BadgerFy.ai Docs

AI Assistant

The AI Assistant is a conversational agent that enables your users to ask questions about your products and services. It leverages a knowledge base you provide to deliver concise, accurate, and brand-consistent replies through a familiar chat interface.

Overview

Easily loadable onto your site via an embed script, the AI Assistant offers extensive styling and configuration options to seamlessly integrate with your website's design and user experience.

Data Sources

The AI Assistant uses two types of data sources to provide intelligent, accurate responses:

  • Knowledge Base (Required): General information about your company, policies, FAQs, and services. This can be a .txt, .json, .pdf, or .csv file, or content from a website scrape.
  • Product Promo Data (Optional): Structured product catalog data that enables intelligent product search and comparison. This allows the assistant to answer complex questions like "What is your most expensive mattress?" or "Which products are best for back support?"
Recommended Setup: For maximum success with your AI Assistant, we recommend combining a website scrape knowledge base (Pro and Business plans) with Product Promo data. This gives your assistant comprehensive company knowledge plus intelligent product search capabilities.

Knowledge Base Tips

  • Curate Content: Remove any extraneous information that isn't directly useful for answering customer questions. A more curated dataset leads to lower token usage and more precise replies.
  • Privacy First: Ensure your dataset does not contain personal or sensitive user information. All data source uploads should be public-facing.
  • Website Scrapes: Utilizing the website scrape method is an excellent and fast way to make comprehensive information about your company available to your assistant. You can index up to 500 URLs as a single data source.
  • Automated Updates: You can use the Consumer API to keep your data source updated on a regular cadence.

Product Promo Data

Product Promo data is a structured JSON format that enables powerful product search and comparison features. When configured, the assistant can intelligently query your product catalog based on user questions. See the Product Promo Data documentation for detailed schema information and setup instructions.

  • Structured Format: Product data uses a standardized schema with fields like productName, category, price, and variants.
  • Custom Metadata: Augment your products with custom metadata fields to help differentiate and compare products. For example, add feature ratings (cooling level, firmness, back support) that help the assistant recommend the right products.
  • Intelligent Search: The assistant automatically generates optimized queries based on user questions, filtering by category, price, or product features.
  • Integration Options: Upload product data manually, connect via Shopify integration, or use the Consumer API for automated syncing.

AI Instructions

The AI Instructions field (also known as a system prompt) allows you to set the tone, style, and guardrails for your assistant. We provide a sensible default that works well for many use cases, but you can customize it to match your brand's unique voice.

Customizing AI Instructions

  1. Define Purpose: Clearly guide the assistant on its primary purposes and responsibilities.
  2. Implement Guardrails: Instruct the assistant to rely solely on its provided knowledge base and prevent it from using its own general knowledge.
  3. Specify Style and Tone: Provide directives on the desired style and tone to ensure responses align with your brand.
  4. Handle Unknowns: Include instructions on how the assistant should respond when it doesn't know the answer.
  5. Use HTML Links: The assistant supports HTML in its responses, particularly links. Instruct it to link product names or relevant pages directly. For example: "Use HTML links for products, linking the product name specifically, when you have the product URL path. Preface these with the domain https://mywebsite.com followed by the product page path and slug."
  6. Leverage LLMs: If you're unsure how to craft effective AI Instructions, ask your favorite Large Language Model to help you create a "system prompt" for your specific use case.

Content Customization

The AI Assistant offers numerous options for visual and content customization:

  • Title & Subtitle: Customize the header text of your chat widget.
  • Logo: Upload a custom logo to be displayed in the header and as the assistant's avatar.
  • Page Position: Choose whether the chat drawer appears on the left or right side of the page.
  • Chat Button: Enable or disable the floating chat button.
  • Custom Chat Trigger: Configure a specific HTML element (by ID) on your page to open the chat widget.
  • Load Delay: Set a delay (in seconds) before the chat button appears.
  • Auto-Load: Configure the assistant to auto-load on the first page a user visits in a session.
  • Suggested Prompts: Add up to three suggested prompts to guide users with common questions.

Styling

The color and overall style of the AI Assistant can be customized using the intuitive WYSIWYG tool in your dashboard. For advanced styling needs, you can add global CSS rules to target the assistant's specific CSS classes.

Tool Calls

The AI Assistant can be configured with optional "tool calls" (also known as functions) to perform dynamic actions on your website based on user prompts. Examples include adding a product to a cart, displaying a specific modal, or fetching real-time data.

ℹ️ Pro & Business Plans Only: Tool Calls are a premium feature available exclusively on Pro and Business subscription plans.
⚠️ Developer Integration Required: Implementing tool calls requires code to define the actual functions that execute on your website. See the Tool Calls Setup Guide for complete implementation instructions.

Analytics

BadgerFy.ai tracks various metrics related to your assistant's usage, including widget opens, chat interactions, link clicks within conversations, and token consumption. You can download your entire conversation history for diagnostic purposes.

Testing

Each AI Assistant comes with a dedicated test page in your dashboard. It's highly recommended to test your assistant's question-answering capabilities in this sandboxed environment before deploying it to your live website.

Test Environment Notes

  • Default Enabled: While your assistant is disabled by default upon creation, it is automatically enabled in the test environment for immediate testing.
  • Tool Calls Disabled: Tool calls and function execution are not active in the test environment. To thoroughly test tool call functionality, deploy to a staging version of your site.
  • Token Usage: Token usage and analytics are recorded in the test environment and count against your included tokens.