Ollama API Wrapper

This project provides a Flask-based API wrapper for Ollama, allowing you to interact with Ollama models using an OpenAI-like API interface. It includes features such as API key authentication, model access control, and JSON response formatting.

Features

• OpenAI-like API for Ollama models
• API key authentication
• Model access control
• Streaming and non-streaming responses
• JSON mode for structured outputs
• Token counting and usage tracking

Prerequisites

• Python 3.7+
• MySQL database
• Ollama running locally or on a accessible server

Installation

1. Clone the repository:

   git clone https://github.com/KillCroop2/ollama-api-wrapper.git
   cd ollama-api-wrapper
   

2. Install the required packages:

   pip install -r requirements.txt
   

3. Set up the environment variables:

   • Copy the .env.template file to .env
   • Fill in the necessary database credentials in the .env file

4. Set up the database:

   • Run the SQL commands in database.sql to create the necessary tables

Configuration

Update the .env file with your database credentials:

DB_HOST=<Your MySQL host>
DB_PORT=<Your MySQL port>
DB_USER=<Your MySQL username>
DB_PASSWORD=<Your MySQL password>
DB_NAME=<Your database name>

Usage

1. Start the Flask server:

   python app.py
   

2. The API will be available at http://localhost:5000

3. Use the following endpoints:

   • /v1/chat/completions: For chat completions
   • /v1/models: To list available models
   • /v1/api_keys: To create a new API key

API Endpoints

Chat Completions

POST /v1/chat/completions

Request body:

{
  "messages": [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Hello, how are you?"}
  ],
  "model": "llama2",
  "stream": false,
  "temperature": 0.7
}

List Models

GET /v1/models

Create API Key

POST /v1/api_keys

Development

• app.py: Main Flask application
• db.py: Database operations
• llm_wrapper.py: Ollama API wrapper
• database.sql: SQL schema

License

This project is licensed under the MIT License.