Getting 500 Internal Server Error FastAPI

[abanerjee-wsu]

Hi all, I am developing a simple Agentic AI App using following Stack

Nextjs → Frontend
FastAPI → Backend
Clerk - Authentication
I am seeing 500 Internal Server error when its fastapi endpoint is being accessed by the frontend

Currently receiving 500 error when accessing the fastapi endpoint on vercel. Expecting a success/200 response

**Build Logs**
 
00:06:06.286 Running build in Washington, D.C., USA (East) – iad1
00:06:06.287 Build machine configuration: 2 cores, 8 GB
00:06:06.422 Retrieving list of deployment files...
00:06:06.900 Downloading 28 deployment files...
00:06:07.675 Restored build cache from previous deployment (Aasp2edaCZy3FmQxfbpohHctPaA5)
00:06:08.196 Running "vercel build"
00:06:08.782 Vercel CLI 48.10.3
00:06:09.130 Installing dependencies...
00:06:10.433 
00:06:10.433 up to date in 1s
00:06:10.434 
00:06:10.434 229 packages are looking for funding
00:06:10.434   run `npm fund` for details
00:06:10.462 Detected Next.js version: 15.5.6
00:06:10.467 Running "npm run build"
00:06:10.653 
00:06:10.653 > saas@0.1.0 build
00:06:10.654 > next build
00:06:10.654 
00:06:11.792    ▲ Next.js 15.5.6
00:06:11.793 
00:06:11.862    Linting and checking validity of types ...
00:06:17.067    Creating an optimized production build ...
00:06:21.553  ✓ Compiled successfully in 1960ms
00:06:21.555    Collecting page data ...
00:06:22.930    Generating static pages (0/5) ...
00:06:23.820    Generating static pages (1/5) 
00:06:23.821    Generating static pages (2/5) 
00:06:23.821    Generating static pages (3/5) 
00:06:23.821  ✓ Generating static pages (5/5)
00:06:24.282    Finalizing page optimization ...
00:06:24.285    Collecting build traces ...
00:06:32.012 
00:06:32.015 Route (pages)                                 Size  First Load JS
00:06:32.015 ┌ ○ / (549 ms)                             3.08 kB         146 kB
00:06:32.016 ├   /_app                                      0 B         142 kB
00:06:32.016 ├ ○ /404                                     180 B         143 kB
00:06:32.016 ├ ○ /product (547 ms)                      45.4 kB         188 kB
00:06:32.016 └ ○ /sign-in/[[...index]] (546 ms)         1.51 kB         144 kB
00:06:32.016 + First Load JS shared by all               147 kB
00:06:32.016   ├ chunks/framework-acd67e14855de5a2.js   57.7 kB
00:06:32.016   ├ chunks/main-62a5bcb5d940e2e2.js        36.8 kB
00:06:32.016   ├ chunks/pages/_app-36c145dffa6388ad.js  46.3 kB
00:06:32.016   └ other shared chunks (total)            5.86 kB
00:06:32.017 
00:06:32.017 ○  (Static)  prerendered as static content
00:06:32.017 
00:06:32.069 Collected static files (public/, static/, .next/static): 3.923ms
00:06:33.753 Using Python 3.12 from pyproject.toml
00:06:35.062 Installing required dependencies from uv.lock...
00:06:35.063 Using uv at "/usr/local/bin/uv"
00:06:36.434 Build Completed in /vercel/output [27s]
00:06:36.662 Deploying outputs...
00:06:46.177 Deployment completed
00:06:46.834 Creating build cache...
00:07:00.688 Created build cache: 13.855s
00:07:00.690 Uploading build cache [170.82 MB]
00:07:02.773 Build cache uploaded: 2.083s

Runtime logs

Python process exited with exit status: 1. The logs above can help with debugging the issue.

api/index.py code

import os
from fastapi import FastAPI, Depends
from fastapi.responses import StreamingResponse
from fastapi.middleware.cors import CORSMiddleware
from fastapi_clerk_auth import ClerkConfig, ClerkHTTPBearer, HTTPAuthorizationCredentials
from openai import OpenAI

print("[DEBUG] Starting api/index.py import...")

app = FastAPI()

# Add CORS middleware
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

print("[DEBUG] FastAPI app and CORS configured")

# Configure Clerk authentication
clerk_config = ClerkConfig(jwks_url=os.getenv("CLERK_JWKS_URL"))
clerk_guard = ClerkHTTPBearer(clerk_config)

print(f"[DEBUG] Clerk configured with JWKS URL: {os.getenv('CLERK_JWKS_URL')}")
print(f"[DEBUG] OpenAI API key present: {bool(os.getenv('OPENAI_API_KEY'))}")


@app.get("/api/health")
@app.get("/health")
async def health_check():
    """Health check endpoint that doesn't require authentication"""
    return {"status": "healthy", "service": "api"}


@app.get("/api/idea")
@app.get("/idea")
def generate_idea(creds: HTTPAuthorizationCredentials = Depends(clerk_guard)):
    """
    Generate a business idea using OpenAI GPT-4o-mini with streaming response.
    Requires Clerk authentication via Authorization header.
    """
    user_id = creds.decoded["sub"]  # User ID from JWT
    print(f"[INFO] Generating idea for user: {user_id}")

    client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
    prompt = [
        {
            "role": "user",
            "content": "Reply with a new business idea for AI Agents, formatted with headings, sub-headings and bullet points"
        }
    ]

    stream = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=prompt,
        max_tokens=500,
        temperature=0.7,
        stream=True
    )

    def event_stream():
        try:
            for chunk in stream:
                if chunk.choices[0].delta.content is not None:
                    text = chunk.choices[0].delta.content
                    # Escape newlines for SSE format
                    safe_text = text.replace('\n', '\\n')
                    yield f"data: {safe_text}\n\n"
            yield "data: [DONE]\n\n"
        except Exception as e:
            print(f"[ERROR] Stream error: {e}")
            yield f"data: [ERROR] {str(e)}\n\n"

    return StreamingResponse(
        event_stream(),
        media_type="text/event-stream",
        headers={
            "Cache-Control": "no-cache",
            "X-Accel-Buffering": "no"
        }
    )


@app.get("/api/user")
@app.get("/user")
def get_user_info(creds: HTTPAuthorizationCredentials = Depends(clerk_guard)):
    """Get information about the authenticated user."""
    return {
        "user_id": creds.decoded.get("sub"),
        "email": creds.decoded.get("email"),
        "claims": creds.decoded
    }


# Vercel serverless function handler
handler = app

print("[DEBUG] Module initialization complete")

Current project structure

Project structure

aas/
├── .env.local
├── .gitignore
├── README.md
├── eslint.config.mjs
├── next-env.d.ts
├── next.config.ts
├── package-lock.json
├── package.json
├── postcss.config.mjs
├── pyproject.toml
├── requirements.txt
├── run-api.sh
├── runtime.txt
├── tsconfig.json
├── tsconfig.tsbuildinfo
├── uv.lock
├── vercel.json
│
├── api/
│   ├── __pycache__/
│   │   └── index.cpython-314.pyc
│   └── index.py
│
├── pages/
│   ├── _app.tsx
│   ├── index.tsx
│   ├── product.tsx
│   │
│   ├── api/
│   │   └── _debug/
│   │
│   └── sign-in/
│       └── [[...index]].tsx
│
├── public/
│   ├── favicon.ico
│   ├── file.svg
│   ├── globe.svg
│   ├── next.svg
│   ├── vercel.svg
│   └── window.svg
│
└── styles/
    └── globals.css

[system]

There’s another community post with 404 debugging tips that might be helpful. Please give these solutions a try and let us know how it goes.

Debugging 404 Errors

Sometimes things don’t go as expected when deploying a website. If you see a 404 on your site, that means the page or resource couldn’t be found. There are many possible causes, and the absence of something can be tricky debug if you don’t know where to look. Here are some things you can do to find the cause and fix it. Debugging Tips Check the error code If you see a mostly white screen with 404: NOT_FOUND along with a Code and and ID then you can click the blue info box below the error deta…

A human should be around soon to offer more advice. But you can also get helpful information quickly by asking v0.

[fmoor]

I tried using the python source you posted and it seems that the handler variable might be the problem here. If you are using a WSGI or ASGI framework like fastapi then you don’t want the handler variable. Vercel interprets handler as a BaseHTTPRequestHandler value. There is a runtime check that asserts that hander is a subclass of BaseHTTPRequestHandler. If this check fails then the API call fails with status 500. The relevant blerb in the docs is here.