Agents

Create intelligent AI agents that can autonomously execute complex multi-step tasks.

Overview

DYPAI Agents are AI-powered automation tools that can:

Understand natural language instructions
Execute multi-step workflows
Access your database, storage, and APIs
Make decisions based on data
Maintain context across interactions

Creating an Agent

Navigate to AI Agents

Go to the AI → Agents section in your dashboard.

Create New Agent

Click "New Agent" and configure the agent settings.

Define Instructions

Write clear instructions that define the agent's behavior.

Select Tools

Choose which tools the agent can use.

Test and Deploy

Test the agent with sample inputs, then deploy.

Agent Configuration

{
  "name": "sales-analyst",
  "description": "Analyzes sales data and generates reports",
  "instructions": """
    You are a sales data analyst. Your job is to:
    1. Query the sales database for relevant data
    2. Analyze trends and patterns
    3. Generate clear, actionable reports
    4. Answer questions about sales performance

    Always provide data-driven insights and cite specific numbers.
  """,
  "tools": [
    "database_query",
    "database_schema",
    "edge_function"
  ],
  "model": "claude-sonnet-4-20250514",
  "max_tokens": 4096
}

Available Tools

Tool	Description
`database_query`	Execute SQL queries
`database_insert`	Insert records
`database_update`	Update records
`storage_read`	Read files
`storage_write`	Write files
`api_call`	Call external APIs
`edge_function`	Run custom functions
`send_email`	Send emails

Running Agents

Via API

// JavaScript/TypeScript
const response = await fetch('https://your-project.dypai.io/ai/v1/agents/sales-analyst/run', {
  method: 'POST',
  headers: {
    'Content-Type': 'application/json',
    'Authorization': 'Bearer ' + apiKey
  },
  body: JSON.stringify({
    input: "Generate a sales report for Q4 2024",
    context: {
      user_id: "user-123",
      timezone: "America/New_York"
    }
  })
});

const result = await response.json();
console.log(result.output);
console.log(result.steps); // Steps the agent took

Via Dashboard

Go to AI → Agents
Select your agent
Enter a prompt in the test interface
View the agent's response and execution steps

Agent Examples

Customer Support Agent

{
  "name": "support-agent",
  "instructions": """
    You are a customer support assistant. Help users by:
    1. Looking up their account information
    2. Checking order status
    3. Answering product questions
    4. Escalating complex issues to human support

    Be friendly, helpful, and professional.
    Always verify the user's identity before sharing account details.
  """,
  "tools": [
    "database_query",
    "api_call",
    "send_email"
  ]
}

Data Processing Agent

{
  "name": "data-processor",
  "instructions": """
    You process incoming data files and load them into the database.
    1. Read the uploaded file from storage
    2. Validate the data format
    3. Transform data as needed
    4. Insert into the appropriate table
    5. Generate a summary report
  """,
  "tools": [
    "storage_read",
    "database_insert",
    "database_query",
    "edge_function"
  ]
}

Content Moderator Agent

{
  "name": "content-moderator",
  "instructions": """
    Review user-generated content for policy violations.
    1. Query new content from the posts table
    2. Analyze each post for violations
    3. Flag or remove violating content
    4. Update the moderation status
    5. Log all decisions for review
  """,
  "tools": [
    "database_query",
    "database_update"
  ]
}

Creating Custom Tools

Define custom tools using Edge Functions:

// Edge Function: calculate-shipping
async def handler(request):
    body = await request.json()
    weight = body.get('weight')
    destination = body.get('destination')

    # Calculate shipping cost
    base_rate = 5.99
    per_kg = 2.50
    cost = base_rate + (weight * per_kg)

    return {
        "shipping_cost": cost,
        "estimated_days": 3,
        "carrier": "Standard Shipping"
    }

# Register as tool
tool_config = {
    "name": "calculate_shipping",
    "description": "Calculate shipping cost for an order",
    "parameters": {
        "weight": {"type": "number", "description": "Weight in kg"},
        "destination": {"type": "string", "description": "Destination country"}
    },
    "function": "calculate-shipping"
}

Execution History

View all agent executions in the dashboard:

Input prompts and outputs
Steps taken during execution
Tools used and their results
Execution time and tokens used
Errors and debugging information

// Query execution history via API
const response = await fetch('https://your-project.dypai.io/ai/v1/agents/sales-analyst/history', {
  headers: {
    'Authorization': 'Bearer ' + apiKey
  }
});

const history = await response.json();
history.executions.forEach(exec => {
  console.log(exec.id, exec.status, exec.created_at);
});

Best Practices

Write clear, specific instructions
Only grant tools the agent needs
Test thoroughly before production
Monitor execution history for issues
Set appropriate token limits
Use context to provide relevant information

Start with simple tasks and gradually add complexity as you understand the agent's behavior.

Next steps