AI Engineering & REST APIs
As an AI Engineer, you do not spend weeks compiling terabytes of data to train a Large Language Model (LLM) from scratch. Instead, you utilize pre-trained models (like Google Gemini, Claude, or GPT) via APIs and integrate them with databases, user interfaces, and business logic.
20.1.1 What is an API?
An API (Application Programming Interface) is a bridge that allows different software applications to talk to one another. When building AI apps:
- Your Python code sends a request containing a user's prompt to the AI provider's server.
- The provider's server processes the prompt through their LLM.
- Their server sends back a response containing the generated text answer.
20.1.2 Request Types: GET vs. POST
Web APIs use standard HTTP request methods to transfer data:
- GET Request: Used to retrieve data from a server. All parameters are passed directly inside the URL query string (e.g., fetching weather forecasts).
- POST Request: Used to send data to a server to create or process something. AI APIs always use POST requests because prompts are long, complex, and need to be sent securely inside the request body.
20.1.3 HTTP Headers & Bearer Tokens
When sending a POST request to an AI provider, your request must include HTTP Headers to provide metadata:
Content-Type: application/json: Tells the server that the request body is formatted as a JSON object.Authorization: Bearer YOUR_API_KEY: Authenticates your request. The API key behaves like a passport, identifying you and checking if you have permissions to use the model.
20.1.4 API Payloads (JSON)
An API payload is the data block sent inside the request body. In AI, this is formatted as a JSON object (which is read in Python as a nested dictionary):
{
"model": "gemini-1.5-flash",
"contents": [
{
"parts": [
{"text": "Write a 3-word poem about coding."}
]
}
]
}
The server processes this JSON body, runs the prompt, and returns a JSON response containing the generated text answer.