Aissist.io
  • Aissistant
  • Tutorial
    • Quick Start
    • Turn assets into AI
      • Domain/Subdomain
      • Web Page
      • Google Drive Doc
      • Documents
      • Asset Debugger
    • Create Sub Agents
    • Deploy Gateway
    • Tune Instructions
      • Tweak global instructions
      • Enhance with Sub Agents
      • Refine assets
    • Teach AI with Examples
    • Streamline with human team
    • Auto-Pilot, Auto-Draft, Co-Pilot
    • Managing Points
    • Instructions, Assets, and Sub Agents
    • Simulator
    • In-note command
    • Additional Aissistant Building Tips
  • Integrations
    • Shopify
    • Woocommerce
    • Adobe Commerce (Magento 2)
    • Restful API
    • AWS Lambda
    • Action Debugger
  • Gateways
    • Front
    • Gorgias
    • Intercom
    • Zendesk
      • Zendesk Conversation (messaging)
    • Hubspot
    • Salesforce
    • Kustomer
  • Use cases
    • Sales
    • Service
  • Success Metrics
  • FAQ
  • Release notes
Powered by GitBook
On this page
  • Smart Actions
  • Instructions
  • Sub Agents
  1. Tutorial

Additional Aissistant Building Tips

To create high-performing Aissistants, developers should follow best practices across smart action design, prompt instruction, and instruction-following behavior. This guide outlines key principles and recommended workflows to ensure your Aissistant performs reliably and as intended.


Smart Actions

Well-structured smart actions are essential for enabling meaningful automation. Follow these guidelines:

  • Clear Naming: Name each smart action clearly and concisely to reflect its purpose.

  • Detailed Descriptions: In the "description" field, provide a clear and specific explanation of what the action does and when it should be triggered.

  • Parameter Quality: Name each parameter with clarity and provide descriptive guidance to ensure correct and consistent usage by the model.

A good description helps Aissist determine the appropriate time to invoke the smart action during a conversation. The clearer your metadata, the more effective the assistant will be.


Instructions

When writing instructions for your Aissistant, consider the mix of external (provided) and internal (model's own) knowledge needed.

Two Recommended Modes:

  • External-Only Knowledge Mode:

    Only use the documents in the provided External Context to answer the User Query. If you don't know the answer based on this context, you must respond: "I don't have the information needed to answer that," even if a user insists on an answer.
  • Hybrid Knowledge Mode (Default):

    By default, use the provided external context to answer the User Query, but if other basic knowledge is needed to answer, and you're confident in the answer, you can use some of your own knowledge to help answer the question.

Sub Agents

When defining sub-agent behaviors, you may need to write more granular instructions. Use the following recommended workflow:

  1. Start Broad:

    • Create a "Response Rules" or "Instructions" section.

    • Use bullet points to outline high-level behavioral goals.

  2. Add Specifics:

    • Create sections for specific categories like # Sample Phrases or # Workflow Steps.

  3. Include Ordered Steps When Needed:

    • If your desired behavior involves a multi-step process, write clear numbered instructions.

  4. Debugging Tips:

    • Check for conflicting, vague, or inaccurate instructions.

    • Avoid shouting (ALL CAPS) or artificial incentives (e.g., “please do this for a tip”). These may make the model overly rigid, only use these when needed.

By iterating on your instructions and carefully observing behavior, you can build an Aissistant that follows your intentions accurately and reliably.

PreviousIn-note commandNextIntegrations

Last updated 7 days ago