	GPT-4o, GPT-4o Mini, GPT-4 Turbo
Azure OpenAI	openai (Azure mode)	GPT-4o, GPT-4o Mini (Azure deployments)
Anthropic	anthropic	Claude Sonnet 4, Claude 3.5 Haiku
AWS Bedrock	boto3	Claude (Anthropic) & Llama 3 (Meta) on Bedrock

Project Structure

├── function_app.py              # Azure Functions entry point (all HTTP triggers)
├── models_config.json           # Model + user storage configuration (editable)
├── host.json                    # Azure Functions host configuration
├── local.settings.json          # Local environment variables (git-ignored)
├── requirements.txt             # Python dependencies
├── .gitignore                   # Git ignore rules
├── services/
│   ├── __init__.py              # Services package init
│   ├── base_provider.py         # Abstract base class for all providers
│   ├── openai_provider.py       # OpenAI API integration
│   ├── azure_openai_provider.py # Azure OpenAI integration
│   ├── anthropic_provider.py    # Anthropic Claude integration
│   ├── bedrock_provider.py      # AWS Bedrock integration (Claude + Llama)
│   ├── provider_factory.py      # Factory to instantiate the correct provider
│   ├── config_loader.py         # Configuration file loader and accessor
│   └── user_service.py          # User CRUD operations via Azure Table Storage
└── README.md

API Endpoints

All endpoints are prefixed with /api.

1. List Models

GET /api/models

Returns all available models from the configuration file.

Response (200):

{
  "models": [
    {
      "modelId": "openai-gpt4o",
      "modelName": "gpt-4o",
      "providerName": "OpenAI",
      "temperature": 0.7,
      "supportsVision": true,
      "description": "OpenAI GPT-4o - multimodal model with vision support"
    }
  ],
  "count": 11
}

2. Text Chat Completion

POST /api/chat
Content-Type: application/json

Sends a text prompt to the specified model and returns the response.

Request Body:

{
  "prompt": "Explain quantum computing in simple terms.",
  "modelId": "openai-gpt4o"
}

Response (200):

{
  "response": "Quantum computing uses quantum bits...",
  "modelId": "openai-gpt4o",
  "providerName": "OpenAI"
}

3. Vision Chat Completion (Text + Image)

POST /api/chat/vision
Content-Type: application/json

Sends a text prompt along with an image to a vision-capable model.

Request Body:

{
  "prompt": "Describe what you see in this image.",
  "imageContent": "<base64-encoded-image-data-or-URL>",
  "modelId": "openai-gpt4o",
  "imageMediaType": "image/png"
}

• imageContent: Base64-encoded image data or an image URL (starting with http).
• imageMediaType (optional): MIME type of the image. Defaults to image/png. Supported: image/png, image/jpeg, image/gif, image/webp.

Response (200):

{
  "response": "The image shows a landscape with mountains...",
  "modelId": "openai-gpt4o",
  "providerName": "OpenAI"
}

4. Ping / Health Check

GET /api/ping

Returns the health status of the function.

Response (200):

{
  "status": "healthy",
  "timestamp": "2026-02-09T10:30:00+00:00",
  "service": "Multi-Provider LLM API"
}

5. Add New User

POST /api/users
Content-Type: application/json

Creates a new user in Azure Table Storage. The password should be sent already base64-encoded (btoa format) from the UI.

Request Body:

{
  "username": "johndoe",
  "password": "cGFzc3dvcmQxMjM=",
  "isAdmin": false
}

Response (201):

{
  "user": {
    "username": "johndoe",
    "password": "cGFzc3dvcmQxMjM=",
    "isAdmin": false
  }
}

Error (409):

{
  "error": "User 'johndoe' already exists."
}

6. Delete User

DELETE /api/users/{username}

Deletes a user by username.

Response (200):

{
  "message": "User 'johndoe' deleted successfully."
}

Error (404):

{
  "error": "User 'johndoe' not found."
}

7. Update User

PUT /api/users/{username}
Content-Type: application/json

Updates a user's password and/or admin flag. At least one field must be provided. The password should be sent already base64-encoded (btoa format) from the UI.

Request Body:

{
  "password": "bmV3cGFzc3dvcmQ=",
  "isAdmin": true
}

Response (200):

{
  "user": {
    "username": "johndoe",
    "password": "bmV3cGFzc3dvcmQ=",
    "isAdmin": true
  }
}

Error (404):

{
  "error": "User 'johndoe' not found."
}

8. Validate User

POST /api/users/validate
Content-Type: application/json

Validates a user's credentials. Takes a username and a base64-encoded (btoa) password, and returns the full user object if the credentials are valid.

Request Body:

{
  "username": "johndoe",
  "password": "cGFzc3dvcmQxMjM="
}

Response (200):

{
  "user": {
    "username": "johndoe",
    "password": "cGFzc3dvcmQxMjM=",
    "isAdmin": false
  }
}

Error (401):

{
  "error": "Invalid username or password."
}

Configuration

Models Configuration (models_config.json)

The configuration file contains two sections: model definitions and user storage settings.

Models

Each model entry has:

Field	Type	Description
modelId	string	Unique identifier used in API requests
modelName	string	Provider-specific model name / deployment name
providerName	string	One of: OpenAI, Azure OpenAI, Anthropic, AWS Bedrock
temperature	float	Default sampling temperature for the model
supportsVision	bool	Whether the model accepts image input
description	string	Human-readable description

To add a new model, simply add a new entry to the models array in the JSON file.

User Storage

The userStorage section configures Azure Table Storage for user management:

Field	Type	Description
tableName	string	Name of the Azure Table Storage table (default: Users)
partitionKey	string	Partition key used for all user entities (default: user)
connectionStringEnvVar	string	Environment variable name that holds the Azure Storage connection string

Environment Variables

Set these in local.settings.json for local development, or in Azure Function Application Settings for production:

Variable	Required For	Description
AZURE_STORAGE_CONNECTION_STRING	User Management	Connection string for Azure Table Storage
OPENAI_API_KEY	OpenAI	API key for OpenAI
AZURE_OPENAI_API_KEY	Azure OpenAI	API key for Azure OpenAI resource
AZURE_OPENAI_ENDPOINT	Azure OpenAI	Azure OpenAI resource endpoint URL
AZURE_OPENAI_API_VERSION	Azure OpenAI	API version (default: 2024-10-21)
ANTHROPIC_API_KEY	Anthropic	API key for Anthropic
AWS_ACCESS_KEY_ID	AWS Bedrock	AWS access key ID
AWS_SECRET_ACCESS_KEY	AWS Bedrock	AWS secret access key
AWS_REGION	AWS Bedrock	AWS region (default: us-east-1)

Note: You only need to configure the environment variables for the providers and features you intend to use. AZURE_STORAGE_CONNECTION_STRING is required for all user management endpoints.

Local Development

Prerequisites

• Python 3.9+
• Azure Functions Core Tools v4

Setup

1. Install dependencies:

   pip install -r requirements.txt

2. Configure environment variables: Copy and edit local.settings.json with your API keys and Azure Storage connection string.

3. Run the function locally:

   func start

4. Test the endpoints:

   # Health check
   curl http://localhost:7071/api/ping
   
   # List models
   curl http://localhost:7071/api/models
   
   # Text chat
   curl -X POST http://localhost:7071/api/chat \
     -H "Content-Type: application/json" \
     -d '{"prompt": "Hello, world!", "modelId": "openai-gpt4o"}'
   
   # Vision chat
   curl -X POST http://localhost:7071/api/chat/vision \
     -H "Content-Type: application/json" \
     -d '{"prompt": "Describe this image.", "imageContent": "<base64-data>", "modelId": "openai-gpt4o"}'
   
   # Add a user
   curl -X POST http://localhost:7071/api/users \
     -H "Content-Type: application/json" \
     -d '{"username": "johndoe", "password": "cGFzc3dvcmQxMjM=", "isAdmin": false}'
   
   # Validate a user
   curl -X POST http://localhost:7071/api/users/validate \
     -H "Content-Type: application/json" \
     -d '{"username": "johndoe", "password": "cGFzc3dvcmQxMjM="}'
   
   # Update a user
   curl -X PUT http://localhost:7071/api/users/johndoe \
     -H "Content-Type: application/json" \
     -d '{"isAdmin": true}'
   
   # Delete a user
   curl -X DELETE http://localhost:7071/api/users/johndoe

Deployment

Deploy to Azure using the Azure Functions Core Tools or the Azure CLI:

func azure functionapp publish <YOUR_FUNCTION_APP_NAME>

Make sure to configure the required environment variables in your Azure Function App's Application Settings before making API calls.

Architecture

Client Request
      │
      ▼
Azure Functions (function_app.py)
      │  Native HTTP Triggers
      ├──────────────────────────────────┐
      │                                  │
      ▼                                  ▼
LLM Endpoints                    User Endpoints
(models, chat, chat/vision)      (add, delete, update, validate)
      │                                  │
      ▼                                  ▼
Config Loader                    User Service
(config_loader.py)               (user_service.py)
      │                                  │
      ▼                                  ▼
Provider Factory                 Azure Table Storage
(provider_factory.py)
      │
      ▼
Provider Service
(e.g., openai_provider.py)
      │
      ▼
LLM Provider API
(OpenAI / Azure / Anthropic / Bedrock)

Error Handling

All endpoints return consistent JSON error responses:

{
  "error": "Descriptive error message."
}

HTTP Status	Meaning
400	Bad request (missing fields, unsupported model features)
401	Invalid credentials (user validation failed)
404	Resource not found (model ID or username)
409	Conflict (user already exists)
500	Internal server error (API call failure, configuration error)