Skip to content

DEPLOYMENT_SETUP.md

Latest commit

 

History

History
98 lines (68 loc) · 2.59 KB

DEPLOYMENT_SETUP.md

File metadata and controls

98 lines (68 loc) · 2.59 KB
Raw

🚀 WORKING LangGraph Deployment Setup

⚡ IMMEDIATE STEPS - DO THIS NOW

1. Create .env file in project root:

cp ".env copy" .env

2. Edit .env with YOUR deployment details:

LANGGRAPH_URL=https://your-actual-deployment-url.langchain.com
LANGCHAIN_API_KEY=lsv2_pt_your-actual-api-key-here
LANGSMITH_API_KEY=lsv2_pt_your-actual-api-key-here
LANGCHAIN_TRACING_V2=true
OPENAI_API_KEY=sk-your-openai-key

3. Create web/.env.local:

NEXT_PUBLIC_COPILOT_API_KEY=
NEXT_PUBLIC_COPILOTKIT_RUNTIME_URL=https://your-actual-deployment-url.langchain.com
NEXT_PUBLIC_COPILOTKIT_AGENT_NAME=agent

4. Get Your Real Values:

From LangSmith (https://smith.langchain.com/):

  • Go to Settings → API Keys
  • Copy key starting with lsv2_pt_

From LangGraph Platform:

  • Copy your deployment URL (ends with .langchain.com)

5. Run:

cd web
npm run dev

⚠️ CRITICAL NOTES:

  • API key MUST start with lsv2_pt_
  • URL MUST be your exact deployment URL
  • Both .env files MUST have the SAME deployment URL

🧪 TEST YOUR CONNECTION (DO THIS FIRST!)

node test-connection.js

This will verify your deployment URL and API key work BEFORE starting the frontend.

🚀 If Test Passes, Run Frontend:

Once configured, run your frontend:

# Install dependencies (if not already done)
cd web && npm install

# Start the development server
npm run dev

Your app will run at http://localhost:3000 and connect to your deployed LangGraph agent!

🔄 How It Works

  1. User Input: User types in the CopilotKit chat interface
  2. API Route: Request goes to /api/copilotkit (route.ts)
  3. LangGraph SDK: Direct connection to your deployed agent using @langchain/langgraph-sdk
  4. Streaming Response: Real-time streaming of agent responses back to the UI

🛠️ Key Changes Made

  • Replaced CopilotKit LangGraphAgent: Now using direct SDK connection
  • Minimal Code Changes: Only route.ts modified for maximum simplicity
  • Environment-Based: Easy switching between local and deployed agents

📝 Notes

  • Your deployed agent on LangGraph Platform is fully managed
  • All tools and capabilities from your agents.py are available
  • Frontend runs locally but connects to your cloud deployment
  • Perfect for development while using production agent infrastructure

🔧 Troubleshooting

  1. Connection Issues: Verify LANGGRAPH_URL is correct and accessible
  2. Auth Errors: Check LANGCHAIN_API_KEY is valid
  3. Agent Not Found: Ensure agent name "agent" matches your deployment

Your deployment is ready! 🎉