AI Inference Backend - Technical Test (aiahura)

A simple AI Inference Backend built with FastAPI, PostgreSQL, Redis, and Docker, designed to register models, accept inference requests (sync/async), implement RBAC via API keys, handle async tasks with Celery/Redis, provide metrics, and include rate limiting.

---

🚀 Setup and Run Instructions

Follow these steps to get the project running locally using Docker.

Prerequisites

• Docker: Install Docker
• Docker Compose: Install Docker Compose

Steps

1. Clone the repository:

   git clone <your-repo-url>
   cd <your-repo-name>

2. Create Environment File: Rename the example environment file. The default values should work for local setup.

   cp .env.example .env

3. Build and Run Containers: This command builds the images and starts the backend, db, redis, and worker services in detached mode.

   docker-compose up --build -d

   Wait a moment for the containers, especially the database, to initialize.

4. Apply Database Migrations: Run Alembic migrations inside the backend container to create the database tables.

   docker-compose exec backend alembic upgrade head

5. Create API Keys: Generate the necessary API keys using the provided script. You'll need at least one admin key to manage models.

   • To create an Admin key:

     docker-compose exec backend python create_key.py --admin

   • To create a regular User key:

     docker-compose exec backend python create_key.py

   Copy the generated key(s) from the output. You will need them for the X-API-Key header.

6. Access the Application:

   • The API is running at http://localhost:8000.
   • Interactive API documentation (Swagger UI) is available at http://localhost:8000/docs.

---

✅ Testing

The project includes unit and integration tests written using pytest.

• Run all tests: Execute the following command:

  docker-compose exec backend pytest

• Test Location: 6 meaningful tests covering model management, inference requests (sync/async), RBAC, and rate limiting are located in the tests/ directory.

---

📚 API Documentation (Swagger UI)

You can interactively explore and test the API endpoints using the auto-generated Swagger documentation:

http://localhost:8000/docs

---

💻 Example cURL Requests

Replace <your_admin_key> and <your_user_key> with the actual keys you generated.

1. Create a Model (Admin Only):

   curl -X 'POST' \
     'http://localhost:8000/models/' \
     -H 'accept: application/json' \
     -H 'X-API-Key: <your_admin_key>' \
     -H 'Content-Type: application/json' \
     -d '{
     "name": "My Sentiment Model",
     "type": "classification",
     "provider": "Local",
     "config": {}
   }'

2. List Models (Any Active User):

   curl -X 'GET' \
     'http://localhost:8000/models/' \
     -H 'accept: application/json' \
     -H 'X-API-Key: <your_user_key>'

3. Run Synchronous Inference (Any Active User): (Assumes a model with ID=1 exists)

   curl -X 'POST' \
     'http://localhost:8000/inferences/' \
     -H 'accept: application/json' \
     -H 'X-API-Key: <your_user_key>' \
     -H 'Content-Type: application/json' \
     -d '{
     "model_id": 1,
     "mode": "sync",
     "input_data": {
       "prompt": "FastAPI is great!"
     }
   }'

4. Run Asynchronous Inference with Webhook (Any Active User): (Assumes a model with ID=1 exists)

   curl -X 'POST' \
     'http://localhost:8000/inferences/' \
     -H 'accept: application/json' \
     -H 'X-API-Key: <your_user_key>' \
     -H 'Content-Type: application/json' \
     -d '{
     "model_id": 1,
     "mode": "async",
     "input_data": {
       "prompt": "Tell me a story."
     },
     "webhook_url": "[https://httpbin.org/post](https://httpbin.org/post)"
   }'

   (The result will be sent to the webhook URL upon completion)

5. Get Metrics Summary (Any Active User):

   curl -X 'GET' \
     'http://localhost:8000/metrics/summary' \
     -H 'accept: application/json' \
     -H 'X-API-Key: <your_user_key>'

---

Thank you,