Validation Gates and Multi-Agent Routing for a LangGraph Support Bot
Add deterministic business-logic gates, a self-correcting validation loop and a supervisor with three specialist subgraphs to an e-commerce LangGraph agent.
This article is published in English.
Picture a customer-support bot for an online shop, called ShopBot here, built on LangGraph. It looks up orders, checks refund eligibility, processes refunds after a human signs off, and keeps the conversation summarised. A basic version of that bot already works. What follows adds three things that a bot with real consequences needs: validation gates for every model output, a loop that corrects the model automatically, and a multi-agent layout in which a supervisor hands work to specialists. The target is a codebase you could bring to a production review without flinching.
The assumed baseline is a graph with the usual modules: tools, a ReAct-style loop, state, and an interrupt() call that pauses the graph for human approval.
The silent assumption in the basic bot
The simple version trusts the language model everywhere. Its agent calls an eligibility check and trusts that the order ID it passes was extracted correctly from the chat. The review node shows that order ID to a human, but the reviewer approves or rejects an action and isn't validating whether the ID exists. The refund tool runs with whatever arguments the model chose. If the model hallucinates "ORD-99X" where "ORD-001" was meant, the tool errors, the agent sees the error, and something reasonable might happen. "Something reasonable might happen" is a weak description of error handling in a system that moves money.
Practitioners who run automated procurement or compliance workflows make the point sharply: probabilistic failures aren't affordable there. The answer isn't to trust the model more. It is to stop trusting its output at all, and to check every structured result with deterministic Python before it touches anything external.
Three additions, and how they fit together
Business logic gates live inside nodes and validate model output before anything acts on it. They are plain Python with no model calls. A gate either lets the data through or rejects it with a specific, structured reason. That's deterministic validation inside a probabilistic system.
Validation loops deal with gate failures. A rejection doesn't crash the graph and doesn't quietly yield a wrong answer. Control moves to a correction node that passes the failure reason to the model and asks for a fixed output. A retry counter in state prevents endless looping, and a hard cap triggers escalation to a human.
A multi-agent architecture replaces the single agent node with three specialists. An Order Agent covers lookups and status questions. A Refund Agent covers eligibility checks and refund processing. A Complaints Agent handles upset customers and escalations. A supervisor node inspects every incoming message and sends it to the right specialist, and every specialist is a compiled subgraph of its own.
These pieces work as one system. The Refund Agent carries a gate, a failing gate can start a validation loop, and the supervisor looks at how that loop ended before choosing to retry, escalate or abort. Put together, they give the finished bot the sturdiness an enterprise setting demands.
Plan the graph before touching code
LangGraph rewards planning, and the larger the graph, the more true that is. Sketch the whole system first. Every box should map to exactly one node or routing function, and every arrow to exactly one edge. The code then becomes a translation of that picture into Python, module by module.
The new file layout
The basic bot fit in a single file. At this size one file gets unwieldy, so the project splits in three.
shopbot_enterprise/
├── gates.py ← Business Logic Gates (pure Python, no LangGraph imports)
├── agents.py ← All three specialist subgraphs
└── shopbot.py ← Supervisor graph + entrypoint (unchanged module structure)
Keeping LangGraph imports out of the gates file is deliberate. Gates become testable with ordinary unit tests: no model, no graph and no running process is needed to confirm that the validation logic is right. You call a gate with sample data and assert on the result, and that is what "deterministic" means in practice.
Extending the state schema
The state needs new fields before anything else. The baseline held three custom fields: a summary, the customer's name and a ticket category. Four more are added.
# ============================================================
# MODULE 2: STATE (Enterprise Version)
# ============================================================
from typing import TypedDict, Annotated
from langgraph.graph import MessagesState
import operator
class SupportState(MessagesState):
# From the original ShopBot - unchanged
summary: str
customer_name: str
ticket_category: str
# NEW: The structured data the LLM extracted from the conversation.
# Gates validate this field. Tools use this field. The LLM never
# writes to external systems directly - it writes here first.
extracted_data: dict
# NEW: The current status of the gate decision for this turn.
# Values: "pending" | "passed" | "failed" | "escalated" | "aborted"
# The supervisor reads this to decide what to do after a gate runs.
gate_status: str
# NEW: A log of every gate failure in this conversation.
# Uses operator.add so entries accumulate rather than replace.
# This is your audit trail. Every compliance system needs one.
validation_errors: Annotated[list[str], operator.add]
# NEW: How many correction attempts have fired in this turn.
# The hard cap is 2. After 2 failed corrections, escalate to human.
# Without this, a broken LLM output creates an infinite loop.
retry_count: int
The key new field is extracted_data. In the basic bot the model produced free-form text and called tools directly, so tool arguments were whatever the model put in its call. Now the model first emits structured data, which goes into extracted_data. The gate checks it, and only after the gate passes does that data become a tool argument. That removes any direct route from the model to external systems.
The validation_errors field is annotated with operator.add, so it accumulates across the full conversation and not just one turn. Suppose a made-up order ID appears in turn three and a poor refund reason in turn seven: both failures end up in the audit log. That matters in compliance reviews.
Writing the gates
A gate is a pure validation function. It takes extracted data and returns a tuple: (True, cleaned_data) on success or (False, error_reason) on failure. That is the whole interface, with no model calls and no graph dependencies. The module below includes an order ID check, a full refund check and a complaint check, plus the business constants they use.
# ============================================================
# gates.py — Business Logic Gates
# Pure Python. No LangGraph. No LLM calls.
# Each gate returns (bool, data_or_error_string).
# ============================================================
import re
# The same fake database as the original ShopBot.
# In production this would be a database query.
ORDERS_DB = {
"ORD-001": {"customer": "Alex", "product": "Wireless Headphones",
"status": "Delivered", "amount": 89.99,
"delivery_date": "2025-06-10"},
"ORD-002": {"customer": "Sam", "product": "Phone Case",
"status": "In Transit", "amount": 14.99,
"delivery_date": "Expected 2025-06-18"},
"ORD-003": {"customer": "Jordan", "product": "Laptop Stand",
"status": "Processing", "amount": 45.00,
"delivery_date": "Expected 2025-06-20"},
}
MAX_REFUND_AMOUNT = 500.00 # Business rule: auto-refunds above this need director sign-off
MIN_REASON_LENGTH = 10 # Business rule: refund reasons must be substantive
def validate_order_id(extracted_data: dict) -> tuple[bool, dict | str]:
"""
Gate 1: Order ID Validation
Checks that the order ID the LLM extracted actually exists in our system
and follows the correct format. This is the most basic hallucination check
in a support workflow. LLMs frequently confuse order IDs, especially when
a customer has mentioned multiple orders or used an informal format like
"order one" instead of "ORD-001".
Why this matters: if the agent calls a tool with a hallucinated order ID,
the tool returns an error, the agent sees the error, and may hallucinate
a recovery that looks plausible but is wrong. Catching it here, before
the tool call, is the correct place to stop the failure.
"""
order_id = extracted_data.get("order_id", "").strip().upper()
# Check 1: Format validation - must match ORD-XXX pattern
if not re.match(r"^ORD-\d{3}quot;, order_id):
return (
False,
f"Order ID '{order_id}' does not match the required format ORD-XXX "
f"(example: ORD-001). The LLM may have extracted the wrong value. "
f"Ask the customer to confirm their exact order number."
)
# Check 2: Existence validation - must exist in our system
if order_id not in ORDERS_DB:
return (
False,
f"Order ID '{order_id}' passed format validation but does not exist "
f"in the orders database. This is likely a hallucination. "
f"Ask the customer to provide the exact order ID from their confirmation email."
)
# Gate passed: return cleaned data with the normalised order ID
cleaned = {**extracted_data, "order_id": order_id}
return True, cleaned
def validate_refund_request(extracted_data: dict) -> tuple[bool, dict | str]:
"""
Gate 2: Refund Request Validation
A compound gate that checks every business rule that must be true
before a refund can be processed. Runs after validate_order_id passes,
so we can assume order_id is valid at this point.
Business rules checked:
1. The order must exist and have Delivered status
2. The refund amount must be below the auto-approval threshold
3. The reason must be substantive (not empty or one word)
Each failed rule produces a specific error message that gets fed back
to the LLM as a correction instruction. Vague errors produce vague
corrections. Specific errors produce specific corrections.
"""
order_id = extracted_data.get("order_id", "").strip().upper()
reason = extracted_data.get("reason", "").strip()
order = ORDERS_DB.get(order_id, {})
errors = []
# Rule 1: Order must be in Delivered status
if order.get("status") != "Delivered":
errors.append(
f"Order {order_id} has status '{order.get('status')}'. "
f"Refunds can only be processed for Delivered orders. "
f"Inform the customer and do not attempt to process a refund."
)
# Rule 2: Amount must be within auto-approval limit
amount = order.get("amount", 0)
if amount > MAX_REFUND_AMOUNT:
errors.append(
f"Refund amount ${amount:.2f} exceeds the auto-approval limit "
f"of ${MAX_REFUND_AMOUNT:.2f}. This requires director-level sign-off. "
f"Escalate to a senior agent - do not process automatically."
)
# Rule 3: Reason must be substantive
if len(reason) < MIN_REASON_LENGTH:
errors.append(
f"Refund reason '{reason}' is too short (minimum {MIN_REASON_LENGTH} characters). "
f"Ask the customer to provide a more detailed reason for their return request."
)
if errors:
# Return all errors joined - the correction node will feed this
# exact string back to the LLM as a correction instruction
return False, " | ".join(errors)
return True, extracted_data
def validate_complaint(extracted_data: dict) -> tuple[bool, dict | str]:
"""
Gate 3: Complaint Validation
Ensures the complaints agent has enough structured data to work with
before attempting any escalation or response. Complaints without
a clear issue category cannot be routed correctly in a real ticketing system.
"""
issue_summary = extracted_data.get("issue_summary", "").strip()
sentiment = extracted_data.get("sentiment", "").strip().lower()
valid_sentiments = {"frustrated", "angry", "neutral", "disappointed"}
if len(issue_summary) < 20:
return (
False,
f"Complaint issue summary is too vague: '{issue_summary}'. "
f"Ask the customer to describe their problem in more detail before escalating."
)
if sentiment not in valid_sentiments:
return (
False,
f"Sentiment '{sentiment}' is not a recognised value. "
f"Valid values are: {', '.join(valid_sentiments)}. "
f"Re-classify the customer's sentiment and try again."
)
return True, extracted_data
Look at what the file lacks: imports from LangGraph, LangChain or any AI library. Any developer on the team can read, test and change it without knowing anything about agent architecture. When an auditor asks what exactly validates a refund request before money moves, this is the file to show, and non-engineers can follow it.
The validation loop
The correction node is what makes gates useful. When a gate fails, execution doesn't stop but routes here. The node takes the specific failure reason from state, builds an instruction telling the model exactly what went wrong, and asks for a corrected structured output.
# ── agents.py (partial) ─────────────────────────────────────
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage, SystemMessage, AIMessage
from pydantic import BaseModel, Field
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
# ── Pydantic Models (Structured LLM Output) ──────────────────
# These are the schemas that define what the LLM must return.
# with_structured_output() enforces these - if the LLM returns
# something that doesn't match, Pydantic raises a validation error
# before the output even reaches a gate.
class OrderExtraction(BaseModel):
order_id: str = Field(description="The order ID in the format ORD-XXX, e.g. ORD-001")
query_type: str = Field(description="What the customer wants to know: 'status', 'delivery_date', 'product', or 'general'")
class RefundExtraction(BaseModel):
order_id: str = Field(description="The order ID in the format ORD-XXX")
reason: str = Field(description="The customer's stated reason for wanting a refund, in their own words")
class ComplaintExtraction(BaseModel):
issue_summary: str = Field(description="A one-sentence summary of the customer's complaint")
sentiment: str = Field(description="The customer's emotional state: 'frustrated', 'angry', 'neutral', or 'disappointed'")
# ── The Correction Node ───────────────────────────────────────
def correction_node(state) -> dict:
"""
The validation loop's core. When a Business Logic Gate fails, this
node receives control. It reads the most recent gate failure from
state, builds a precise correction instruction for the LLM, and
asks for a new extraction attempt.
This node is shared across all three specialist agents. The
ticket_category in state tells it which Pydantic model to use
for the re-extraction, so the same node handles order corrections,
refund corrections, and complaint corrections without duplicating logic.
Critical design decision: the correction instruction goes into the
messages list as a HumanMessage, not a SystemMessage. This makes it
part of the visible conversation history. If the LLM makes the same
mistake twice, the history shows both the first failure and the second
correction attempt, which helps diagnose systematic prompt problems.
"""
# Read the most recent validation error from the audit log
errors = state.get("validation_errors", [])
last_error = errors[-1] if errors else "Unknown validation error."
retry_count = state.get("retry_count", 0)
# Hard stop: after 2 retries, do not attempt again - escalate
if retry_count >= 2:
return {
"gate_status": "escalated",
"messages": [
AIMessage(
content=(
f"I was unable to process your request correctly after {retry_count} attempts. "
f"I'm escalating this to a senior support agent who will contact you directly. "
f"I apologise for the inconvenience."
)
)
]
}
# Build the correction instruction - be maximally specific
correction_instruction = (
f"Your previous response contained a validation error that must be corrected "
f"before we can proceed. The exact error was:\n\n"
f"{last_error}\n\n"
f"Please re-read the conversation carefully and provide a corrected response "
f"that addresses this specific error. Do not repeat the same mistake."
)
category = state.get("ticket_category", "order_inquiry")
# Choose the right structured output model based on what kind of
# correction is needed
if category == "refund_request":
structured_llm = llm.with_structured_output(RefundExtraction)
elif category == "complaint":
structured_llm = llm.with_structured_output(ComplaintExtraction)
else:
structured_llm = llm.with_structured_output(OrderExtraction)
messages = (
[SystemMessage(content="You are a customer support assistant. Extract the requested information accurately.")]
+ state["messages"]
+ [HumanMessage(content=correction_instruction)]
)
try:
corrected = structured_llm.invoke(messages)
return {
"extracted_data": corrected.model_dump(),
"gate_status": "pending",
"retry_count": retry_count + 1,
"messages": [HumanMessage(content=correction_instruction)],
}
except Exception as e:
# Structured output itself failed (Pydantic validation error)
# Count this as a retry and surface the raw error
return {
"gate_status": "failed",
"retry_count": retry_count + 1,
"validation_errors": [f"Structured output failed: {str(e)}"],
}
The retry check at the top is the most important safety mechanism in the whole system. Without it, a truly broken input (a customer who won't supply a valid order ID, or a model configuration that keeps hallucinating) would spin forever, bouncing between the correction node and the gate. Two retries makes a sensible default for support work, since it gives the model two chances to fix itself before a human gets involved. The number can be tuned according to how much latency versus accuracy is tolerable.
The three specialist subgraphs
Each specialist is a compiled subgraph with its own tools, extraction logic, gate and internal routing. The parent supervisor graph neither knows nor cares about that internal structure. It sees a node that receives state and returns state.
The Order Agent
# ── agents.py (continued) ────────────────────────────────────
from langchain_core.tools import tool
from langgraph.graph import StateGraph, START, END
from langgraph.prebuilt import ToolNode
from gates import validate_order_id, ORDERS_DB
@tool
def lookup_order(order_id: str) -> str:
"""Look up the details of a customer's order by order ID.
Use this when the customer wants to know their order status,
product name, or expected delivery date.
"""
order = ORDERS_DB.get(order_id.upper())
if not order:
return f"No order found with ID '{order_id}'."
return (
f"Order {order_id.upper()}: {order['product']} | "
f"Status: {order['status']} | "
f"Amount: ${order['amount']:.2f} | "
f"Delivery: {order['delivery_date']}"
)
order_tools = [lookup_order]
order_tool_node = ToolNode(order_tools)
order_llm = llm.bind_tools(order_tools)
def order_extract_node(state) -> dict:
"""
Step 1 for the Order Agent: extract the order ID from the conversation
using structured output. The result goes into state['extracted_data'].
The gate runs next. The tool call only happens if the gate passes.
"""
structured_llm = llm.with_structured_output(OrderExtraction)
messages = (
[SystemMessage(content="Extract the order ID and query type from the customer's message.")]
+ state["messages"]
)
try:
result = structured_llm.invoke(messages)
return {
"extracted_data": result.model_dump(),
"gate_status": "pending",
"ticket_category": "order_inquiry",
}
except Exception as e:
return {
"gate_status": "failed",
"validation_errors": [f"Extraction failed: {str(e)}"],
"ticket_category": "order_inquiry",
}
def order_gate_node(state) -> dict:
"""
Step 2: Run the Business Logic Gate on the extracted data.
Either passes gate_status to 'passed' and lets the tool run,
or sets it to 'failed' and logs the reason to validation_errors.
"""
extracted = state.get("extracted_data", {})
passed, result = validate_order_id(extracted)
if passed:
return {"gate_status": "passed", "extracted_data": result}
return {
"gate_status": "failed",
"validation_errors": [result], # operator.add accumulates this
}
def order_agent_node(state) -> dict:
"""
Step 3 (only runs after gate passes): call the LLM with tools bound.
Now the LLM uses the validated order_id from extracted_data rather
than whatever it decided to extract ad hoc.
"""
extracted = state.get("extracted_data", {})
order_id = extracted.get("order_id", "")
messages = (
[SystemMessage(content=(
"You are the Order Agent for ShopBot. Your only job is to look up "
"order information and report it clearly to the customer. "
f"The validated order ID is {order_id}. Use lookup_order to get the details."
))]
+ state["messages"]
)
response = order_llm.invoke(messages)
return {"messages": [response]}
def route_order_gate(state) -> str:
status = state.get("gate_status", "pending")
if status == "passed":
return "order_agent_node"
if state.get("retry_count", 0) >= 2:
return "order_agent_node" # correction_node already added the escalation message
return "correction_node"
def route_order_agent(state) -> str:
last = state["messages"][-1]
if hasattr(last, "tool_calls") and last.tool_calls:
return "order_tools"
return "__end__"
# ── Order subgraph assembly ──────────────────────────────────
order_builder = StateGraph(state_schema=type(None)) # Uses parent state directly
order_builder = StateGraph(dict)
order_builder.add_node("order_extract_node", order_extract_node)
order_builder.add_node("order_gate_node", order_gate_node)
order_builder.add_node("correction_node", correction_node)
order_builder.add_node("order_agent_node", order_agent_node)
order_builder.add_node("order_tools", order_tool_node)
order_builder.add_edge(START, "order_extract_node")
order_builder.add_edge("order_extract_node", "order_gate_node")
order_builder.add_conditional_edges(
"order_gate_node",
route_order_gate,
{"order_agent_node": "order_agent_node", "correction_node": "correction_node"}
)
order_builder.add_edge("correction_node", "order_gate_node")
order_builder.add_conditional_edges(
"order_agent_node",
route_order_agent,
{"order_tools": "order_tools", "__end__": END}
)
order_builder.add_edge("order_tools", "order_agent_node")
order_agent = order_builder.compile()
Trace the loop in this subgraph. The order gate node feeds the correction node on failure, and the correction node feeds back into the gate. That cycle is the validation loop, and it runs at most twice before the counter produces the escalation message.
The Refund Agent
This specialist matters most because it handles money. It runs two gates, first order ID validation and then the full refund validation, and the human-approval interrupt stays inside this subgraph.
# ── agents.py (continued) ────────────────────────────────────
from gates import validate_order_id, validate_refund_request
from langgraph.types import interrupt, Command
from langchain_core.messages import ToolMessage
from datetime import datetime
@tool
def check_refund_eligibility(order_id: str) -> str:
"""Check whether an order is eligible for a refund.
An order is eligible only if its status is Delivered.
"""
order = ORDERS_DB.get(order_id.upper())
if not order:
return f"Cannot check refund: order '{order_id}' not found."
if order["status"] == "Delivered":
return (
f"Order {order_id.upper()} IS eligible for a refund. "
f"Amount: ${order['amount']:.2f}."
)
return (
f"Order {order_id.upper()} is NOT eligible. "
f"Status: {order['status']}."
)
@tool
def process_refund(order_id: str, reason: str) -> str:
"""Process a refund for a delivered order.
Only call this after human approval has been confirmed.
"""
order = ORDERS_DB.get(order_id.upper())
if not order:
return f"Refund failed: order '{order_id}' not found."
ref = f"REF-{order_id.upper()}-{datetime.now().strftime('%H%M%S')}"
return (
f"Refund APPROVED and PROCESSED. Reference: {ref}. "
f"${order['amount']:.2f} will be returned within 3 to 5 business days."
)
refund_tools = [check_refund_eligibility, process_refund]
refund_tool_node = ToolNode(refund_tools)
refund_llm = llm.bind_tools(refund_tools)
def refund_extract_node(state) -> dict:
structured_llm = llm.with_structured_output(RefundExtraction)
messages = (
[SystemMessage(content="Extract the order ID and refund reason from the customer's message.")]
+ state["messages"]
)
try:
result = structured_llm.invoke(messages)
return {
"extracted_data": result.model_dump(),
"gate_status": "pending",
"ticket_category": "refund_request",
}
except Exception as e:
return {
"gate_status": "failed",
"validation_errors": [f"Refund extraction failed: {str(e)}"],
"ticket_category": "refund_request",
}
def refund_gate_node(state) -> dict:
"""
Compound gate: runs both order ID validation and full refund validation.
Both must pass for execution to continue.
"""
extracted = state.get("extracted_data", {})
# Gate 1: Order ID
id_passed, id_result = validate_order_id(extracted)
if not id_passed:
return {"gate_status": "failed", "validation_errors": [id_result]}
# Gate 2: Full refund rules (eligibility, amount, reason)
ref_passed, ref_result = validate_refund_request(id_result)
if not ref_passed:
return {"gate_status": "failed", "validation_errors": [ref_result]}
return {"gate_status": "passed", "extracted_data": ref_result}
def refund_review_node(state) -> dict:
"""
The human approval gate, carried over from the original ShopBot
and now living inside the Refund Agent subgraph.
interrupt() pauses the entire graph here. The parent graph's
checkpointer saves the state. A human reviews the details.
Command(resume=...) restarts from this exact line.
"""
last_message = state["messages"][-1]
refund_tool_call = None
if hasattr(last_message, "tool_calls"):
for tc in last_message.tool_calls:
if "refund" in tc["name"].lower():
refund_tool_call = tc
break
extracted = state.get("extracted_data", {})
human_decision = interrupt({
"message": "Refund approval required",
"order_id": extracted.get("order_id", "Unknown"),
"reason": extracted.get("reason", "Not provided"),
"tool_being_called": refund_tool_call["name"] if refund_tool_call else "process_refund",
"arguments": refund_tool_call["args"] if refund_tool_call else extracted,
"summary": state.get("summary", "No summary yet"),
"options": ["approve", "reject", "escalate"],
})
if human_decision == "approve":
return {}
if human_decision in ("reject", "escalate"):
label = "declined" if human_decision == "reject" else "escalated to a senior agent"
return {
"messages": [
ToolMessage(
content=f"Refund request was {label} by the support team.",
tool_call_id=refund_tool_call["id"] if refund_tool_call else "unknown",
)
]
}
return {}
def refund_agent_node(state) -> dict:
extracted = state.get("extracted_data", {})
messages = (
[SystemMessage(content=(
"You are the Refund Agent for ShopBot. Your job is to check refund "
"eligibility and process approved refunds. "
f"Validated order ID: {extracted.get('order_id', '')}. "
f"Customer reason: {extracted.get('reason', '')}. "
"Always check eligibility first, then process if eligible and approved."
))]
+ state["messages"]
)
response = refund_llm.invoke(messages)
return {"messages": [response]}
def route_refund_gate(state) -> str:
status = state.get("gate_status", "pending")
if status == "passed":
return "refund_agent_node"
if state.get("retry_count", 0) >= 2:
return "refund_agent_node"
return "correction_node"
def route_refund_agent(state) -> str:
last = state["messages"][-1]
if hasattr(last, "tool_calls") and last.tool_calls:
tool_names = [tc["name"] for tc in last.tool_calls]
if "process_refund" in tool_names:
return "refund_review_node"
return "refund_tools"
return "__end__"
def route_after_review(state) -> str:
last = state["messages"][-1]
if isinstance(last, ToolMessage):
return "refund_agent_node"
return "refund_tools"
# ── Refund subgraph assembly ─────────────────────────────────
refund_builder = StateGraph(dict)
refund_builder.add_node("refund_extract_node", refund_extract_node)
refund_builder.add_node("refund_gate_node", refund_gate_node)
refund_builder.add_node("correction_node", correction_node)
refund_builder.add_node("refund_agent_node", refund_agent_node)
refund_builder.add_node("refund_review_node", refund_review_node)
refund_builder.add_node("refund_tools", refund_tool_node)
refund_builder.add_edge(START, "refund_extract_node")
refund_builder.add_edge("refund_extract_node", "refund_gate_node")
refund_builder.add_conditional_edges(
"refund_gate_node",
route_refund_gate,
{"refund_agent_node": "refund_agent_node", "correction_node": "correction_node"}
)
refund_builder.add_edge("correction_node", "refund_gate_node")
refund_builder.add_conditional_edges(
"refund_agent_node",
route_refund_agent,
{
"refund_review_node": "refund_review_node",
"refund_tools": "refund_tools",
"__end__": END,
}
)
refund_builder.add_conditional_edges(
"refund_review_node",
route_after_review,
{"refund_tools": "refund_tools", "refund_agent_node": "refund_agent_node"}
)
refund_builder.add_edge("refund_tools", "refund_agent_node")
refund_agent = refund_builder.compile()
The Complaints Agent
This is the least complicated specialist. It acknowledges the customer's frustration, confirms there is enough detail to open a ticket, and hands off to a human through interrupt() when the sentiment runs strong.
# ── agents.py (continued) ────────────────────────────────────
from gates import validate_complaint
complaints_llm = llm # Complaints agent doesn't need tool-calling
def complaints_extract_node(state) -> dict:
structured_llm = llm.with_structured_output(ComplaintExtraction)
messages = (
[SystemMessage(content=(
"Extract the complaint summary and the customer's emotional sentiment "
"from the conversation. Sentiment must be one of: frustrated, angry, "
"neutral, disappointed."
))]
+ state["messages"]
)
try:
result = structured_llm.invoke(messages)
return {
"extracted_data": result.model_dump(),
"gate_status": "pending",
"ticket_category": "complaint",
}
except Exception as e:
return {
"gate_status": "failed",
"validation_errors": [f"Complaint extraction failed: {str(e)}"],
"ticket_category": "complaint",
}
def complaints_gate_node(state) -> dict:
extracted = state.get("extracted_data", {})
passed, result = validate_complaint(extracted)
if passed:
return {"gate_status": "passed", "extracted_data": result}
return {"gate_status": "failed", "validation_errors": [result]}
def complaints_respond_node(state) -> dict:
"""
For high-sentiment complaints (frustrated, angry), trigger a human
escalation via interrupt(). For lower-sentiment complaints, the LLM
responds empathetically and the turn ends.
This mirrors the Part 3 pattern but applied conditionally based on
the validated sentiment field, not a keyword scan of free-form text.
The gate already confirmed the sentiment is a valid value, so this
routing decision is deterministic.
"""
extracted = state.get("extracted_data", {})
sentiment = extracted.get("sentiment", "neutral")
if sentiment in ("frustrated", "angry"):
human_decision = interrupt({
"message": "High-sentiment complaint requires human review",
"issue_summary": extracted.get("issue_summary", ""),
"sentiment": sentiment,
"options": ["escalate_to_senior", "provide_compensation", "close_ticket"],
})
if human_decision == "escalate_to_senior":
response_text = (
"I completely understand your frustration, and I want to make sure "
"you receive the best possible support. I am escalating your case to "
"a senior support agent who will contact you within 2 business hours. "
"Your ticket has been flagged as priority."
)
elif human_decision == "provide_compensation":
response_text = (
"I sincerely apologise for the experience you have had. As a gesture "
"of goodwill, our team will be in touch within 24 hours with a resolution "
"that includes appropriate compensation for the inconvenience caused."
)
else:
response_text = (
"Thank you for bringing this to our attention. Your feedback has been "
"recorded and will help us improve our service."
)
from langchain_core.messages import AIMessage
return {"messages": [AIMessage(content=response_text)]}
# Lower-sentiment complaints: LLM responds empathetically
messages = (
[SystemMessage(content=(
"You are the Complaints Agent for ShopBot. The customer has a complaint. "
"Respond with genuine empathy, acknowledge their specific issue, and "
"explain what steps will be taken. Be warm and professional."
))]
+ state["messages"]
)
response = complaints_llm.invoke(messages)
return {"messages": [response]}
def route_complaints_gate(state) -> str:
status = state.get("gate_status", "pending")
if status == "passed":
return "complaints_respond_node"
if state.get("retry_count", 0) >= 2:
return "complaints_respond_node"
return "correction_node"
# ── Complaints subgraph assembly ─────────────────────────────
complaints_builder = StateGraph(dict)
complaints_builder.add_node("complaints_extract_node", complaints_extract_node)
complaints_builder.add_node("complaints_gate_node", complaints_gate_node)
complaints_builder.add_node("correction_node", correction_node)
complaints_builder.add_node("complaints_respond_node", complaints_respond_node)
complaints_builder.add_edge(START, "complaints_extract_node")
complaints_builder.add_edge("complaints_extract_node", "complaints_gate_node")
complaints_builder.add_conditional_edges(
"complaints_gate_node",
route_complaints_gate,
{
"complaints_respond_node": "complaints_respond_node",
"correction_node": "correction_node",
}
)
complaints_builder.add_edge("correction_node", "complaints_gate_node")
complaints_builder.add_edge("complaints_respond_node", END)
complaints_agent = complaints_builder.compile()
The supervisor graph
With the subgraphs compiled, the supervisor graph turns out remarkably simple. It only routes. It never does specialist work. The principle behind multi-agent patterns is exactly this: supervisors route and specialists execute.
# ============================================================
# shopbot.py — Supervisor Graph (Enterprise Version)
# ============================================================
# ── MODULE 1: IMPORTS & CONFIGURATION ───────────────────────
import os
from typing import Literal
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage, SystemMessage, RemoveMessage, AIMessage
from langchain.chat_models import init_chat_model
from langgraph.graph import StateGraph, MessagesState, START, END
from langgraph.checkpoint.memory import MemorySaver
from langgraph.types import Command
# Import the compiled subgraphs from agents.py
from agents import order_agent, refund_agent, complaints_agent
from state import SupportState # The extended state from Module 2 above
load_dotenv()
# The supervisor uses a plain LLM - no tools, just routing decisions
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
# ── MODULE 4: SUPERVISOR NODE ────────────────────────────────
SUPERVISOR_SYSTEM = """You are the ShopBot Supervisor. Your only job is to read
the customer's message and decide which specialist should handle it.
Available specialists:
order_agent - handles order status, delivery date, and product questions
refund_agent - handles refund requests and return eligibility
complaints_agent - handles complaints, frustration, and negative feedback
FINISH - use this only when the conversation is fully resolved
Respond with exactly one word: the name of the specialist, or FINISH."""
def supervisor_node(state: SupportState) -> dict:
"""
Reads the full conversation and routes to the correct specialist.
Returns a next_agent string that the routing function reads.
Uses plain LLM with no tools - the supervisor decides, it never acts.
This separation keeps the routing logic deterministic and auditable.
"""
messages = [SystemMessage(content=SUPERVISOR_SYSTEM)] + state["messages"]
response = llm.invoke(messages)
decision = response.content.strip().lower()
# Normalise the response in case the LLM adds punctuation or casing
if "refund" in decision:
next_agent = "refund_agent"
elif "complaint" in decision or "frustrated" in decision:
next_agent = "complaints_agent"
elif "order" in decision:
next_agent = "order_agent"
elif "finish" in decision:
next_agent = "FINISH"
else:
next_agent = "order_agent" # Safe default for ambiguous routing
return {"next_agent": next_agent}
# ── MODULE 4 CONTINUED: SUMMARIZE NODE ───────────────────────
# Carried over from the original ShopBot without changes.
# The enterprise version still needs conversation compression.
def summarize_node(state: SupportState) -> dict:
existing_summary = state.get("summary", "")
instruction = (
f"Current summary:\n{existing_summary}\n\nExtend with new messages. "
"Under 5 sentences. Include customer name, issue, orders mentioned, actions taken."
if existing_summary
else "Summarise this support conversation in under 5 sentences."
)
messages = state["messages"] + [HumanMessage(content=instruction)]
response = llm.invoke(messages)
to_delete = [RemoveMessage(id=m.id) for m in state["messages"][:-2]]
return {"summary": response.content, "messages": to_delete}
# ── MODULE 5: ROUTING ────────────────────────────────────────
def route_supervisor(state: SupportState) -> Literal[
"order_agent", "refund_agent", "complaints_agent", "summarize_node", "__end__"
]:
"""
Reads next_agent from state and routes accordingly.
Also checks whether the conversation is long enough to need summarisation.
"""
next_agent = state.get("next_agent", "order_agent")
if next_agent == "FINISH":
if len(state["messages"]) > 6:
return "summarize_node"
return "__end__"
return next_agent
# ── MODULE 6: GRAPH ASSEMBLY ─────────────────────────────────
graph_builder = StateGraph(SupportState)
graph_builder.add_node("supervisor", supervisor_node)
graph_builder.add_node("order_agent", order_agent) # compiled subgraph
graph_builder.add_node("refund_agent", refund_agent) # compiled subgraph
graph_builder.add_node("complaints_agent", complaints_agent) # compiled subgraph
graph_builder.add_node("summarize_node", summarize_node)
graph_builder.add_edge(START, "supervisor")
graph_builder.add_conditional_edges(
"supervisor",
route_supervisor,
{
"order_agent": "order_agent",
"refund_agent": "refund_agent",
"complaints_agent": "complaints_agent",
"summarize_node": "summarize_node",
"__end__": END,
}
)
# After each specialist finishes, always return to supervisor.
# The supervisor re-evaluates and decides: another specialist, or FINISH.
graph_builder.add_edge("order_agent", "supervisor")
graph_builder.add_edge("refund_agent", "supervisor")
graph_builder.add_edge("complaints_agent", "supervisor")
graph_builder.add_edge("summarize_node", END)
memory = MemorySaver()
shopbot = graph_builder.compile(checkpointer=memory)
# ── MODULE 7: ENTRYPOINT ──────────────────────────────────────
def run_shopbot():
print("=" * 60)
print(" ShopBot Enterprise - Multi-Agent Customer Support")
print("=" * 60)
print("Type your message. Type 'exit' to quit. Type 'audit' to")
print("see the validation error log for this session.\n")
config = {"configurable": {"thread_id": "enterprise-session-001"}}
while True:
user_input = input("You: ").strip()
if not user_input:
continue
if user_input.lower() == "exit":
print("ShopBot: Thank you for contacting ShopBot. Have a great day!")
break
if user_input.lower() == "audit":
snapshot = shopbot.get_state(config)
errors = snapshot.values.get("validation_errors", [])
print(f"\n[AUDIT LOG] {len(errors)} validation event(s) this session:")
for i, err in enumerate(errors, 1):
print(f" {i}. {err}")
print()
continue
result = shopbot.invoke(
{"messages": [HumanMessage(content=user_input)]},
config=config,
)
while "__interrupt__" in result:
interrupt_data = result["__interrupt__"][0].value
print("\n" + "=" * 60)
print(" HUMAN REVIEW REQUIRED")
print("=" * 60)
for key, value in interrupt_data.items():
if key != "options":
print(f" {key.replace('_', ' ').title()}: {value}")
options = interrupt_data.get("options", [])
print(f"\n Options: {' | '.join(f'[{o[0]}] {o}' for o in options)}")
raw = input(" Your decision: ").strip().lower()
chosen = next((o for o in options if o.startswith(raw)), options[-1])
result = shopbot.invoke(Command(resume=chosen), config=config)
last = result["messages"][-1]
print(f"\nShopBot: {last.content}\n")
if __name__ == "__main__":
run_shopbot()
What changes in behaviour
Both the basic and the enterprise version are valid implementations of the standard LangGraph patterns, but they make different assumptions, and the difference counts once you leave the sandbox for a system with real consequences.
The basic agent assumed the model was right. If the model extracted "ORD-0O1", with a letter O in place of a zero, the tool would error, the agent would probably recover, and the customer might get a slightly confusing reply. Probably fine.
The enterprise agent assumes the model will sometimes be wrong and builds a recovery path before that mistake reaches anything external. The gate catches "ORD-0O1" at once and produces an error message naming precisely what was wrong and why. The correction node hands that exact message back to the model, and the model gets a second try with an exact instruction. A repeat failure hands the conversation to a person. No tool ran with bad data, and no external system ever saw the hallucination.
Thanks to validation_errors in state, a senior engineer reviewing a session that gave a wrong answer has a full log of every gate failure, correction attempt and retry count. The failure can be reproduced and understood, and then either the prompt is fixed or the gate is tightened. With no such log, debugging a probabilistic system means guessing.
The three-file structure pays off with auditors too. Asked what validates a refund before money moves, the answer is one file of pure Python with no agent architecture to explain: clear, auditable business rules in a language any developer can read.
Glossary of the new pieces
- Business logic gate: plain Python that accepts extracted data and gives back either
(True, cleaned_data)or(False, error_reason). It has its own file free of LangGraph imports and executes in a node ahead of any tool call. with_structured_output(PydanticModel): makes the model return data matching a Pydantic schema and not free text. The schema is checked before the gate even runs, so structured output plus a gate gives two layers of validation.extracted_data: dict: the state field where structured model output lands before gates examine it. Tools read from here and never from the model's raw output.validation_errors: Annotated[list[str], operator.add]: the audit trail. Each gate failure appends to it, and it accumulates across the whole conversation.retry_count: int: protection against infinite loops. The correction node increments it, the gate's routing function checks it, and exceeding the maximum means escalation and not another attempt.correction_node: the recovery node. It reads the latest validation error, builds a specific correction instruction, re-invokes the model with structured output and routes back to the gate. All three specialists share it.gate_status: str: the channel between gate nodes and routing functions, with values "pending", "passed", "failed", "escalated" and "aborted". Routing functions read it instead of parsing model output.supervisor_node: reads conversation state and returns a specialist's name. It never calls tools and never acts.- A compiled subgraph used as a node: added with
graph_builder.add_node("name", compiled_subgraph). The parent sees a black box, while the subgraph handles its own routing, gate logic and tool calls.
Closing the gap
The difficult parts of enterprise agent work lie elsewhere than the first graph layout: deterministic error handling, state validation at scale, and an orchestration layer that isn't tied to one model vendor. All three get an answer here. Validation is deterministic. A failure triggers a correction cycle and not a crash. Every decision can be inspected through the audit trail. And nothing in the orchestration layer, whether gates, routing functions, supervisor logic or the correction node, imports anything vendor-specific. Change the configured provider from ChatOpenAI to ChatAnthropic, or to anything else init_chat_model supports, and each gate, route and business rule runs exactly as before. The model is a pluggable module, and the architecture is not.