trellis-pipelines 0.1.1

Modular agentic pipeline system for document and data workflows

Trellis

An agentic solution for pipeline planning and execution within controlled environments.

Quick Start

Installation

pip install -e .

Usage

CLI

# Validate a pipeline
trellis validate pipelines/example.yaml

# Run a pipeline
trellis run pipelines/example.yaml

# Generate a plan from a goal
trellis plan "Extract and analyze market data"

# Generate a pipeline from a goal
trellis generate "Extract and analyze market data" -o output.yaml

API

# Start the API server
python -m trellis_api.main

# Create a pipeline
curl -X POST http://localhost:8000/pipelines \
  -H "Content-Type: application/json" \
  -d '{
    "id": "example_pipeline",
    "goal": "Process data",
    "tasks": []
  }'

Python Library

from trellis.models.pipeline import Pipeline
from trellis.execution.dag import DAGExecutor

pipeline = Pipeline(
    id="my_pipeline",
    goal="Extract and process data",
    tasks=[]
)

executor = DAGExecutor()
results = executor.execute({})

Architecture

See docs/architecture.md for detailed architecture documentation.

Project Structure

trellis/
├── trellis/                    # Core library (shared)
│   ├── models/             # Pydantic DSL models
│   ├── validation/         # DAG and contract validation
│   ├── execution/          # Execution engine
│   └── tools/              # Tool protocol and implementations
├── trelis_api/             # FastAPI server
├── trelis_mcp/             # MCP server
├── trelis_cli/             # CLI interface
├── data/                   # Dataset generation
├── tests/                  # Test suite
├── docs/                   # Documentation
└── scripts/                # Utility scripts

Development

Testing

pytest tests/ -v

Linting

black .
isort .
mypy trellis/

Building

python -m build

License

See LICENSE file for details.