> ## Documentation Index
> Fetch the complete documentation index at: https://docs.portkey.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# LangGraph
> Use Prisma AIRS AI Gateway with LangGraph to take your AI agent workflows to production
## Introduction
LangGraph is a library for building stateful, multi-actor applications with LLMs, designed to make developing complex agent workflows easier. It provides a flexible framework to create directed graphs where nodes process information and edges define the flow between them.
The AI Gateway enhances LangGraph with production-readiness features, turning your experimental agent workflows into robust systems by providing:
* **Complete observability** of every agent step, tool use, and state transition
* **Built-in reliability** with fallbacks, retries, and load balancing
* **Cost tracking and optimization** to manage your AI spend
* **Access to 3,000+ LLMs** through a single integration
* **Guardrails** to keep agent behavior safe and compliant
* **Version-controlled prompts** for consistent agent performance
Learn more about LangGraph's core concepts and features
### Installation & Setup
```bash theme={"system"}
pip install -U langgraph langchain langchain_openai
```
Depending on your use case, you may also need additional packages:
* For search capabilities: `pip install langchain_community`
* For memory functionality: `pip install langgraph[checkpoint]`
Create an AI Gateway API key with optional budget/rate limits from the [Strata Cloud Manager](https://stratacloudmanager.paloaltonetworks.com/). You can attach configurations for reliability, caching, and more to this key.
For a simple setup, configure a LangChain ChatOpenAI instance to use the AI Gateway:
## Basic Agent Implementation
Let's create a simple LangGraph chatbot using the AI Gateway. This example shows how to set up a basic conversational agent:
```python theme={"system"}
from typing import Annotated
from langchain_openai import ChatOpenAI
from typing_extensions import TypedDict
from langgraph.graph import StateGraph
from langgraph.graph.message import add_messages
# Define state structure with message history
class State(TypedDict):
messages: Annotated[list, add_messages]
# Initialize graph builder
graph_builder = StateGraph(State)
# Set up LLM with the AI Gateway
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY",
base_url="https://aigw.portkey.ai/v1",
default_headers={
"x-portkey-provider": "@YOUR_PROVIDER",
"x-portkey-trace-id": "chat-session-123", # Optional
}
)
# Define chatbot node function
def chatbot(state: State):
return {"messages": [llm.invoke(state["messages"])]}
# Add node to graph and set entry/exit points
graph_builder.add_node("chatbot", chatbot)
graph_builder.set_entry_point("chatbot")
graph_builder.set_finish_point("chatbot")
graph = graph_builder.compile()
# Function to handle streaming updates
def stream_graph_updates(user_input: str):
for event in graph.stream({"messages": [{"role": "user", "content": user_input}]}):
for value in event.values():
print("Assistant:", value["messages"][-1].content)
# Interactive chat loop
while True:
try:
user_input = input("User: ")
if user_input.lower() in ["quit", "exit", "q"]:
print("Goodbye!")
break
stream_graph_updates(user_input)
except Exception as e:
print(f"Error: {e}")
break
```
This basic implementation:
1. Creates a state graph with a message history
2. Configures a ChatOpenAI model with the AI Gateway
3. Defines a simple chatbot node that processes messages with the LLM
4. Compiles the graph and provides a streaming interface for chat
## Advanced Features
### 1. Adding Tools to Your Agent
LangGraph can be enhanced with tools to allow your agent to perform actions. Here's how to add the Tavily search tool:
```python [expandable] theme={"system"}
from langgraph.prebuilt import ToolNode, tools_condition
from typing import Annotated
from langchain_openai import ChatOpenAI
from typing_extensions import TypedDict
from langgraph.graph import StateGraph
from langgraph.graph.message import add_messages
from langchain_community.tools.tavily_search import TavilySearchResults
class State(TypedDict):
messages: Annotated[list, add_messages]
graph_builder = StateGraph(State)
# Initialize the Tavily search tool
# Note: Requires TAVILY_API_KEY environment variable or passed as argument
tool = TavilySearchResults(max_results=2)
tools = [tool]
# Set up LLM with the AI Gateway
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY",
base_url="https://aigw.portkey.ai/v1",
default_headers={
"x-portkey-provider": "@YOUR_PROVIDER",
"x-portkey-trace-id": "search-agent-session", # Optional
}
)
# Bind tools to the LLM
llm_with_tools = llm.bind_tools(tools)
def chatbot(state: State):
return {"messages": [llm_with_tools.invoke(state["messages"])]}
graph_builder.add_node("chatbot", chatbot)
# Add tool node to handle tool execution
tool_node = ToolNode(tools=[tool])
graph_builder.add_node("tools", tool_node)
# Add conditional routing based on tool usage
graph_builder.add_conditional_edges(
"chatbot",
tools_condition,
)
# Return to chatbot after tool execution
graph_builder.add_edge("tools", "chatbot")
graph_builder.set_entry_point("chatbot")
graph = graph_builder.compile()
```
This example requires a Tavily API key for the search functionality. You can sign up for one at [Tavily's website](https://tavily.com/).
### 2. Creating Custom Tools
You can create custom tools for your agents using the `@tool` decorator. Here's how to create a simple multiplication tool:
```python [expandable] theme={"system"}
from langchain_core.tools import tool
from pydantic import BaseModel, Field
from langgraph.prebuilt import ToolNode, tools_condition
from typing import Annotated
from langchain_openai import ChatOpenAI
from typing_extensions import TypedDict
from langgraph.graph import StateGraph
from langgraph.graph.message import add_messages
class State(TypedDict):
messages: Annotated[list, add_messages]
# Define input schema for the tool
class MultiplyInputSchema(BaseModel):
"""Multiply two numbers"""
a: int = Field(description="First operand")
b: int = Field(description="Second operand")
# Create the tool
@tool("multiply_tool", args_schema=MultiplyInputSchema)
def multiply(a: int, b: int) -> int:
return a * b
graph_builder = StateGraph(State)
tools = [multiply]
# Set up LLM with the AI Gateway
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY",
base_url="https://aigw.portkey.ai/v1",
default_headers={"x-portkey-provider": "@YOUR_PROVIDER"}
)
llm_with_tools = llm.bind_tools(tools)
def chatbot(state: State):
return {"messages": [llm_with_tools.invoke(state["messages"])]}
graph_builder.add_node("chatbot", chatbot)
tool_node = ToolNode(tools=tools)
graph_builder.add_node("tools", tool_node)
graph_builder.add_conditional_edges(
"chatbot",
tools_condition,
)
graph_builder.add_edge("tools", "chatbot")
graph_builder.set_entry_point("chatbot")
graph = graph_builder.compile()
```
This example:
1. Defines a Pydantic model for the tool's input schema
2. Creates a custom multiplication tool with the `@tool` decorator
3. Integrates it into LangGraph with a tool node
### 3. Adding Memory to Your Agent
For persistent conversations, you can add memory to your LangGraph agents:
```python [expandable] theme={"system"}
from typing import Annotated
from langchain_openai import ChatOpenAI
from langchain_community.tools.tavily_search import TavilySearchResults
from typing_extensions import TypedDict
from langgraph.prebuilt import ToolNode, tools_condition
from langgraph.checkpoint.memory import MemorySaver
from langgraph.graph import StateGraph
from langgraph.graph.message import add_messages
class State(TypedDict):
messages: Annotated[list, add_messages]
graph_builder = StateGraph(State)
# Set up tools and LLM with the AI Gateway
tool = TavilySearchResults(max_results=2)
tools = [tool]
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY",
base_url="https://aigw.portkey.ai/v1",
default_headers={
"x-portkey-provider": "@YOUR_PROVIDER",
"x-portkey-trace-id": "memory-agent-session", # Optional
}
)
llm_with_tools = llm.bind_tools(tools)
def chatbot(state: State):
return {"messages": [llm_with_tools.invoke(state["messages"])]}
graph_builder.add_node("chatbot", chatbot)
tool_node = ToolNode(tools=[tool])
graph_builder.add_node("tools", tool_node)
graph_builder.add_conditional_edges(
"chatbot",
tools_condition,
)
graph_builder.add_edge("tools", "chatbot")
graph_builder.set_entry_point("chatbot")
# Initialize memory saver for persistent state
memory = MemorySaver()
graph = graph_builder.compile(checkpointer=memory)
# Configuration for memory thread id
config = {"configurable": {"thread_id": "1"}}
# Function to handle streaming updates with memory
def stream_graph_updates(user_input: str):
events = graph.stream(
{"messages": [{"role": "user", "content": user_input}]},
config,
stream_mode="values",
)
for event in events:
if "messages" in event:
# Use pretty_print method for better formatting
event["messages"][-1].pretty_print()
```
The `thread_id` in the config allows you to maintain separate conversation threads for different users or contexts.
## Production Features
### 1. Enhanced Observability
The AI Gateway provides comprehensive observability for your LangGraph agents, helping you understand exactly what's happening during each execution.
Traces provide a hierarchical view of your agent's execution, showing the sequence of LLM calls, tool invocations, and state transitions.
```python theme={"system"}
import json
# Add trace_id to enable hierarchical tracing in the AI Gateway
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY",
base_url="https://aigw.portkey.ai/v1",
default_headers={
"x-portkey-provider": "@YOUR_LLM_PROVIDER",
# Add unique trace ID
"x-portkey-trace-id": "unique-session-id",
"x-portkey-metadata": json.dumps({"request_type": "user_query"}),
}
)
```
LangGraph also offers its own tracing via LangSmith, which can be used alongside the AI Gateway for even more detailed workflow insights.
AI Gateway logs every interaction with LLMs, including:
* Complete request and response payloads
* Latency and token usage metrics
* Cost calculations
* Tool calls and function executions
All logs can be filtered by metadata, trace IDs, models, and more, making it easy to debug specific agent runs.
The AI Gateway provides built-in dashboards that help you:
* Track cost and token usage across all agent runs
* Analyze performance metrics like latency and success rates
* Identify bottlenecks in your agent workflows
* Compare different agent configurations and LLMs
You can filter and segment all metrics by custom metadata to analyze specific agent types, user groups, or use cases.
Add custom metadata to your LangGraph agent calls to enable powerful filtering and segmentation:
```python theme={"system"}
import json
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY",
base_url="https://aigw.portkey.ai/v1",
default_headers={
"x-portkey-provider": "@YOUR_LLM_PROVIDER",
"x-portkey-metadata": json.dumps({
"agent_type": "search_agent",
"environment": "production",
"_user": "user_123", # Special _user field for user analytics
"graph_id": "complex_workflow"
}),
}
)
```
This metadata can be used to filter logs, traces, and metrics on Strata Cloud Manager, allowing you to analyze specific agent runs, users, or environments.
### 2. Reliability - Keep Your Agents Running Smoothly
When running agents in production, things can go wrong - API rate limits, network issues, or provider outages. The AI Gateway's reliability features ensure your agents keep running smoothly even when problems occur.
Enable fallback in your LangGraph agents by using an AI Gateway Config:
This configuration will automatically try Claude if the GPT-4o request fails, ensuring your agent can continue operating.
Handles temporary failures automatically. If an LLM call fails, the AI Gateway will retry the same request for the specified number of times - perfect for rate limits or network blips.
Prevent your agents from hanging. Set timeouts to ensure you get responses (or can fail gracefully) within your required timeframes.
Send different requests to different providers. Route complex reasoning to GPT-4, creative tasks to Claude, and quick responses to Gemini based on your needs.
Keep running even if your primary provider fails. Automatically switch to backup providers to maintain availability.
Spread requests across multiple API keys or providers. Great for high-volume agent operations and staying within rate limits.
### 3. Prompting in LangGraph
The AI Gateway's Prompt Engineering Studio helps you create, manage, and optimize the prompts used in your LangGraph agents. Instead of hardcoding prompts or instructions, use the AI Gateway's prompt rendering API to dynamically fetch and apply your versioned prompts.
Prompt Playground is a place to compare, test and deploy perfect prompts for your AI application. It's where you experiment with different models, test variables, compare outputs, and refine your prompt engineering strategy before deploying to production. It allows you to:
1. Iteratively develop prompts before using them in your agents
2. Test prompts with different variables and models
3. Compare outputs between different prompt versions
4. Collaborate with team members on prompt development
This visual environment makes it easier to craft effective prompts for each step in your LangGraph agent's workflow.
The Prompt Render API retrieves your prompt templates with all parameters configured:
You can:
* Create multiple versions of the same prompt
* Compare performance between versions
* Roll back to previous versions if needed
* Specify which version to use in your code:
```python theme={"system"}
# Use a specific prompt version
prompt_data = portkey_admin.prompts.render(
prompt_id="YOUR_PROMPT_ID@version_number",
variables={
"user_input": "Tell me about quantum computing"
}
)
```
AI Gateway prompts use Mustache-style templating for easy variable substitution:
```
You are an AI assistant specialized in {{agent_role}}.
User question: {{user_input}}
Please respond in a {{tone}} tone and include {{required_elements}}.
```
When rendering, simply pass the variables:
```python theme={"system"}
prompt_data = portkey_admin.prompts.render(
prompt_id="YOUR_PROMPT_ID",
variables={
"agent_role": "search navigator",
"user_input": "Find information about climate change",
"tone": "informative",
"required_elements": "recent scientific findings"
}
)
```
### 4. Guardrails for Safe Agents
Guardrails ensure your LangGraph agents operate safely and respond appropriately in all situations.
**Why Use Guardrails?**
LangGraph agents can experience various failure modes:
* Generating harmful or inappropriate content
* Leaking sensitive information like PII
* Hallucinating incorrect information
* Generating outputs in incorrect formats
The AI Gateway's guardrails add protection for both inputs and outputs.
**Implementing Guardrails**
The AI Gateway's guardrails can:
* Detect and redact PII in both inputs and outputs
* Filter harmful or inappropriate content
* Validate response formats against schemas
* Check for hallucinations against ground truth
* Apply custom business logic and rules
Explore the AI Gateway's guardrail features to enhance agent safety
### 5. User Tracking with Metadata
Track individual users through your LangGraph agents using the AI Gateway's metadata system.
**What is Metadata in the AI Gateway?**
Metadata allows you to associate custom data with each request, enabling filtering, segmentation, and analytics. The special `_user` field is specifically designed for user tracking.
**Filter Analytics by User**
With metadata in place, you can filter analytics by user and analyze performance metrics on a per-user basis:
This enables:
* Per-user cost tracking and budgeting
* Personalized user analytics
* Team or organisation-level metrics
* Environment-specific monitoring (staging vs. production)
Explore how to use custom metadata to enhance your analytics
### 6. Caching for Efficient Agents
Implement caching to make your LangGraph agents more efficient and cost-effective:
Simple caching performs exact matches on input prompts, caching identical requests to avoid redundant model executions.
Semantic caching considers the contextual similarity between input requests, caching responses for semantically similar inputs.
### 7. Model Interoperability
LangGraph works with multiple LLM providers, and the AI Gateway extends this capability by providing access to over 200 LLMs through a unified interface. You can easily switch between different models without changing your core agent logic:
The AI Gateway provides access to LLMs from providers including:
* OpenAI (GPT-4o, GPT-4 Turbo, etc.)
* Anthropic (Claude 3.5 Sonnet, Claude 3 Opus, etc.)
* Mistral AI (Mistral Large, Mistral Medium, etc.)
* Google Vertex AI (Gemini 1.5 Pro, etc.)
* Cohere (Command, Command-R, etc.)
* AWS Bedrock (Claude, Titan, etc.)
* Local/Private Models
See the full list of LLM providers supported by the AI Gateway
## Set Up Enterprise Governance for LangGraph
**Why Enterprise Governance?**
If you are using LangGraph inside your organisation, you need to consider several governance aspects:
* **Cost Management**: Controlling and tracking AI spending across teams
* **Access Control**: Managing which teams can use specific models
* **Usage Analytics**: Understanding how AI is being used across the organisation
* **Security & Compliance**: Maintaining enterprise security standards
* **Reliability**: Ensuring consistent service across all users
The AI Gateway adds a comprehensive governance layer to address these enterprise needs. Let's implement these controls step by step.
To create a new LLM integration:
Go to [Integrations](https://stratacloudmanager.paloaltonetworks.com/) in Strata Cloud Manager. Set budget / rate limits, model access if required and save the integration.
This creates a "AI Gateway Provider" that you can then use in any of your AI Gateway requests without having to send auth details for that LLM provider again.
Configs in the AI Gateway define how your requests are routed, with features like advanced routing, fallbacks, and retries.
To create your config:
1. Go to [Configs](https://stratacloudmanager.paloaltonetworks.com/) in Strata Cloud Manager
2. Create new config with:
```json theme={"system"}
{
"provider":"@YOUR_PROVIDER_FROM_STEP1",
"override_params": {
"model": "gpt-4o" // Your preferred model name
}
}
```
3. Save and note the Config name for the next step
Now create an AI Gateway API key and attach the config you created in Step 2:
1. Go to [API Keys](https://stratacloudmanager.paloaltonetworks.com/) in the AI Gateway and Create new API key
2. Select your config from `Step 2`
3. Generate and save your API key
After setting up your AI Gateway API key with the attached config, connect it to your LangGraph agents:
```python theme={"system"}
from langchain_openai import ChatOpenAI
# Configure LLM with your AI Gateway API key
llm = ChatOpenAI(
api_key="YOUR_PORTKEY_API_KEY", # The API key with attached config from step 3
base_url="https://aigw.portkey.ai/v1"
)
```
### Step 1: Implement Budget Controls & Rate Limits
Integrations enable granular control over LLM access at the team/department level. This helps you:
* Set up [budget limits](/docs/aigw/product/model-catalog/integrations#3-budget-%26-rate-limits)
* Prevent unexpected usage spikes using Rate limits
* Track departmental spending
#### Setting Up Department-Specific Controls:
1. Navigate to [Integrations](https://stratacloudmanager.paloaltonetworks.com/) in Strata Cloud Manager and create a new Integration.
2. Provision this Integration for each department with their budget limits and rate limits
3. Configure model access if required.
### Step 2: Define Model Access Rules
As your AI usage scales, controlling which teams can access specific models becomes crucial. AI Gateway Configs provide this control layer with features like:
#### Access Control Features:
* **Model Restrictions**: Limit access to specific models
* **Data Protection**: Implement guardrails for sensitive data
* **Reliability Controls**: Add fallbacks and retry logic
#### Example Configuration:
Here's a basic configuration to route requests to OpenAI, specifically using GPT-4o:
```json theme={"system"}
{
"strategy": {
"mode": "single"
},
"targets": [
{
"provider":"@YOUR_OPENAI_PROVIDER",
"override_params": {
"model": "gpt-4o"
}
}
]
}
```
Create your config on the [Configs page](https://stratacloudmanager.paloaltonetworks.com/) in Strata Cloud Manager.
Configs can be updated anytime to adjust controls without affecting running applications.
### Step 3: Implement Access Controls
Create User-specific API keys that automatically:
* Track usage per user/team with the help of metadata
* Apply appropriate configs to route requests
* Collect relevant metadata to filter logs
* Enforce access permissions
Create API keys through:
* [Strata Cloud Manager](https://stratacloudmanager.paloaltonetworks.com/)
* [API Key Management API](/docs/api-reference/admin-api/control-plane/api-keys/create-api-key)
Example using Python SDK:
For detailed key management instructions, see our [API Keys documentation](/docs/api-reference/admin-api/control-plane/api-keys/create-api-key).
### Step 4: Deploy & Monitor
After distributing API keys to your team members, your enterprise-ready LangGraph setup is ready to go. Each team member can now use their designated API keys with appropriate access levels and budget controls.
Monitor usage in Strata Cloud Manager:
* Cost tracking by department
* Model usage patterns
* Request volumes
* Error rates
### Enterprise Features Now Available
**Your LangGraph integration now has:**
* Departmental budget controls
* Model access governance
* Usage tracking & attribution
* Security guardrails
* Reliability features
## Frequently Asked Questions
The AI Gateway adds production-readiness to LangGraph through comprehensive observability (traces, logs, metrics), reliability features (fallbacks, retries, caching), and access to 3,000+ LLMs through a unified interface. This makes it easier to debug, optimize, and scale your agent workflows.
Yes! The AI Gateway integrates seamlessly with existing LangGraph applications. You just need to replace your LLM initialization code with the gateway-enabled version. The rest of your graph code remains unchanged.
The AI Gateway supports all LangGraph features, including tools, memory, conditional routing, and complex multi-node workflows. It adds observability and reliability without limiting any of the framework's functionality.
The AI Gateway allows you to add custom metadata and trace IDs to your LLM calls, which you can then use for filtering. Add fields like `graph_id`, `workflow_type`, or `session_id` to easily find and analyze specific graph executions.
Yes! The examples in this documentation show how to use LangGraph's `MemorySaver` checkpointer with gateway-enabled LLMs. All the memory and state management features work seamlessly with the AI Gateway.
## Resources