stacksense 0.1.0

AI infrastructure monitoring for tracking usage, cost, and performance across OpenAI, Claude, ElevenLabs, Pinecone, and more. Supports SQLite (default) and PostgreSQL databases.

AI cost monitoring for Python. Two lines of code. Thirteen providers.

     

Quickstart  •   Providers  •   Features  •   Middleware  •   Config  •   Contributing

The Problem

You're shipping AI features. Costs are invisible until the invoice hits.

Month 1    $12        "No big deal."
Month 3    $480       "Wait, what?"
Month 5    $2,100     "Which call is doing this??"

StackSense wraps your existing AI clients and tracks every call — tokens, latency, cost — with zero config and zero code changes to your business logic.

Quickstart

pip install stacksense

from stacksense import StackSense
import openai

ss = StackSense()
client = ss.monitor(openai.OpenAI())

# Use client exactly as before — every call is now tracked
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}]
)

print(ss.get_metrics())
# {'total_calls': 1, 'total_tokens': 28, 'total_cost': 0.0004, ...}

That's it. No dashboards to configure. No agents to deploy. Just monitor() and go.

Supported Providers

Pass any supported client to ss.monitor() — the provider is auto-detected.

OpenAI GPT-4o • o1 • o3 • Embeddings	Anthropic Opus 4 • Sonnet 4 • Haiku	Google Gemini 2.0 Flash • 1.5 Pro	Mistral Large • Small • Codestral
Cohere Command R/R+ • Embed v4	DeepSeek Chat • Reasoner	AI21 Labs Jamba 1.5 Large/Mini	Together AI Llama 3.1 • Mixtral
Groq Llama 3.3 • Mixtral • Gemma2	Perplexity Sonar Pro • Reasoning	Replicate Llama • SDXL • any model	ElevenLabs Voice models • per-character
Pinecone Vector ops • per-query	More coming soon — request a provider

Features

Multi-Provider Cost Breakdown

Track spend across providers from a single StackSense instance.

ss = StackSense()

oai = ss.monitor(openai.OpenAI())
claude = ss.monitor(anthropic.Anthropic())

oai.chat.completions.create(model="gpt-4o", messages=[...])
claude.messages.create(model="claude-sonnet-4-20250514", messages=[...])

ss.get_cost_breakdown()
# {'openai': 0.003, 'anthropic': 0.002}

Decorator API

Track any function without wrapping a client:

import stacksense

@stacksense.track(provider="openai", model="gpt-4o")
def generate(prompt):
    return client.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": prompt}]
    )

Works with async functions too.

Alerts & Webhooks

Get notified when costs spike.

from stacksense.alerts import AlertManager, AlertRule

alerts = AlertManager(tracker=ss.tracker)
alerts.add_rule(AlertRule(
    name="Cost spike",
    metric="cost",
    threshold=5.0,
    window="1h",
))
alerts.add_webhook("https://hooks.slack.com/services/...")
alerts.check()

Export

from stacksense.exporters import Exporter

exporter = Exporter(ss.tracker)
exporter.to_csv("metrics.csv")
exporter.to_json("metrics.json")

Framework Middleware

Drop-in middleware for popular frameworks — automatically tracks all AI calls per request.

FastAPI from stacksense.middleware import ( FastAPIMiddleware ) app.add_middleware( FastAPIMiddleware, stacksense=ss )

Flask from stacksense.middleware import ( FlaskMiddleware ) FlaskMiddleware(app, stacksense=ss)

Django # settings.py MIDDLEWARE = [ ..., 'stacksense.middleware' '.DjangoMiddleware', ]

CLI

stacksense status              # View current metrics
stacksense dashboard           # Launch web dashboard
stacksense export csv -o out.csv
stacksense db init             # Initialize database

Configuration

SQLite by default — zero config. PostgreSQL for production:

pip install stacksense[postgresql]

# Environment variables
STACKSENSE_PROJECT_ID=my-project
STACKSENSE_ENABLE_DB=true
STACKSENSE_DB_URL=postgresql://user:pass@host:5432/stacksense
STACKSENSE_ENVIRONMENT=production
STACKSENSE_DEBUG=false

Contributing

git clone https://github.com/Abdulkvng/stacksense.git
cd stacksense
pip install -e ".[dev]"
pytest tests/ -v

PRs welcome. Please open an issue first for large changes.

_{MIT License © 2025 StackSense Contributors}