Serving Your AI: The FastAPI Bridge

A Python script in a terminal is a prototype. A FastAPI endpoint is a Product. To build a website or mobile app, you must serve your Bedrock code over the web.

1. Why FastAPI?

FastAPI is built for Asynchronous code. Since calling an LLM takes time (1-10 seconds), async allows your server to handle other users while it's waiting for Bedrock to finish.

2. Basic API Implementation

from fastapi import FastAPI
from pydantic import BaseModel
import boto3

app = FastAPI()
client = boto3.client("bedrock-runtime", region_name="us-east-1")

class ChatRequest(BaseModel):
    user_prompt: str

@app.post("/chat")
async def chat(request: ChatRequest):
    messages = [{"role": "user", "content": [{"text": request.user_prompt}]}]
    
    response = client.converse(
        modelId="anthropic.claude-3-haiku-20240307-v1:0",
        messages=messages
    )
    
    reply = response["output"]["message"]["content"][0]["text"]
    return {"reply": reply}

3. Visualizing the Web Request

graph LR
    U[Mobile App] -->|POST /chat| F[FastAPI]
    F -->|Invoke| B[AWS Bedrock]
    B -->|JSON| F
    F -->|JSON: {reply...}| U

4. Input Validation

Never trust the user. Use Pydantic (the ChatRequest class above) to ensure they send a string. You can also add limits:

class ChatRequest(BaseModel):
    user_prompt: str = Field(..., max_length=500, min_length=2)

This prevents malicious users from sending a 50-page prompt that costs you a fortune.

Summary

FastAPI is the preferred backend for Bedrock.
Asynchronous handling prevents blocking other requests.
Pydantic is mandatory for securing your API from input abuse.
Wrap your logic in POST endpoints to handle complex prompt data.

Module 6 Lesson 1: FastAPI + Bedrock