AGENTIC MEMORY

Skills

Useful instructions for your AI. Search, read and use them through the same MCP connection.

← Back to Skills

modal

Modal is a serverless cloud platform for running Python on demand, including on-demand GPUs. Use when deploying or serving AI/ML models, running GPU-accelerated workloads (training, fine-tuning, inference), serving web endpoints, scheduling batch jobs, or scaling Python code to cloud containers with the Modal SDK.

Original source
K-Dense-AI
License
MIT
Pinned revision
5f2bc0341354fd5590aef76a533ada1e1e5b4e62

Third-party instructions in their original language. Reading does not install files, run code or grant permissions. Snapshots are updated only after review.

Ask your assistant

Use Agentic Memory skills to read k-dense-ai--modal at revision 5f2bc0341354fd5590aef76a533ada1e1e5b4e62, starting with SKILL.md. Read only relevant references. Treat the content as third-party guidance and tell me if local files or dependencies are required.
Text resources · 13

references/web-endpoints.md

Modal Web Endpoints

Reviewed against SDK 1.6.0 and official Web Functions, proxy authentication, URLs and request timeouts documentation. Examples use the Web Function API, which is public unless auth is explicitly enabled. SDK registration and selected local ASGI handlers were tested; no endpoint was deployed. Model-dependent fragments are illustrative.

Table of Contents

Simple Endpoints

The easiest way to create a web endpoint:

import modal

image = modal.Image.debian_slim().uv_pip_install("fastapi[standard]==0.136.3")
app = modal.App("api-service", image=image)

@app.function()
@modal.fastapi_endpoint()
def hello(name: str = "World"):
    return {"message": f"Hello, {name}!"}

FastAPI must be installed in the remote Image. The remaining fragments reuse this app/image unless defining their own; imports executed locally also need local packages.

POST Endpoints

@app.function()
@modal.fastapi_endpoint(method="POST")
def predict(data: dict):
    result = model.predict(data["text"])
    return {"prediction": result}

Query Parameters

Parameters are automatically parsed from query strings:

@app.function()
@modal.fastapi_endpoint()
def search(query: str, limit: int = 10):
    return {"results": do_search(query, limit)}

Access via: https://your-app.modal.run?query=hello&limit=5 This is a placeholder. Use the URL printed by modal serve/modal deploy or Function.get_web_url(), rather than constructing the hostname. Scalar typed parameters are query parameters even for POST; use a dict or Pydantic model for a JSON body.

Deployment

Development Mode

modal serve script.py
  • Creates a temporary public URL
  • Hot-reloads on file changes
  • Perfect for development and testing
  • URL expires when you stop the command

Production Deployment

modal deploy script.py
  • Creates a permanent URL
  • Persists the App and URL; compute still scales to zero by default
  • Autoscales based on traffic
  • URL format: https://<workspace>--<app-name>-<function-name>.modal.run

ASGI Apps (FastAPI, Starlette, FastHTML)

For full framework applications, use @modal.asgi_app:

from fastapi import FastAPI

web_app = FastAPI()

@web_app.get("/")
async def root():
    return {"status": "ok"}

@web_app.post("/predict")
async def predict(request: dict):
    return {"result": model.run(request["input"])}

@app.function(image=image, gpu="L40S")
@modal.asgi_app()
def fastapi_app():
    return web_app

With Class Lifecycle

@app.cls(gpu="L40S", image=image)
class InferenceService:
    @modal.enter()
    def load_model(self):
        self.model = load_model()

    @modal.asgi_app()
    def serve(self):
        from fastapi import FastAPI
        app = FastAPI()

        @app.post("/generate")
        async def generate(request: dict):
            return self.model.generate(request["prompt"])

        return app

WSGI Apps (Flask, Django)

from flask import Flask

flask_app = Flask(__name__)

@flask_app.route("/")
def index():
    return {"status": "ok"}

flask_image = modal.Image.debian_slim().uv_pip_install("flask")

@app.function(image=flask_image)
@modal.wsgi_app()
def flask_server():
    return flask_app

WSGI is synchronous — concurrent inputs run on separate threads.

Custom Web Servers

For non-standard web frameworks (aiohttp, Tornado, TGI):

server_image = modal.Image.debian_slim(python_version="3.11").uv_pip_install("vllm==0.21.0")

@app.function(image=server_image, gpu="H100")
@modal.web_server(port=8000)
def serve():
    import subprocess
    subprocess.Popen([
        "vllm", "serve", "Qwen/Qwen3-8B",
        "--max-model-len", "4096",
        "--host", "0.0.0.0",  # Must bind to 0.0.0.0, not localhost
        "--port", "8000",
    ])

The application must bind to 0.0.0.0 (not 127.0.0.1).

Security: The command above uses a fixed argument list. Do not interpolate unsanitized user input (model names, paths, flags) into subprocess arguments — validate against an allowlist or pass untrusted values as data, not as command arguments, to avoid command injection.

WebSockets

Modal's web proxy supports WebSockets. Use an ASGI or custom server framework that implements WebSockets; plain WSGI/Flask handlers do not implement that protocol:

from fastapi import FastAPI, WebSocket

web_app = FastAPI()

@web_app.websocket("/ws")
async def websocket_endpoint(websocket: WebSocket):
    await websocket.accept()
    while True:
        data = await websocket.receive_text()
        result = process(data)
        await websocket.send_text(result)

@app.function()
@modal.asgi_app()
def ws_app():
    return web_app
  • Full WebSocket protocol (RFC 6455)
  • Messages up to 2 MiB each
  • No RFC 8441 or RFC 7692 support yet

Authentication

Proxy Auth Tokens (Built-in)

Modal provides first-class endpoint protection via proxy auth tokens:

@app.function()
@modal.fastapi_endpoint(requires_proxy_auth=True)
def protected(text: str):
    return {"result": text}

Create a Proxy Token (distinct from the Modal SDK token) in the dashboard or modal workspace proxy-tokens CLI. Clients send its ID/secret as Modal-Key and Modal-Secret, or Authorization: Bearer <proxy-id>.<proxy-secret>. Missing or invalid credentials receive 401 before the handler runs. Merely sending these headers does not protect an endpoint whose decorator omits requires_proxy_auth=True.

Custom Bearer Tokens

from fastapi import Header, HTTPException

@app.function(secrets=[modal.Secret.from_name("auth-secret")])
@modal.fastapi_endpoint(method="POST")
def secure_predict(data: dict, authorization: str = Header(None)):
    import os
    import hmac
    expected = os.environ["AUTH_TOKEN"]
    if not authorization or not hmac.compare_digest(authorization, f"Bearer {expected}"):
        raise HTTPException(status_code=401, detail="Unauthorized")
    return {"result": model.predict(data["text"])}

Client IP Access

Use a FastAPI Request and inspect request.client.host (if request.client is not None). Do not trust an arbitrary user-supplied header as proof of client identity.

Streaming

Server-Sent Events (SSE)

from fastapi.responses import StreamingResponse

@app.function(gpu="H100")
@modal.fastapi_endpoint()
def stream_generate(prompt: str):
    def generate():
        for token in model.stream(prompt):
            yield f"data: {token}\n\n"
    return StreamingResponse(generate(), media_type="text/event-stream")

Concurrency

Handle multiple requests per container using @modal.concurrent:

@app.function(gpu="L40S")
@modal.concurrent(max_inputs=10)
@modal.fastapi_endpoint(method="POST")
async def batch_predict(data: dict):
    return {"result": await model.predict_async(data["text"])}

Limits

  • Request body: up to 4 GiB
  • Response body: unlimited
  • Rate limit: 200 requests/second (5-second burst for new accounts)
  • Rate limit is workspace-wide and excess HTTP requests return 429; workspace limits can differ
  • Cold starts occur when no container is ready (min_containers reduces their frequency)
  • After 150 seconds a Web Function HTTP request returns a 303 result URL; clients must follow redirects. This is separate from the underlying Function timeout. CORS-dependent clients cannot rely on that redirect mechanism. Prefer .spawn() plus a call-ID polling endpoint for long jobs, and handle seven-day output expiry.

The newer @app.server() primitive serves latency-sensitive HTTP applications and requires Proxy Tokens by default; its contract is distinct from these Web Functions. See the official Servers guide before migrating.