Skip to main content

Configuring an Agent

After creating an agent, you need to configure its behavior to match your business needs.

Accessing Configuration

  1. Go to Agents in the sidebar
  2. Click on the agent you want to configure
  3. Select the Configuration tab

Configuration Options

Safety Level

The safety level controls how cautious your agent is when responding:

LevelDescriptionBest For
LowMore creative, less restrictive responsesInformal conversations
MediumBalanced approach (recommended)Most use cases
HighConservative, fact-focused responsesSensitive industries
Recommendation

Start with Medium and adjust based on customer feedback and conversation quality.

Conversation Turns

Set the maximum number of back-and-forth exchanges before the agent suggests human handoff:

  • Minimum: 3 turns
  • Recommended: 10-15 turns
  • Maximum: 50 turns

This prevents conversations from running indefinitely and ensures complex queries get human attention.

Response Language

Configure which languages your agent can respond in:

  • English (default)
  • Spanish
  • Arabic

The agent can detect the customer's language and respond accordingly if multiple languages are enabled.

Knowledge Base Connection

Link your agent to specific datasets from your Knowledge Base:

  1. Click Select Datasets
  2. Choose which document collections the agent should use
  3. Click Save

The agent will only reference documents from selected datasets when answering questions.

Advanced Settings

Workflow Backend

PropPilot uses a workflow system to process messages. Advanced users can configure:

SettingDescription
Workflow URLCustom endpoint for message processing
TimeoutMaximum wait time for responses (default: 30s)
Default Configuration

Most users should leave these at default values. Only modify if you have a custom n8n or automation setup.

Rate Limiting

Each agent has built-in rate limiting to prevent abuse:

  • Default: 10 requests per second per tenant
  • Prevents excessive API calls
  • Protects against spam attacks

Testing Your Agent

Before going live, test your agent:

  1. Enable the agent on a test channel
  2. Send various types of messages
  3. Check responses for accuracy and tone
  4. Verify knowledge base information is correct
  5. Test edge cases and unusual questions

Test Scenarios

Try these scenarios:

  • ✅ Basic property inquiry
  • ✅ Pricing questions
  • ✅ Appointment scheduling request
  • ✅ Off-topic question (should redirect gracefully)
  • ✅ Multiple questions in one message
  • ✅ Messages in different languages

Saving Changes

After making configuration changes:

  1. Review all settings
  2. Click Save Changes
  3. Wait for confirmation toast

Changes take effect immediately for new conversations. Ongoing conversations may need to restart to see updates.

Best Practices

Do's

  • ✅ Keep knowledge base documents up to date
  • ✅ Use medium safety level to start
  • ✅ Set reasonable conversation turn limits
  • ✅ Test thoroughly before activating
  • ✅ Monitor conversations regularly

Don'ts

  • ❌ Set safety level too low without testing
  • ❌ Connect agents to irrelevant datasets
  • ❌ Activate without testing
  • ❌ Ignore conversation logs

What's Next?

Configure which channels your agent responds on:

Agent Channels