dagloom 0.4.0

A lightweight pipeline/workflow engine. Weave data processing nodes into DAG workflows with decorators and the >> operator.

🧶 Dagloom

Like a loom weaving threads into fabric, Dagloom weaves data processing nodes into DAG workflows.

A lightweight pipeline/workflow engine for Python. Define nodes with decorators, connect them with the >> operator, visualize and edit in a drag-and-drop Web UI.

中文文档

---

✨ Why Dagloom?

Problem	Competitors	Dagloom
Overkill installation	Airflow needs PostgreSQL + Redis + Celery + Webserver	pip install dagloom && dagloom serve
Too many concepts	Dagster: Assets, Ops, Jobs, Resources, IO Managers...	Just @node and >>
Code/visual disconnect	Airflow UI is read-only	True bidirectional sync
Can't resume from failure	Re-run the entire pipeline	dagloom resume picks up where it left off
Shell-only nodes	Dagu only supports shell commands	Native Python objects (DataFrames, dicts, classes)

🚀 Quick Start

Installation

pip install dagloom

Your First Pipeline

from dagloom import node, Pipeline

@node
def greet(name: str) -> str:
    """Create a greeting message."""
    return f"Hello, {name}!"

@node
def shout(message: str) -> str:
    """Convert message to uppercase."""
    return message.upper()

@node
def add_emoji(message: str) -> str:
    """Add emoji to the message."""
    return f"🎉 {message} 🎉"

# Build DAG with >> operator
pipeline = greet >> shout >> add_emoji

# Run the pipeline
result = pipeline.run(name="World")
print(result)  # 🎉 HELLO, WORLD! 🎉

Conditional Branching

Use the | operator to create mutually exclusive branches — the runtime selects which branch to execute based on the upstream output:

from dagloom import node

@node
def classify(text: str) -> dict:
    """Route to different processors."""
    if "urgent" in text:
        return {"branch": "urgent_handler", "text": text}
    return {"branch": "normal_handler", "text": text}

@node
def urgent_handler(data: dict) -> str:
    return f"🚨 URGENT: {data['text']}"

@node
def normal_handler(data: dict) -> str:
    return f"📋 Normal: {data['text']}"

pipeline = classify >> (urgent_handler | normal_handler)
result = pipeline.run(text="urgent: server down!")
# 🚨 URGENT: urgent: server down!

Streaming Nodes (Generator)

Node functions can be generators — yielded values are automatically collected into a list:

@node
def stream_data(url: str):
    """Yield data chunks."""
    for i in range(5):
        yield {"chunk": i, "url": url}

@node
def aggregate(chunks: list[dict]) -> int:
    return len(chunks)

pipeline = stream_data >> aggregate
result = pipeline.run(url="https://example.com")
# 5

Execution Hooks

Monitor node execution with on_node_start / on_node_end callbacks:

import asyncio
from dagloom import node, AsyncExecutor

@node
def step(x: int) -> int:
    return x + 1

pipeline = step

def my_hook(node_name, ctx):
    print(f"  → {node_name}: {ctx.get_node_info(node_name).status}")

executor = AsyncExecutor(
    pipeline,
    on_node_start=my_hook,
    on_node_end=my_hook,
)
result = asyncio.run(executor.execute(x=1))

Pipeline Scheduling

Schedule pipelines to run automatically on cron expressions or fixed intervals:

from dagloom import node, Pipeline

@node
def fetch(url: str = "https://example.com/data.csv") -> list:
    return [1, 2, 3]

@node
def process(data: list) -> int:
    return sum(data)

# Set schedule via Pipeline constructor
pipeline = Pipeline(name="daily_etl", schedule="0 9 * * *")

# Or use interval shorthand
pipeline = Pipeline(name="frequent_check", schedule="every 30m")

# Or set after construction
pipeline = fetch >> process
pipeline.name = "my_pipeline"
pipeline.schedule = "0 9 * * 1-5"  # Weekdays at 9am

The scheduler runs in-process with dagloom serve — schedules are persisted to SQLite and auto-restored on restart.

Advanced Features

@node(retry=3, cache=True, timeout=30.0)
def fetch_data(url: str) -> pd.DataFrame:
    """Fetch CSV data with retry and caching."""
    return pd.read_csv(url)

@node(cache=True)
def clean(df: pd.DataFrame) -> pd.DataFrame:
    """Remove rows with missing values."""
    return df.dropna()

@node
def save(df: pd.DataFrame) -> str:
    """Persist cleaned data to parquet file."""
    path = "output/cleaned.parquet"
    df.to_parquet(path)
    return path

pipeline = fetch_data >> clean >> save
pipeline.run(url="https://example.com/data.csv")

Start the Web UI

dagloom serve
# Open http://localhost:8000 in your browser

🏗️ Architecture

Single Process Architecture
┌─────────────────────────────────────┐
│  CLI / Web UI                       │
├─────────────────────────────────────┤
│  FastAPI (REST API + WebSocket)     │
├─────────────────────────────────────┤
│  Scheduler (APScheduler + asyncio)  │
├─────────────────────────────────────┤
│  Core (@node + Pipeline + DAG)      │
├─────────────────────────────────────┤
│  SQLite (embedded, zero config)     │
└─────────────────────────────────────┘

📦 Project Structure

dagloom/
├── core/       # @node decorator, Pipeline class, DAG validation
├── scheduler/  # Cron/interval scheduler, asyncio executor, caching, checkpoint
├── connectors/ # PostgreSQL, MySQL, S3, HTTP connectors
├── server/     # FastAPI REST API + WebSocket
├── store/      # SQLite storage layer
└── cli/        # Click CLI (serve, run, list, inspect, scheduler)

📖 Documentation

• 中文文档
• Architecture Guide | 架构文档
• Getting Started (EN) | 快速入门 (中文)

🤝 Contributing

Contributions are welcome! Please feel free to:

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add some amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

Apache License 2.0 — see LICENSE for details.