django-nocturne 0.1.2

Intelligent observability for Django. Detect anomalies, diagnose with AI, alert via webhooks — plug in, migrate, done.

django-nocturne

Intelligent observability for Django. Detect anomalies, diagnose with AI, alert via webhooks — plug in, migrate, done.

---

What is django-nocturne?

django-nocturne is a plug-and-play APM (Application Performance Monitoring) library for Django. It silently observes every HTTP request, statistically detects error spikes, asks an AI to diagnose them, fires webhook alerts, and renders a real-time dashboard — all with zero instrumentation changes to your application code.

Features

• Zero-code observability — one middleware entry records every request automatically
• Statistical anomaly detection — Z-score analysis across 5-minute buckets; three severity tiers (MEDIUM / HIGH / CRITICAL)
• Multi-signal health scoring — composite 0–100 score from error rate, response time, and volume signals
• AI-powered root-cause analysis — LangChain interface supports Anthropic Claude, OpenAI GPT, Google Gemini, and local Ollama models
• Webhook alerting — HMAC-SHA256 signed payloads delivered to any HTTP endpoint when thresholds are breached
• Health trend tracking — per-service snapshots over time; dashboard shows improving/degrading/stable trends
• Real-time dashboard — Chart.js visualisations, log explorer with syntax-highlighted stacktraces, anomaly detail modals
• Timeframe filter — global 15M / 30M / 1H / 3H / 6H / 12H / 24H / 7D filter controlling all charts simultaneously
• REST API — full DRF API for health, logs, anomalies, webhooks, and dashboard data

---

Screenshots

Dashboard Overview

Real-time system overview with total logs, error rate, and active anomalies

Error Rate & Service Health

Error rate over time per service + health scores (0–100)

Anomaly Detection

Detected anomalies with severity badges and Z-scores

Anomaly Detail with AI Diagnosis

Z-score meter, plain English explanation, and LLM root cause analysis

Log Explorer

Paginated log table with colored level badges and full-text search

Stacktrace Viewer

Expanded log row showing exception badge, syntax-highlighted stacktrace, and AI analysis button

AI Root-Cause Analysis

On-demand LLM diagnosis with root cause, fix, and prevention steps

Webhook Activity

Live delivery feed showing status, payload, and response for every webhook event

Health Trends

Per-service trend cards with ↑/↓/→ indicators and 0–100 composite scores

Timeframe Filter

Global 15M–7D pill filter that updates all charts simultaneously

Django Admin — Log Entries

Django admin list view for LogEntry with filtering and search

Django Admin — Anomaly Events

Django admin list view for AnomalyEvent with severity and resolution status

REST API (DRF Browsable)

DRF browsable API health endpoint response

---

Requirements

• Python 3.9+
• Django 4.2 or 5.1+
• Django REST Framework 3.14+
• numpy

---

Quick Start

1. Install

# Core (no AI)
pip install django-nocturne

# With your preferred AI backend
pip install "django-nocturne[anthropic]"   # Claude
pip install "django-nocturne[openai]"      # GPT-4o
pip install "django-nocturne[ollama]"      # Local Llama via Ollama
pip install "django-nocturne[gemini]"      # Google Gemini

2. Configure (settings.py)

INSTALLED_APPS = [
    # ... your apps ...
    "nocturne",
]

MIDDLEWARE = [
    # ... other middleware ...
    "nocturne.middleware.NocturneMiddleware",   # add last
]

NOCTURNE = {
    "SERVICE_NAME": "my-api",
    "ANOMALY_THRESHOLD": 2.0,              # Z-score cutoff
    "AI_BACKEND": "anthropic",             # 'anthropic' | 'openai' | 'ollama' | 'gemini'
    "AI_DIAGNOSIS_ENABLED": True,
    "ANTHROPIC_API_KEY": env("ANTHROPIC_API_KEY"),
    "ANTHROPIC_MODEL": "claude-sonnet-4-6",
}

3. Wire up URLs (urls.py)

from django.urls import include, path

urlpatterns = [
    path("nocturne/", include("nocturne.urls")),
    # ...
]

4. Migrate and run

python manage.py migrate
python manage.py runserver

Open http://localhost:8000/nocturne/dashboard/ — log in as a superuser.

---

Configuration Reference

All settings live inside the NOCTURNE dict in settings.py.

Key	Default	Description
SERVICE_NAME	"default"	Tag applied to all log entries by the middleware
ANOMALY_THRESHOLD	2.0	Z-score threshold to trigger anomaly detection
RETENTION_DAYS	30	Auto-purge log entries older than N days
EXCLUDE_PATHS	["/health", "/static", "/favicon.ico"]	Paths skipped by middleware
LOGIN_URL	"/admin/login/"	Redirect for unauthenticated dashboard access
AI_BACKEND	"ollama"	Active LLM backend: anthropic / openai / ollama / gemini
AI_DIAGNOSIS_ENABLED	True	Master toggle for all AI calls
ANTHROPIC_API_KEY	""	Claude API key
ANTHROPIC_MODEL	"claude-sonnet-4-6"	Claude model ID
OLLAMA_BASE_URL	"http://localhost:11434"	Ollama server URL
OLLAMA_MODEL	"llama3.2"	Ollama model name
OPENAI_API_KEY	""	OpenAI API key
OPENAI_MODEL	"gpt-4o"	OpenAI model ID
OPENAI_BASE_URL	"https://api.openai.com/v1"	Override for Azure / Groq / vLLM
GEMINI_API_KEY	""	Google Gemini API key
GEMINI_MODEL	"gemini-1.5-flash"	Gemini model ID
WEBHOOK_SECRET	""	HMAC-SHA256 secret for webhook signature validation

---

API Reference

All endpoints are prefixed with /nocturne/ (or wherever you mounted the URLs).

Method	Path	Auth	Description
GET	api/health/	view	Health scores, error rate, anomaly counts
GET	api/logs/	view	Paginated log entries (no stacktrace)
POST	api/logs/ingest/	view	Ingest a log entry externally
GET	api/logs/<id>/	view	Full log entry including stacktrace
POST	api/logs/<id>/analyse/	view	AI root-cause analysis for a log
GET	api/anomalies/	view	List anomaly events
PATCH	api/anomalies/<id>/	admin	Resolve an anomaly
POST	api/detect/	admin	Trigger anomaly detection scan
GET	api/dashboard/data/	view	All chart data in one response
GET	api/webhooks/	admin	List webhook configurations
POST	api/webhooks/	admin	Create webhook configuration
PUT	api/webhooks/<id>/	admin	Update webhook configuration
DELETE	api/webhooks/<id>/	admin	Delete webhook configuration
GET	api/webhooks/events/	view	Webhook delivery history
POST	api/webhooks/test/	admin	Send test webhook to all active configs
POST	api/webhook/receive/	admin	Simulated receiver endpoint
GET	dashboard/	view	Browser dashboard (HTML)

Auth tiers:

• view — superuser or user with nocturne.view_nocturne permission
• admin — superuser only

---

Webhook Payload

When an anomaly is detected, Nocturne fires this payload to every active webhook config:

{
  "event": "anomaly.detected",
  "timestamp": "2026-06-19T03:22:27Z",
  "watchdog_version": "0.1.0",
  "anomaly": {
    "id": 42,
    "service": "auth-service",
    "severity": "CRITICAL",
    "z_score": 4.7,
    "error_count": 50,
    "window_start": "...",
    "window_end": "...",
    "ai_diagnosis": "..."
  },
  "health": {
    "score_before": 85.0,
    "score_after": 20.0,
    "trend": "degrading"
  },
  "action_required": true,
  "dashboard_url": "/nocturne/dashboard/"
}

Headers sent:

• Content-Type: application/json
• X-Watchdog-Event: anomaly.detected
• X-Watchdog-Severity: CRITICAL
• X-Watchdog-Signature: sha256=<hmac> (when WEBHOOK_SECRET is set)

---

Management Commands

Commands included with the package (available after pip install django-nocturne):

Command	Description
nocturne_config	Print resolved NOCTURNE settings + DB stats
test_ai_diagnosis	Verify AI backend is working
create_nocturne_user	Create a viewer user with correct permissions
snapshot_health	Take a health snapshot for all active services

python manage.py nocturne_config
python manage.py test_ai_diagnosis
python manage.py test_ai_diagnosis --backend openai
python manage.py create_nocturne_user
python manage.py create_nocturne_user --username viewer1 --password secret
python manage.py snapshot_health

Commands available in the example project only:

Command	Description
generate_demo_logs	Seed 975 demo logs with 3 anomaly spikes, WebhookEvents, and HealthSnapshots

python manage.py generate_demo_logs

---

Running the Example Project

git clone https://github.com/rishav00a/django-nocturne
cd django-nocturne
python -m venv .venv && source .venv/bin/activate
pip install -e ".[anthropic]"   # or [ollama], [openai], [gemini]
cd example_project
python manage.py migrate
python manage.py createsuperuser
python manage.py generate_demo_logs
python manage.py runserver

Visit http://localhost:8000/nocturne/dashboard/ and log in.

---

Running Tests

pip install -e ".[dev]"
pytest tests/ -v

---

Contributing

See CONTRIBUTING.md.

---

Changelog

See CHANGELOG.md.

---

License

MIT — see LICENSE.