Skip to main content

How to create a Knowledge Agent engine

1. What is a Knowledge Agent Engine?

A Knowledge Agent is designed to answer user questions based on your specific content, acting as an intelligent assistant for your website visitors. But before it can answer questions, you need to define its "brain", this is called an Engine.

An Engine for a Knowledge Agent is a sophisticated set of instructions that tells the Bridged AI how to find information, how to formulate answers, and how to handle different types of user queries. It ensures the AI stays grounded in your content, matches your brand's voice, and provides accurate, helpful responses.

Think of it as the complete operating manual for your AI assistant. This guide will walk you through creating a robust Engine for a Knowledge Agent.

2. Where to start

To begin creating an Engine for a Knowledge Agent, you must first navigate to the correct section of the Bridged Admin panel.

  1. Log in to your Bridged Admin account.

  2. From the main navigation sidebar on the left, locate and click on AI Labs .

  3. Under the AI Labs section, click on the sub-menu item Engines .

  4. You will be taken to the main Engines dashboard. Here, you can see a list of existing engines.

  5. To start the creation process, click the button to create a New Engine (indicated by the "< AI Lab | New Engine >" breadcrumb).

3. Step 1: Starting a New Engine

After clicking "New Engine," you will be presented with a detailed, multi-tab form to configure your Knowledge Agent. The process is designed to give you granular control over the AI's behavior.

4. Step 2: Defining the Basics

4.1. Choosing the Engine Type

The first and most critical choice is defining what kind of agent this engine is for.

  • Locate the Choose engine type dropdown at the top of the page.

  • From the options, select Knowledge agent . This tells the system to prepare the engine for question-answering tasks, grounding responses in your content.

4.2. Naming Your Engine

  • In the Engine title field, enter a clear and descriptive name for your engine.

  • Tip: Use names that reflect its purpose or knowledge domain, such as:

    • "Customer Support Knowledge Base"

    • "Event FAQ Assistant"

    • "Product Documentation Guide"

  • A good name will help you and your team select the right engine later when creating a new Knowledge Agent.

5. Step 3: The Core of Your Engine - Writing Guidelines

This is the most critical and detailed part of the process. The interface provides several distinct sections, or tabs, for different types of guidelines. Understanding each one is key to building a powerful engine.

5.1. Non-Editable Core Rules

This section contains fundamental, technical rules set by Bridged to ensure the AI's output is safe, accurate, and functional. You cannot edit these rules, but you should be familiar with them. For a Knowledge Agent.

These rules form the foundation of a trustworthy, non-hallucinating AI assistant.

5.2. Answering Guidelines

This is the primary text box where you define the overall tone, style, and structure of the AI's answers.

  • Click on the Answering Guidelines tab.

  • In the text editor, write instructions in plain English. For example:

    • "Always start with a friendly greeting."

    • "If the answer is complex, break it down into bullet points for clarity."

    • "Maintain a professional and helpful tone."

    • "If you are unsure, state that clearly and offer to connect the user with a human."

5.3. Using "Help Me Write"

If you need inspiration or want to refine your guidelines, use the AI-powered assistant.

  1. Click the Help me write link near the guidelines text box.

  2. A pop-up will appear with two fields:

    • "How should the response sound?" : Describe the desired tone. Example: "Keep it short and friendly, use bullet points, and write in UK English."

    • "Want the output to match a specific style of a URL? (Optional)" : Paste a link to a page that embodies the tone and style you want to emulate.

  3. Click Suggest guideline , and the AI will generate draft guidelines for you to use.

5.4. Intent Guidelines & Intents

Intents allow your Knowledge Agent to categorize user queries and potentially route them to different CTA handlers.

  • Click on the Intent Guidelines tab.

  • Here, you can define specific user intentions.

To add an Intent:

  1. Click the + Add Intent button.

  2. A new panel will appear where you can define:

    • Intent Title: Use lowercase, single words, or snake_case (e.g., pricing_query, technical_support, partnership_inquiry).

    • Intent Description: Explain what user queries should match this intent. For example, for pricing_query, you might write: "Any question about subscription costs, plans, billing, or payment methods."

  3. Click Save .

You can add multiple intents to help the AI better understand and categorize the questions it receives.

Link Intents to Specific CTAs

During the Knowledge Agent setup, you can connect intents to specific CTA features like banners, forms, lead capture options, and meeting bookings to streamline user interactions. For more information check Knowledge agent setup guide

5.5. Fallback Guidelines

This section defines what the agent should do when it cannot confidently answer a user's question based on its knowledge base.

  • Click on the Fallback Guidelines tab.

  • Write instructions for this scenario. For example:

    • "Apologize politely and explain that you couldn't find an answer to their specific question."

    • "Suggest they rephrase their query."

    • "Provide a link to contact human support or email help@bridged.media."

5.6. FAQ Guidelines

This section allows you to provide specific instructions for handling frequently asked questions, which might be treated differently from complex, open-ended queries.

  • Click on the FAQ Guidelines tab.

  • Write instructions for concise, direct answers. For example:

    • "For FAQs, provide the answer immediately in 1-2 sentences."

    • "If the FAQ has a known article, link to it at the end of the response."

6. Step 4: Testing Your Engine

Before saving, it's crucial to test your engine in a safe environment. The "Simulator" section allows you to do just that.

6.1. Selecting the Simulator

You have different simulation modes to test various aspects of your engine.

  • Locate the Select the Simulator section.

  • Choose between:

    • KA Simulator : A general-purpose simulator for testing standard Knowledge Agent queries.

    • FAQ Simulation : A simulator specifically designed to test how your engine creates Frequently Asked Questions for each page that is running on.

Configure Your FAQs

FAQs can be manual or AI-generated depending on your configuration. If you add FAQs manually on your Knowledge agent, the guidelines will not be applied.

6.2. Selecting a Collection

A "Collection" is a group of articles or documents that the Knowledge Agent will use as its source of truth.

  • Use the Select the Collection dropdown to choose the specific knowledge base you want to test your engine against (e.g., "KA for BE website").

6.3. Running a Simulation

Once you have selected a simulator and a collection:

  1. In the chat interface provided in the simulator (e.g., the "Bridge" assistant window), type a test question.

  2. Press send.

  3. Review the AI's response carefully. Does it follow your Answering, Intent, Fallback, and FAQ guidelines? Is it accurate based on the collection's content?

  4. Iterate on your guidelines and test again until you are satisfied with the behavior.

6.4. Bulk Simulation: Testing at Scale

What it is:
Bulk Simulation allows you to test your Knowledge Agent Engine against hundreds of questions at once by uploading a file.

How to use it:

  1. Prepare a file: Download the template and create an XLSX or CSV with one column named question. Add your test questions in the rows below.

  2. Upload: Drag and drop your file into the upload area.

  3. Run: Click to start the simulation. The system will process all questions and generate responses.

  4. Review: Check the output history to evaluate performance and consistency.

Why use it:

  • Quality assurance: Validate your engine handles a wide range of queries correctly.

  • Regression testing: After updating guidelines, rerun the same test to ensure nothing broke.

  • Confidence: Know exactly how your engine will perform before going live.

7. Step 5: Saving and Publishing Your Engine

Once you have configured all the guideline sections and are happy with the simulation results:

  1. Review all your settings.

  2. Click the Publish button (located at the bottom or top of the page). If you are not ready to publish, you may have a Save as Draft or Cancel option.

Your new Knowledge Agent Engine will now be saved and available for use.

How to Use Your New Engine:
When you go to create a new Knowledge Agent, you will be able to select your newly published custom engine from the dropdown menu in the design step . The agent will then use your sophisticated guidelines to power its interactions with users.