cloudmeter 0.5.0

Per-endpoint cloud cost attribution for Python web apps — see exactly what each API route costs on AWS, GCP, or Azure

cloudmeter (Python)

Per-endpoint cloud cost monitoring for Python web apps. Supports Flask, FastAPI, and Django — one import, one line.

[Your Flask/FastAPI/Django app]
        │
        │  in-process (zero overhead, no sidecar)
        ▼
[cloudmeter buffer]  →  http://localhost:7777
                     "GET /api/export costs $340/month"

---

Install

pip install cloudmeter

Zero required dependencies. No sidecar binary. No external process. The cost projector and dashboard run entirely in-process.

---

Quick start

Flask

from flask import Flask
from cloudmeter.flask import CloudMeterFlask

app = Flask(__name__)
CloudMeterFlask(app, provider="AWS", target_users=1000)

@app.route("/api/users/<int:user_id>")
def get_user(user_id):
    return {"id": user_id}

Open http://localhost:7777 — costs appear as soon as requests hit your app.

FastAPI

from fastapi import FastAPI
from cloudmeter.fastapi import CloudMeterMiddleware

app = FastAPI()
app.add_middleware(CloudMeterMiddleware, provider="AWS", target_users=1000)

@app.get("/api/users/{user_id}")
def get_user(user_id: int):
    return {"id": user_id}

Django — settings.py

MIDDLEWARE = [
    'cloudmeter.django.CloudMeterMiddleware',
    # ... your other middleware
]

CLOUDMETER = {
    "provider":     "AWS",
    "region":       "us-east-1",
    "target_users": 1000,
}

---

Configuration options

All options are optional — defaults work out of the box.

Option	Default	Description
provider	"AWS"	Cloud provider: "AWS", "GCP", or "AZURE"
region	"us-east-1"	Cloud region (e.g. "us-central1", "eastus")
target_users	1000	Concurrent users to project cost at
requests_per_user_per_second	1.0	Scaling factor for the projection
budget_usd	0.0	Monthly budget alert per endpoint (0 = off)
port	7777	Port for the live cost dashboard

Flask — app factory pattern

from cloudmeter.flask import CloudMeterFlask

cm = CloudMeterFlask()

def create_app():
    app = Flask(__name__)
    cm.init_app(app, provider="GCP", region="us-central1", target_users=5000)
    return app

FastAPI — full options example

app.add_middleware(
    CloudMeterMiddleware,
    provider="AWS",
    region="eu-west-1",
    target_users=50_000,
    budget_usd=200.0,
)

Django — full options in settings.py

CLOUDMETER = {
    "provider":                     "AZURE",
    "region":                       "eastus",
    "target_users":                 2000,
    "requests_per_user_per_second": 2.0,
    "budget_usd":                   500.0,
}

---

Dashboard panels

Open http://localhost:7777 after starting your app:

Panel	What it shows
Cost table	Each route with projected monthly cost, cost per user, recommended instance
Cost curve	Cost vs concurrent-user count chart (100 → 1M users) for any selected route
Variance panel	p50/p95/p99 per-request cost; p95/p50 > 1.5× flags N+1 queries or missing indexes
Recording controls	Start/stop metric collection; first 30 s is warmup (excluded from projections)

---

How route templates work

CloudMeter reports route templates, not raw paths — so all traffic to /api/users/1, /api/users/42, and /api/users/999 is grouped as a single endpoint with a single cost projection.

Framework	Template source	Example
Flask	request.url_rule	GET /api/users/<int:user_id>
FastAPI	scope["route"].path	GET /api/users/{user_id}
Django	request.resolver_match.route	GET api/users/<int:pk>/

---

What's measured

For each request:

• Route template — not the raw URL
• HTTP method — GET, POST, etc.
• HTTP status code — 200, 404, 500, etc.
• Wall-clock duration — in milliseconds
• Response size — egress bytes (from Content-Length header)

CPU accuracy note: Python cannot attribute per-request CPU time because of the GIL. Wall-clock time is used as a proxy. Cost accuracy is ±40% vs the Java agent's ±20%. Egress and instance selection are not affected.

---

Privacy and security

• Request and response bodies are never captured — only route, method, status, duration, and byte counts
• The dashboard binds to 127.0.0.1 only — not reachable from outside the host
• No authentication on the dashboard — do not expose port 7777 publicly
• No cloud credentials ever required — pricing uses static embedded tables

---

Running tests

pip install pytest pytest-asyncio httpx flask "fastapi[standard]" django
cd clients/python
pytest tests/ -v

86 tests covering Flask, FastAPI, Django middleware, the reporter buffer, cost projector, and dashboard server.

---

Troubleshooting

Dashboard shows no data: Make sure your app is receiving real HTTP requests (or use its test client). Endpoints only appear after they have been hit.

First 30 seconds shows "warmup excluded": Normal — the first 30 s after recording starts are excluded to avoid JIT/cold-start noise inflating cost projections. Data will appear automatically after warmup.

Port conflict (Address already in use): Change the dashboard port:

CloudMeterFlask(app, port=17777)

---

License

Apache 2.0 — see LICENSE at the repository root.