deadpipe 0.1.3

Dead simple pipeline monitoring. Know when your pipelines die.

Deadpipe Python SDK

Dead simple pipeline monitoring. Know when your pipelines die.

Installation

pip install deadpipe

Quick Start

Option 1: Decorator (Recommended)

from deadpipe import Deadpipe

dp = Deadpipe("your-api-key")

@dp.heartbeat("daily-sales-etl")
def run_pipeline():
    # Your pipeline code here
    process_data()
    return {"records_processed": 1500}  # Optional: track records

# That's it! Deadpipe will ping on success or failure.
run_pipeline()

Option 2: Context Manager

from deadpipe import Deadpipe

dp = Deadpipe("your-api-key")

with dp.pipeline("hourly-sync"):
    # Your code here
    sync_data()

Option 3: Manual Ping

from deadpipe import Deadpipe

dp = Deadpipe("your-api-key")

try:
    run_my_job()
    dp.ping("my-job", status="success", records_processed=1000)
except Exception as e:
    dp.ping("my-job", status="failed")
    raise

Option 4: Environment Variable

Set DEADPIPE_API_KEY and use module-level functions:

import deadpipe

@deadpipe.heartbeat("my-pipeline")
def my_job():
    pass

Async Support (FastAPI, aiohttp, asyncio)

For async applications, install with async extras:

pip install deadpipe[async]

Async Decorator

from deadpipe import AsyncDeadpipe

dp = AsyncDeadpipe("your-api-key")

@dp.heartbeat("async-pipeline")
async def my_async_job():
    await fetch_data()
    return {"records_processed": 500}

Async Context Manager

async with dp.pipeline("async-etl"):
    await process_async_data()

Async Manual Ping

await dp.ping("my-job", status="success")

FastAPI Example

from fastapi import FastAPI
from deadpipe import AsyncDeadpipe

app = FastAPI()
dp = AsyncDeadpipe()  # Uses DEADPIPE_API_KEY env var

@app.on_event("startup")
async def startup():
    # Optionally track app startup
    await dp.ping("fastapi-app", status="success")

@app.on_event("shutdown") 
async def shutdown():
    await dp.close()  # Clean up HTTP session

@app.post("/process")
@dp.heartbeat("data-processor")
async def process_data():
    result = await heavy_processing()
    return {"records_processed": len(result)}

Session Management

AsyncDeadpipe uses connection pooling via aiohttp.ClientSession. For long-running apps, use as a context manager or call close():

# Option 1: Context manager (auto-closes)
async with AsyncDeadpipe("your-key") as dp:
    await dp.ping("my-pipeline")

# Option 2: Manual close
dp = AsyncDeadpipe("your-key")
try:
    await dp.ping("my-pipeline")
finally:
    await dp.close()

Airflow Integration

from deadpipe import Deadpipe

dp = Deadpipe(api_key=Variable.get("DEADPIPE_API_KEY"))

@dp.heartbeat("{{ dag.dag_id }}")
def my_task():
    ...

Or add to the end of any task:

from deadpipe import ping

def my_task():
    # ... your code ...
    ping("daily-etl", status="success")

dbt Integration

Add to your dbt_project.yml on-run-end hook:

on-run-end:
  - "{{ deadpipe_heartbeat('dbt-run') }}"

Or call from Python:

# In your dbt runner script
from deadpipe import Deadpipe

dp = Deadpipe("your-api-key")

with dp.pipeline("dbt-daily"):
    subprocess.run(["dbt", "run"], check=True)

API Reference

Deadpipe(api_key, base_url, timeout)

Create a client instance.

• api_key: Your API key (or set DEADPIPE_API_KEY env var)
• base_url: Override for self-hosted (default: https://www.deadpipe.com/api/v1)
• timeout: Request timeout in seconds (default: 10)

dp.ping(pipeline_id, status, duration_ms, records_processed, app_name)

Send a heartbeat.

• pipeline_id: Unique identifier for this pipeline
• status: "success" or "failed"
• duration_ms: How long the run took (optional)
• records_processed: Number of records (optional)
• app_name: Group pipelines under an app (optional)

@dp.heartbeat(pipeline_id, app_name, on_error)

Decorator that auto-sends heartbeats.

• on_error: What to do on exception:
  • "ping" (default): Send failed heartbeat, then re-raise
  • "raise": Re-raise without heartbeat
  • "ignore": Send success heartbeat anyway

with dp.pipeline(pipeline_id, app_name)

Context manager for heartbeats.

AsyncDeadpipe(api_key, base_url, timeout)

Async client with the same API as Deadpipe, but all methods are async:

• await dp.ping(...) - Async heartbeat
• @dp.heartbeat(...) - Decorator for async functions
• async with dp.pipeline(...) - Async context manager
• await dp.run(pipeline_id, fn, *args) - Run async function with heartbeat
• await dp.close() - Close HTTP session
• async with AsyncDeadpipe(...) as dp: - Auto-close on exit

Dependencies

• Sync: Zero dependencies (uses Python standard library only)
• Async: Requires aiohttp (install with pip install deadpipe[async])

License

Deadpipe SDK License - see LICENSE file.