feat(examples): rewrite LangGraph integration with customer support domain

[sanjeed5]

Example Update: langgraph

File: python-sdk/examples/integrations/langgraph_integration.py
Framework version: langgraph 1.0.10
Status: Updated

Summary

• Rewrote the LangGraph example from scratch to match the customer support domain used by other integration examples (ADK, PydanticAI)
• Upgraded from deprecated create_react_agent (langgraph.prebuilt) to create_agent (langchain.agents)
• Added three comprehensive scenarios showcasing key LangGraph patterns

Patterns Covered

• Agent names/hierarchy: yes (support_agent, order_specialist, policy_specialist, support_agent_with_memory)
• System instructions: yes (each agent has explicit system prompts)
• User input/model output: yes
• Tool calls (args + results): yes (lookup_order_status, lookup_policy with ORD-1001 through ORD-1003)
• Session continuity: yes (Scenario 3 uses MemorySaver checkpointer with thread_id)
• Multi-agent delegation: yes (Scenario 2: coordinator routes to specialist subgraphs)
• Routing with structured output: yes (RouteDecision Pydantic model with conditional edges)

Changes Made

• Replaced generic calculator/knowledge-base scenarios with customer support domain
• Added Scenario 1: Single agent with tools via create_agent
• Added Scenario 2: Multi-agent delegation with coordinator StateGraph + specialist subgraph agents
• Added Scenario 3: Multi-turn conversation with MemorySaver checkpointing
• Switched imports from langchain.messages to langchain_core.messages
• Added langchain to install deps (required for create_agent)
• Wrapped all execution in main() function with proper force_flush()

Tracing Issues Found

• None found. 60 events captured correctly: 1 session, 38 chain, 15 model, 6 tool. All agent names, tool calls, and hierarchy properly traced.

Links

• UI trace: https://fe.testing-cp.honeyhive.ai/datastore/sessions/96426abe-b2b0-4488-80bd-c40fabead802/peQ9-_5k2zXezKJ_cwO20KYx

---

[snyk-io]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scanner	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

[claude]

Claude finished @sanjeed5's task in 3m 6s —— View job

---

• Read PR diff and changed files
• Analyze code for issues, best practices, and patterns
• Post inline comments for specific feedback
• Post top-level review summary

Result

Reviewed the LangGraph integration rewrite. This is a clean, well-structured update that properly migrates to the new create_agent API from LangChain 1.0 and aligns the example with the customer support domain used by other integration examples. No blocking issues found — left two minor inline nits on type annotations and a positive note on the os.getenv improvement.

[github-actions]

📚 Documentation preview built — Download artifact

Review instructions & validation status

How to Review

1. Download the artifact from the link above
2. Extract the files
3. Open index.html in your browser

Validation Status

• API validation: ✅ Passed
• Build process: ✅ Successful
• Import tests: ✅ All imports working

[claude]

Review Summary

Overall: Looks great — this is a clean, well-structured rewrite that significantly improves the LangGraph example.

What looks good:

• Clean migration from deprecated create_react_agent (langgraph.prebuilt) to the new create_agent (langchain.agents), aligned with LangChain 1.0
• The customer support domain is consistent with the PydanticAI and ADK examples — same lookup_order_status / lookup_policy tools and mock data
• Proper main() entry point with if __name__ == "__main__" guard (previously ran at module level)
• try/finally ensures force_flush() is called even on error — good pattern for examples
• Switch from os.environ[] (crash on missing key) to os.getenv() (graceful None) follows the SDK's graceful degradation principle
• Three scenarios demonstrate progressively more advanced patterns: single agent → multi-agent delegation → stateful multi-turn

No blocking issues found. Two minor inline nits left on the diff.

[github-actions]

📚 Documentation preview built — Download artifact

Review instructions & validation status

How to Review

1. Download the artifact from the link above
2. Extract the files
3. Open index.html in your browser

Validation Status

• API validation: ✅ Passed
• Build process: ✅ Successful
• Import tests: ✅ All imports working

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 3 additional findings.