synapse-step-workflow

# Step-based Workflow System

Synapse SDK provides a step-based workflow system for complex actions that need:
- Multi-phase execution with progress tracking
- Automatic rollback on failure
- State sharing between steps
- Conditional step execution

## Core Components

| Component | Purpose |
|-----------|---------|
| `BaseStep` | Abstract step definition |
| `StepResult` | Step execution result |
| `StepRegistry` | Ordered step registration |
| `Orchestrator` | Step execution with rollback |
| `BaseStepContext` | State sharing between steps |

## Quick Start

```python
from dataclasses import dataclass, field
from synapse_sdk.plugins.steps import (
    BaseStep,
    StepResult,
    StepRegistry,
    Orchestrator,
    BaseStepContext,
)
from synapse_sdk.plugins.context import RuntimeContext

# 1. Define context for state sharing
@dataclass
class ProcessContext(BaseStepContext):
    data: list = field(default_factory=list)
    processed: int = 0

# 2. Define steps
class LoadStep(BaseStep[ProcessContext]):
    @property
    def name(self) -> str:
        return 'load'

    @property
    def progress_weight(self) -> float:
        return 0.3

    def execute(self, ctx: ProcessContext) -> StepResult:
        ctx.data = load_data()
        return StepResult(success=True, data={'count': len(ctx.data)})

class ProcessStep(BaseStep[ProcessContext]):
    @property
    def name(self) -> str:
        return 'process'

    @property
    def progress_weight(self) -> float:
        return 0.7

    def execute(self, ctx: ProcessContext) -> StepResult:
        for item in ctx.data:
            process(item)
            ctx.processed += 1
            ctx.set_progress(ctx.processed, len(ctx.data))
        return StepResult(success=True)

# 3. Register and run
registry = StepRegistry[ProcessContext]()
registry.register(LoadStep())
registry.register(ProcessStep())

context = ProcessContext(runtime_ctx=runtime_ctx)
orchestrator = Orchestrator(registry, context)
result = orchestrator.exe