Implementation:Truera Trulens StateGraph Compile

Knowledge Sources	LangGraph Docs TruLens
Domains	Agent_Architecture, LLM_Applications
Last Updated	2026-02-14 08:00 GMT

Overview

External tool for constructing and compiling LangGraph stateful agent workflows, provided by the langgraph library. This is a Wrapper Doc — TruLens instruments but does not define this API.

Description

StateGraph is the primary builder class in LangGraph for defining stateful agent workflows. Users add nodes (processing functions), edges (execution order), and conditional routing logic, then call .compile() to produce a CompiledStateGraph (Pregel) object that can be invoked.

TruLens wraps the compiled result with TruGraph for tracing and evaluation. The StateGraph API itself is external to TruLens.

Usage

Use StateGraph when building multi-step agentic workflows with LangGraph. Define your graph, compile it, then wrap with TruGraph for evaluation.

Code Reference

Source Location

• Repository: langgraph (external)
• File: langgraph/graph/state.py (external)

Signature

class StateGraph:
    def __init__(self, state_schema: Type):
        """
        Args:
            state_schema: Pydantic model or TypedDict defining the graph state.
        """

    def add_node(self, name: str, function: Callable) -> StateGraph: ...
    def add_edge(self, start: str, end: str) -> StateGraph: ...
    def add_conditional_edges(self, source: str, condition: Callable, mapping: Dict) -> StateGraph: ...
    def set_entry_point(self, name: str) -> StateGraph: ...
    def compile(self) -> CompiledStateGraph: ...

Import

from langgraph.graph import StateGraph

External Reference

• LangGraph Documentation

I/O Contract

Inputs

Name	Type	Required	Description
state_schema	Type	Yes	Pydantic model or TypedDict defining the graph state
nodes	Dict[str, Callable]	Yes	Named processing functions added via add_node
edges	List[Tuple]	Yes	Connections between nodes via add_edge/add_conditional_edges

Outputs

Name	Type	Description
compile() returns	CompiledStateGraph	Compiled graph (Pregel) ready for invocation and TruGraph wrapping

Usage Examples

Build and Compile Agent Graph

from langgraph.graph import StateGraph
from typing import TypedDict, List

class AgentState(TypedDict):
    messages: List[dict]

def agent_node(state: AgentState) -> dict:
    return {"messages": state["messages"] + [{"role": "assistant", "content": "..."}]}

def tool_node(state: AgentState) -> dict:
    return {"messages": state["messages"] + [{"role": "tool", "content": "..."}]}

def should_continue(state: AgentState) -> str:
    last = state["messages"][-1]
    if "tool_call" in last.get("content", ""):
        return "continue"
    return "end"

graph = StateGraph(AgentState)
graph.add_node("agent", agent_node)
graph.add_node("tools", tool_node)
graph.add_conditional_edges("agent", should_continue, {"continue": "tools", "end": "__end__"})
graph.add_edge("tools", "agent")
graph.set_entry_point("agent")

compiled = graph.compile()

# Now wrap with TruGraph for evaluation
from trulens.apps.langgraph import TruGraph
tru_agent = TruGraph(compiled, app_name="Agent_v1", feedbacks=[...])

Related Pages

Implements Principle

• Principle:Truera_Trulens_LangGraph_Agent_Construction

Requires Environment

• Environment:Truera_Trulens_LangChain_LangGraph_Environment