In this guide, we'll learn how to build a Python API using FastAPI and integrate it with OpenAI's ChatGPT. By the end of this post, you'll be able to create RESTful endpoints and utilize the power of OpenAI's ChatGPT. Let's get started!

Step 1: Setup

First, we'll need to install the required packages. For this project, we're going to use FastAPI and Uvicorn for creating and running our API, and OpenAI to use the GPT-3 model for text generation.

Create a new file requirements.txt and add the following lines:

fastapi
uvicorn
openai

Install these packages using pip:

pip install -r requirements.txt

Step 2: Creating the API

We start by initializing our FastAPI app in the main.py file. Additionally, we import the Pydantic BaseModel, which allows us to define how our data should be modeled.

from fastapi import FastAPI
from pydantic import BaseModel
from utils import generate_description

app = FastAPI()

Next, we define our data models for the product and order. We're using Pydantic's BaseModel for this:

Pydantic:

Pydantic is a data validation library that uses Python type annotations. The principal advantage of Pydantic is the ease with which complex data schemas can be declared and validated. Pydantic's BaseModel forms the foundation for all models. It provides functionality for model initialization, serialization to JSON, model validation, etc.

Here, Order and Product are subclasses of Pydantic's BaseModel. By declaring our data classes in this way, Pydantic will automatically handle data validation, serialization, and documentation.

• Order class: This model describes the structure of an order in our application. It has two fields: product, which should be a string, and units, which should be an integer. When we create an instance of Order, Pydantic will ensure these types are respected, throwing an error if we attempt to assign an inappropriate value.

• Product class: Similarly, this model describes a product in our application. It has two fields: name, which should be a string, and notes, which should also be a string.

class Order(BaseModel):
    product: str
    units: int

class Product(BaseModel):
    name: str
    notes: str

Python Type Hints:

Python 3.5 introduced optional "type hints". You can specify the expected type of function arguments and return values. They don't affect the runtime behavior of your program but serve as documentation and are used by static type checkers, linters, and IDE features.

In the context of FastAPI, these type hints are used for:

• Data validation: For example, if we specify a function parameter to be of type int, and the client sends a string, FastAPI will send a helpful and descriptive error message.

• Data serialization: FastAPI can convert complex data types (like datetime objects) into formats that can be easily converted to JSON.

• API documentation: FastAPI uses these type hints to automatically generate API documentation.

So, in our Order and Product models, str and int are type hints, declaring what type of data each attribute should hold.

These concepts form the backbone of FastAPI and contribute to its fast, flexible, and developer-friendly nature.

class Order(BaseModel):
    product: str
    units: int

class Product(BaseModel):
    name: str
    notes: str

With our API and data models ready, we can now create our endpoints:

1. A GET endpoint /ok which returns a simple "ok" message.

2. A GET endpoint /hello which takes an optional query parameter name and returns a personalized greeting.

3. Two POST endpoints /orders and /orders_pydantic which take product information and return a confirmation message. The difference between these two endpoints is that /orders takes query parameters while /orders_pydantic takes a JSON body.

@app.get("/ok")
async def ok_endpoint():
    return {"message": "ok"}

@app.get("/hello")
async def hello_endpoint(name: str = 'World'):
    return {"message": f"Hello, {name}!"}

@app.post("/orders")
async def place_order(product: str, units: int):
    return {"message": f"Order for {units} units of {product} placed successfully."}

@app.post("/orders_pydantic")
async def place_order(order: Order):
    return {"message": f"Order for {order.units} units of {order.product} placed successfully."}

Step 3: Integrating with OpenAI ChatGPT

We're going to use OpenAI's GPT-3 model to generate product descriptions.

In utils.py, we're initializing the OpenAI API key and defining a function generate_description which takes in product details and returns a generated description.

import openai

openai.api_key = ""  # Add your OpenAI API key here

def generate_description(input):
    messages = [
        {"role": "user",
         "content": """As a Product Description Generator, Generate multi paragraph rich text product description with emojis from the information provided to you' \n"""},
    ]

    messages.append({"role": "user", "content": f"{input}"})
    completion = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=messages
    )
    reply = completion.choices[0].message.content
    return reply

Back in main.py, we create a POST endpoint /product_description which takes in product details and returns the generated description:

@app.post("/product_description")
async def generate_product_description(product: Product):
    description = generate_description(f"Product name: {product.name}, Notes: {product.notes}")
    return {"product_description": description}

Step 4: Running the API and Making Requests

Run your API using Uvicorn:

uvicorn main:app --reload

Now you can make requests to your API. Here are some examples of how to do it in Python using the requests library:

For the /orders endpoint, which uses query parameters:

import requests

url = 'http://127.0.0.1:8000/orders'
headers = {
    'accept': 'application/json',
}
params = {
    'product': 'laptop',
    'units': '1'
}

response = requests.post(url, headers=headers, params=params)

print(response.json())

For the /orders_pydantic endpoint, which uses a JSON body:

import requests
import json

url = 'http://127.0.0.1:8000/orders_pydantic'
headers = {
    'accept': 'application/json'
}
params = {
    'product': 'laptop',
    'units': '1'
}

response = requests.post(url, headers=headers, data=json.dumps(params))

print(response.json())

import requests
import json

url = 'http://127.0.0.1:8000/product_description'
headers = {
    'accept': 'application/json',
    'Content-Type': 'application/json'
}
data = {
    "name": "Laptop",
    "notes": "4GB RAM . 256 GB Disk"
}

response = requests.post(url, headers=headers, data=json.dumps(data))

print(response.json())

Accessing the Swagger UI Documentation

FastAPI provides out-of-the-box support for generating interactive API documentation with Swagger UI. After you've started your application using uvicorn main:app --reload, navigate to localhost:8000/docs in your web browser.

Here, you'll find a list of your defined endpoints (/ok, /hello, /orders, /orders_pydantic, and /product_description). Swagger UI allows you to experiment with your API directly from the browser: you can "Try it out", fill in the necessary fields, and then "Execute" to send a request.

This intuitive interface also presents the schemas of our Pydantic models (Order and Product), making it an excellent resource for anyone who needs to understand or interact with your API.

That's it! You now have a fully functional API with FastAPI, integrated with OpenAI's ChatGPT. To learn more, check out the accompanying video tutorial (put the link of the video here) on YouTube.

Full Code : https://github.com/PradipNichite/Youtube-Tutorials/tree/main/fastapi_openai/app

FastAPI is a powerful tool for creating APIs, and when combined with OpenAI's ChatGPT, it becomes even more potent. By understanding these fundamentals, you can create more complex and robust APIs