Creating Your Agent

To begin building an AI agent in Shipable:

  1. Navigate to the Playground section in your Shipable dashboard
  2. Click on “New Project” or select an existing draft project
  3. You’ll be taken to the agent configuration screen

Configuring Your Agent

Step 1: Select an AI Model

The first step is choosing which AI model will power your agent:

  1. From the dropdown menu, select your preferred model (e.g., OpenAI/GPT-4-turbo, Claude)
  2. For models requiring API keys, enter your key in the provided field

Step 2: Define Agent Instructions

Instructions are crucial for shaping how your agent behaves:

  1. Choose a preset persona (Financial Analyst, Legal Assistant, Travel Guide, or Personal Companion) or create your own
  2. Write clear instructions in the editor, including:
    • The agent’s role and purpose
    • How it should respond to queries
    • Any specific guidelines for tone and style
    • Limitations or boundaries for responses

Step 3: Add Knowledge Base

Power your agent with relevant data:

  1. Click “Add Data” in the Knowledge Base section
  2. Choose your data source:
    • Local File: Upload documents (.pdf, .doc, .docx, .txt)
    • Website: Connect to web content
    • Q&A: Create custom question-answer pairs
    • Integrations: Connect with Google Drive, Notion, and other platforms
  3. For uploaded files, Nbula will automatically process and tokenize the content
  4. You can view all uploaded data and token usage in the All Data section

Step 4: Connect Tools

Enhance your agent’s capabilities with tools:

  1. Click “Add Tool” to browse available tools
  2. Select tools relevant to your use case:
    • Jira: Search and manage tasks
    • Gmail: Search and send emails
    • Web Content: Retrieve information from websites
    • Github: Access repositories and code
  3. For each tool, complete the required authentication steps
  4. Configure tool-specific settings (e.g., for Gmail, specify allowed email addresses)

Step 5: Enable Additional Features

Customize your agent with additional capabilities:

  1. Voice Chat: Toggle on to allow users to interact with your agent via voice
  2. Guardrail: Enable to ensure your agent strictly follows its instructions
  3. Collect User Information: Configure what user data to collect (name, email, phone number)