Documentation

Your complete guide to deploying AI-powered support across multiple sites with intelligent collection routing.

Table of Contents

1) Sign In

Log in to your account and access the admin dashboard to configure your chatbot. After authentication, you'll see the main admin panel with multiple tabs for managing your AI system.

2) Admin Panel Overview

The admin dashboard contains the following sections, accessed via tabs at the top. These features allow you to manage your AI chatbot and knowledge sources:

  • General: Configure AI behavior including response creativity level (how varied the answers are), model selection, smart search settings, and API settings. Set up allowed URLs for widget deployment and customize rate limiting and caching.
  • Knowledge Files: Upload and manage TXT/CSV files directly. Assign keywords to files for intelligent retrieval, organize files into collections, and track file sizes and modification dates.
  • Knowledge Bases: Connect to external databases through XML feeds. Supports dynamic content syncing from OpenCart, WordPress, WHMCS, and other platforms with custom XML export options.
  • Routing: Configure intelligent collection routing rules. Map specific keywords and page URL patterns to different knowledge collections, ensuring the AI selects the correct information source.
  • Widget: Generate unique API keys for each website deployment. Customize widget appearance per site with logos, button styles, colors, welcome messages, and language settings.
  • Logs: Monitor all conversation activity and system events. Review which collections were used, track user engagement patterns, and identify frequently asked questions.
  • Profile: Update your user profile information including email, address, phone, and company details.

3) Organize Your Knowledge

  • Create data collections: Organize knowledge into separate collections for different products, services, or websites.
  • Upload files: Add TXT, CSV, XLSX, PDF, DOC, or DOCX documents to each collection via the Knowledge Files tab.
  • Link XML feeds: Connect dynamic content feeds (see API Reference for XML format) that auto-sync with your databases through the Knowledge Bases tab.
  • Set routing rules: Assign keywords and page URLs to each collection so the bot automatically selects the right knowledge source based on context.
  • Keyword assignments: Add specific keywords to knowledge files to improve relevance matching. The system uses these keywords to select appropriate content when answering questions.
  • Collection management: Assign each file to a specific collection for advanced intelligent search and organization.

4) Configure AI Behavior

  • Model selection: Choose from available GPT models (gpt-4o, gpt-4.1, gpt-5, etc.) in the General tab. Each model has different capabilities, pricing, and context window sizes.
  • Response Creativity Level: Adjust how varied and creative the AI responses are on a scale of 0.0-2.0. Lower values (0.0-0.5) = more accurate and consistent answers. Higher values (1.5-2.0) = more varied and creative responses. Middle range (0.7-1.0) = balanced approach.
  • System prompt: Define custom system instructions to guide how the AI behaves, what tone it uses, and how it prioritizes information.
  • Search Intelligence: Configure the search accuracy level for intelligent knowledge retrieval (Standard or Advanced).
  • Rate limiting: Set API call limits per IP address and per API key to control usage and prevent abuse.
  • Cache settings: Configure cache TTL (time-to-live) to optimize response times for frequently asked questions.
  • Zero setup required: GPT usage is included—no API keys needed by default (can be overridden per user if needed).

5) Deploy on Multiple Sites

  • Generate API keys: Create unique keys for each website in the Widget tab. Each key can have different permissions and rate limits.
  • Customize per site: Configure different colors, logos, button styles, and welcome messages for each deployment through the Widget tab.
  • Widget configuration: Upload custom logos and button icons specific to each site's branding requirements.
  • Embed the script: Add the loader to any site: <script src="https://chat.sbzsystems.com/chat_widget.js?apikey=YOUR_KEY"></script>
  • Add language support: Include &lang=en (or your locale) to localize the interface.
  • Automatic routing: The widget automatically selects the correct data collection based on the page URL and conversation keywords.
  • Collection-aware deployment: Each widget can be configured to use specific collections, allowing fine-grained control over which knowledge sources are available on each site.

6) Monitor and Optimize

  • Test conversations: Ask typical customer questions and verify the bot cites correct sources from the right collection.
  • Review analytics: Access the Logs tab to track conversation volumes, user engagement patterns, and conversion events.
  • Collection performance: Monitor which collections are being used most frequently and identify knowledge gaps.
  • Measure performance: Identify which collections get the most use and which questions need better content.
  • Refine content: Add missing information to underperforming collections based on log insights and user questions.
  • System metrics: Track cache hit rates, API response times, and search performance.

7) Advanced Features

  • Knowledge Base Synchronization: Set up periodic sync with external databases through XML feeds. Supports automatic updates when source data changes.
  • Response Formatting: Apply custom normalization and formatting rules to ensure consistent output quality.
  • Per-user settings: Each user can customize their own settings for models and knowledge sources.

8) Best Practices

  • Collection strategy: Create separate collections for distinct product lines, service areas, or geographic regions. Organize by function rather than arbitrary groupings.
  • Keyword optimization: Use specific, relevant keywords to ensure accurate collection routing (e.g., "shipping", "warranty", "pricing"). Include synonyms and common variations.
  • URL patterns: Map collections to URL patterns (e.g., /products/* → Product Collection, /support/* → Support Collection) for automatic routing.
  • Plain text content: Keep descriptions and XML feeds in plain text—no HTML or special formatting.
  • Incremental updates: Always update the modified timestamp in XML items to enable efficient syncing.
  • Multi-site branding: Customize widget appearance per site to match each brand's visual identity. Ensure logos and button colors align with site design.
  • Creativity level tuning: Start with default 0.7 creativity level. Lower it for support/FAQ domains requiring accuracy; raise it for creative domains like copywriting or brainstorming.
  • Model selection: Use cost-effective models (gpt-4o-mini, gpt-4.1-mini) for high-volume, routine queries. Reserve premium models (gpt-5.2) for complex reasoning tasks.
  • Cache optimization: Set appropriate cache TTL based on your knowledge update frequency. Higher TTL = better performance but staler data.
  • Rate limit tuning: Configure limits based on expected traffic. Too strict limits may throttle legitimate traffic; too loose may enable abuse.
  • Search Intelligence: Use standard search for most use cases. Consider advanced search for higher accuracy if performance permits.
  • Regular audits: Review logs monthly to identify underutilized collections, frequently unanswered questions, and opportunities for content improvement.