Skip to content

QUICK_BEDROCK_SETUP.md

Latest commit

 

History

History
189 lines (131 loc) · 5.23 KB

QUICK_BEDROCK_SETUP.md

File metadata and controls

189 lines (131 loc) · 5.23 KB

🚀 Quick AWS Bedrock Agent Setup

Let's get your real Bedrock Agent running in 10 minutes!

✅ Prerequisites (Already Done!)

  • ✅ AWS credentials configured
  • ✅ Region: eu-west-2
  • ✅ AWS SDK installed

🎯 Step-by-Step Setup

Step 1: Enable Claude 3 Sonnet (2 minutes)

  1. Open AWS Bedrock Console:

    https://eu-west-2.console.aws.amazon.com/bedrock/home?region=eu-west-2#/modelaccess

  2. Click "Enable specific models"

  3. Find "Anthropic Claude 3 Sonnet" and check the box

  4. Click "Submit" - Usually approved instantly

  5. Wait for status to show "Access granted"

Step 2: Create IAM Role (3 minutes)

  1. Go to IAM Console:

    https://console.aws.amazon.com/iam/home?region=eu-west-2#/roles

  2. Click "Create role"

  3. Select "AWS service" → "Bedrock" → "Bedrock - Agent"

  4. Click "Next" (permissions are auto-attached)

  5. Role name: AmazonBedrockExecutionRoleForAgents_shopping

  6. Click "Create role"

  7. Copy the Role ARN (you'll need this)

Step 3: Create Bedrock Agent (5 minutes)

  1. Go to Bedrock Agents Console:

    https://eu-west-2.console.aws.amazon.com/bedrock/home?region=eu-west-2#/agents

  2. Click "Create Agent"

  3. Fill in details:

    • Agent name: autonomous-shopping-agent
    • Description: AI agent for autonomous shopping
    • Agent resource role: Select the role you just created
    • Foundation model: Anthropic Claude 3 Sonnet

  4. Agent instructions (copy this):

    You are an autonomous AI shopping assistant for a sustainable blockchain marketplace. 

CAPABILITIES:
- Search products using natural language queries
- Provide detailed product information and sustainability scores
- Create orders automatically when users request
- Process payments through blockchain smart contracts
- Track order status and handle customer service

PERSONALITY:
- Friendly, helpful, and knowledgeable about sustainability
- Proactive in suggesting eco-friendly alternatives
- Clear about pricing, fees, and transaction details

WORKFLOW:
1. SEARCH: When users ask about products, use searchProducts function
2. DETAILS: Provide comprehensive product information
3. ORDER: When users want to buy, use createOrder function
4. PAYMENT: Process payments using processPayment function
5. FOLLOW-UP: Check order status and provide updates

RULES:
- Always confirm details before processing payments
- Explain sustainability scores and certifications
- Be transparent about blockchain transaction fees
- Prioritize products with high sustainability scores (80%+)

  5. Click "Create"

Step 4: Add Action Group (Skip for Now)

For the hackathon demo, we'll use the agent without action groups first. The agent will work with natural language responses.

Step 5: Create Agent Alias

  1. In your agent, click "Create Alias"

  2. Alias name: production

  3. Description: Production alias for shopping agent

  4. Click "Create"

Step 6: Update Environment Variables

  1. Copy your Agent ID (from the agent details page)

  2. Copy your Agent Alias ID (from the alias you just created)

  3. Update your .env file:

    BEDROCK_AGENT_ID=ABCD123456
BEDROCK_AGENT_ALIAS_ID=EFGH789012

🧪 Test Your Setup

Option 1: Use the Checker Script

node scripts/check-bedrock-setup.js

Option 2: Test in Browser

  1. Go to http://localhost:3001/marketplace
  2. Click "Shopping Assistant" → "Autonomous Agent"
  3. Try: "Find sustainable electronics under $100"

🎯 Expected Results

✅ Working: Agent responds with natural language ✅ Working: Agent understands shopping requests ✅ Working: Agent provides helpful responses

❌ Not yet: Function calling (requires Lambda setup) ❌ Not yet: Real order processing (requires action groups)

🚀 Quick Test Commands

Test the agent directly:

# Test 1: Basic conversation
curl -X POST http://localhost:3001/api/ai/autonomous-agent \
  -H "Content-Type: application/json" \
  -d '{"message": "Hello, can you help me find products?"}'

# Test 2: Product search
curl -X POST http://localhost:3001/api/ai/autonomous-agent \
  -H "Content-Type: application/json" \
  -d '{"message": "Find sustainable electronics under $100"}'

🔧 Troubleshooting

"Model access denied"

  • Go back to Step 1 and enable Claude 3 Sonnet

"Agent not found"

  • Check your Agent ID in .env file
  • Make sure you're in the eu-west-2 region

"Permission denied"

  • Check your IAM role has Bedrock permissions
  • Verify your AWS credentials

"Still using mock responses"

  • Restart your Next.js server after updating .env
  • Check the console logs for "Real Bedrock Agent" vs "Using mock responses"

🎉 Success Indicators

When working correctly, you'll see in the console:

🤖 Real Bedrock Agent: Processing "your message"
✅ Real Agent Response: {...}

Instead of:

🎭 Using mock responses (Bedrock Agent not configured)

Ready to start? Begin with Step 1 and work through each step! 🚀

The whole process should take about 10 minutes, and you'll have a real AWS Bedrock Agent powering your autonomous shopping assistant!