Skip to main content

luml.integrations.langgraph.packaging

save_langgraph​

def save_langgraph(
        graph: Pregel | Callable[[], Pregel] | str,
        path: str | None = None,
        dependencies: Literal["default"] | Literal["all"]
    | list[str] = "default",
        extra_dependencies: list[str] | None = None,
        extra_code_modules: list[str] | Literal["auto"] | None = "auto",
        env_vars: list[str] | None = None,
        manifest_model_name: str | None = None,
        manifest_model_version: str | None = None,
        manifest_model_description: str | None = None,
        manifest_extra_producer_tags: list[str] | None = None
) -> ModelReference

Save LangGraph application to LUML format for deployment.

Packages a LangGraph workflow with its dependencies, environment variables, and metadata for production deployment or model registry.

Arguments:

  • graph - LangGraph Pregel instance, callable returning Pregel, or import path string.
  • path - Output file path. Auto-generated if not provided.
  • dependencies - Dependency management strategy:
    • "default": Auto-detect dependencies
    • "all": Include all detected dependencies
    • list: Custom dependency list
  • extra_dependencies - Additional pip packages to include.
  • extra_code_modules - Local code modules to bundle.
    • "auto": Auto-detect local dependencies (default)
    • list: Specific modules to include
    • None: Don't include local modules
  • env_vars - List of environment variable names to mark as runtime secrets.
  • manifest_model_name - Model name for metadata.
  • manifest_model_version - Model version for metadata.
  • manifest_model_description - Model description for metadata.
  • manifest_extra_producer_tags - Additional tags for model metadata.

Returns:

  • ModelReference - Reference to the saved model package with embedded Mermaid diagram.

Example:

from langgraph.graph import StateGraph
from luml.integrations.langgraph.packaging import save_langgraph

# Define your LangGraph
def create_graph():
    workflow = StateGraph(...)
    # ... add nodes and edges
    return workflow.compile()

graph = create_graph()

# Save graph
model_ref = save_langgraph(
    graph,
    path="my_agent.luml",
    env_vars=["OPENAI_API_KEY"],
    manifest_model_name="customer_support_agent",
    manifest_model_version="2.0.0"
)