create-state

Create LangGraph State Schema

Generate a TypedDict state schema with proper type hints, Annotated fields, and optional Pydantic validation.

Instructions for Claude

1. Gather Information

Ask user for:

• State class name (if not provided)
• State fields (if --fields not provided):
  • Field name
  • Type (str, int, float, List[...], dict, etc.)
  • Whether append-only (use Annotated[List, operator.add])
  • Whether custom reducer needed

2. Read Settings

Check .claude/langgraph-dev.local.md for:

• code_style (pydantic_v2 or pydantic_v1)
• include_type_hints (default: true)
• include_docstrings (default: true)

3. Generate State Schema

Create file {state_name.lower()}_state.py with:

from typing import TypedDict, List, Annotated, Optional
from langchain.schema import BaseMessage
import operator

class {StateName}(TypedDict):
    \"\"\"State schema for {state_name} workflow.\"\"\"
    # Add fields based on user input
    messages: Annotated[List[BaseMessage], operator.add]  # If append-only
    current_step: str
    # ... other fields

4. Add Validators (if Pydantic requested)

If user wants validation:

from pydantic import BaseModel, Field, field_validator

class {StateName}State(BaseModel):
    messages: List[BaseMessage] = Field(description="Conversation messages")
    confidence: float = Field(ge=0, le=1, description="Confidence score 0-1")

    @field_validator('confidence')
    def validate_confidence(cls, v):
        if not 0 <= v <= 1:
            raise ValueError('Confidence must be between 0 and 1')
        return v

5. Add Helper Functions

Include:

def create_initial_state(**kwargs) -> {StateName}:
    \"\"\"Factory for creating initial state with defaults.\"\"\"
    return {
        "messages": [],
        "current_step": "start",
        **kwargs
    }

def validate_state(state: {StateName}) -> bool:
    \"\"\"Validate state integrity.\"\"\"
    # Add validation logic
    return True