Appearance

Django — full setup ​

Complete Django integration: middleware, OpenAI in views, environment config, shutdown.

Install ​

bash
pip install amlexia django openai

Environment ​

env
AMLEXIA_SDK_KEY=am_...
AMLEXIA_INGEST_URL=https://ingest.amlexia.com
AMLEXIA_ENVIRONMENT=production
AMLEXIA_SERVICE_NAME=api
DJANGO_SECRET_KEY=...
OPENAI_API_KEY=sk-...

Settings (myproject/settings.py) ​

python
MIDDLEWARE = [
    "django.middleware.security.SecurityMiddleware",
    "amlexia.django_integration.AmlexiaMiddleware",
    "django.middleware.common.CommonMiddleware",
    # ... rest of your middleware
]

AmlexiaMiddleware uses AmlexiaClient.from_env() and AMLEXIA_SERVICE_NAME (default api).

Optional explicit client ​

python
from amlexia import AmlexiaClient

AMLEXIA_CLIENT = AmlexiaClient.from_env()

MIDDLEWARE = [
    "django.middleware.security.SecurityMiddleware",
    # Custom wrapper if you inject client — otherwise default middleware is enough
    "amlexia.django_integration.AmlexiaMiddleware",
]

Views (myapp/views.py) ​

python
import time

from django.http import JsonResponse
from django.views.decorators.csrf import csrf_exempt
from django.views.decorators.http import require_http_methods
from openai import OpenAI

from amlexia import AmlexiaClient
from amlexia.openai_helper import track_openai_completion

client = AmlexiaClient.from_env()
openai = OpenAI()


@require_http_methods(["GET"])
def health(request):
    return JsonResponse({"ok": True})


@require_http_methods(["GET"])
def get_user(request, user_id):
    return JsonResponse({"id": user_id})


@csrf_exempt
@require_http_methods(["POST"])
def stripe_webhook(request):
    return JsonResponse({"received": True})


@csrf_exempt
@require_http_methods(["POST"])
def chat(request):
    import json

    body = json.loads(request.body)
    start = time.time()
    try:
        completion = openai.chat.completions.create(
            model="gpt-4o-mini",
            messages=body["messages"],
        )
        track_openai_completion(
            client,
            model=completion.model,
            status_code=200,
            latency_ms=int((time.time() - start) * 1000),
            usage=completion.usage,
        )
        return JsonResponse(completion.model_dump())
    except Exception as e:
        client.track(
            endpoint="POST /api/chat",
            method="POST",
            status_code=500,
            latency_ms=int((time.time() - start) * 1000),
            provider="openai",
            error_message=str(e)[:500],
        )
        return JsonResponse({"error": str(e)}, status=500)

URLs (myapp/urls.py) ​

python
from django.urls import path
from . import views

urlpatterns = [
    path("health/", views.health),
    path("api/users/<int:user_id>/", views.get_user),
    path("webhooks/stripe/", views.stripe_webhook),
    path("api/chat/", views.chat),
]

Middleware behavior ​

FeatureDetail
Tracingtraceparent on response
PathsNormalized (/users/42/ → /users/:id/)
SessionHTTP_X_SESSION_ID, HTTP_X_USER_ID
ProviderAuto-detect from path hints

Place after SecurityMiddleware, before view runs.

Shutdown (ASGI / management commands) ​

For long-running runserver or gunicorn, register shutdown on worker exit:

python
# myproject/apps.py or wsgi.py
import atexit
from amlexia import AmlexiaClient

_client = AmlexiaClient.from_env()
atexit.register(_client.shutdown)

Or call client.shutdown() in gunicorn worker_exit hook.

Verify ​

bash
amlexia health
python manage.py runserver
curl http://127.0.0.1:8000/health/