# Knowledge User Guide

### Overview <a href="#overview" id="overview"></a>

Knowledge is an AI-powered knowledge retrieval tool designed to help teams efficiently access information from diverse sources—including uploaded files, databases, and online resources. Knowledge delivers accurate, contextual answers with transparent source citations, enabling users to resolve queries quickly and confidently. A key feature of Knowledge is its continuous improvement cycle: users can provide feedback on retrieved answers, allowing knowledge managers to refine and expand the knowledge base over time.

With its intuitive interface, advanced natural language understanding, and seamless integration capabilities, Knowledge enhances knowledge accessibility, streamlines operations, and provides fast, reliable solutions for teams across various industries, including customer support, HR, and IT.

***

### Benefits for End Users <a href="#benefits-for-end-users" id="benefits-for-end-users"></a>

#### Quick and Reliable Responses <a href="#quick-and-reliable-responses" id="quick-and-reliable-responses"></a>

Knowledge retrieves answers exclusively from the knowledge stored within it, ensuring responses are accurate and relevant.

#### Answer Verification via Screenshots and Thumbnails <a href="#answer-verification-via-screenshots-and-thumbnails" id="answer-verification-via-screenshots-and-thumbnails"></a>

Users can cross-Check answers with attached visual references, providing an extra layer of validation before responding to customers.

#### Follow-Up Questions (Threaded Conversations) <a href="#follow-up-questions-threaded-conversations" id="follow-up-questions-threaded-conversations"></a>

Instead of treating each question as standalone, Knowledge supports conversational continuity, allowing users to ask follow-up questions within the same context.

#### PII Data Masking for Privacy Protection <a href="#pii-data-masking-for-privacy-protection" id="pii-data-masking-for-privacy-protection"></a>

To enhance security, Knowledge can mask personally identifiable information (PII) within requests, responses, and stored knowledge.

#### Knowledge Base Organization (Applications and Knowledges) <a href="#knowledge-base-organization-applications-and-knowledges" id="knowledge-base-organization-applications-and-knowledges"></a>

Knowledge is structured into applications and Knowledges, allowing separation of data across different departments or use cases.

#### Role-Based Access Control <a href="#role-based-access-control" id="role-based-access-control"></a>

* **Advisor:** Access is restricted to the chat interface.
* **Knowledge Manager:** Can upload, edit, and delete knowledge, modify knowledge chunks, and manage advisors for their Knowledge.
* **Admin:** Has full permissions, including creating Knowledges and assigning Knowledge Managers and Advisors.

#### Reporting and Analytics <a href="#reporting-and-analytics" id="reporting-and-analytics"></a>

A reporting section provides insights into costs, user feedback, and performance metrics.

#### Optimized Knowledge Chunking <a href="#optimized-knowledge-chunking" id="optimized-knowledge-chunking"></a>

Knowledge allows the optimization of chunking strategies based on knowledge structure and use case, ensuring efficient retrieval and response accuracy.

***

### Supported Knowledge Sources <a href="#supported-knowledge-sources" id="supported-knowledge-sources"></a>

Knowledge can retrieve information from the following sources:

* **Documents:** TXT, XLS, PPTX, PDF (maximum 10 files per upload, total size limit of 300MB).
* **Public URLs:** Knowledge can scrape content from specified web pages.
* **Knowledge Base Connectors:** Currently supporting Confluence, with more integrations planned.

***

### Querying Best Practices <a href="#querying-best-practices" id="querying-best-practices"></a>

To maximize the effectiveness of Knowledge, users should follow these best practices:

1. **Use Specific Subject Names** – Instead of pronouns like "our," "it," or "we," explicitly state the name of the object or subject.
2. **Review Rephrased Questions** – If Knowledge reformulates a question and the answer is unsuitable, refine the phrasing to improve accuracy.

***

### Querying Best Practices <a href="#querying-best-practices.1" id="querying-best-practices.1"></a>

To maximize the effectiveness of Knowledge, users should follow these best practices:

1. **Use Specific Subject Names** – Instead of pronouns like "our," "it," or "we," explicitly state the name of the object or subject.
2. **Review Rephrased Questions** – If Knowledge reformulates a question and the answer is unsuitable, refine the phrasing to improve accuracy.
3. **Provide Context** – Ensure questions include enough details for Knowledge to understand the request accurately.
4. **Avoid Jargon** – Use clear and simple language to enhance understanding.
5. **Ask One Thing at a Time** – Keep questions focused on a single topic for more precise answers.
6. **Rephrase If Necessary** – If an answer is not useful, try rewording the question to improve relevance.
7. **Do Not Use Knowledge as a Chatbot** – Avoid greetings, casual conversation, and subjective phrasing; ensure each query is direct and relevant.
8. **If something was removed during the automatic rephrase part,** try to reformulate the question or use the follow-up functionality to provide additional context.

***

### User Roles and Responsibilities <a href="#user-roles-and-responsibilities" id="user-roles-and-responsibilities"></a>

#### **Advisor** <a href="#advisor" id="advisor"></a>

* **Purpose:** Primarily interacts with the chatbot to retrieve knowledge-based responses.
* **Responsibilities:**
  * Can only access the chat page.
  * Uses Knowledge to answer queries within assigned applications.
  * Provides feedback on responses to improve accuracy.

#### **Knowledge Manager** <a href="#knowledge-manager" id="knowledge-manager"></a>

* **Purpose:** Maintains and curates the knowledge base to ensure data accuracy and completeness.
* **Responsibilities:**
  * Uploads, edits, and deletes knowledge entries.
  * Directly modifies chunks of knowledge to improve information clarity.
  * Manages and assigns Advisors within their designated Knowledge.

#### **Admin** <a href="#admin" id="admin"></a>

* **Purpose:** Oversees Knowledge setup and user role assignments.
* **Responsibilities:**
  * All permissions of a Knowledge Manager.
  * Can create and configure Knowledges.
  * Assigns Knowledge Managers and Advisors to specific Knowledges.
  * Monitors overall system usage and reporting.

***

### Complete Workflow <a href="#complete-workflow" id="complete-workflow"></a>

1. **Super Admin creates the application** and assigns an Admin from the client’s organization.
2. **Admin sets up one or multiple Knowledges** within the application and assigns Knowledge Managers to them.
3. **Knowledge Managers upload knowledge sources** using file uploads, web scraping, or Confluence integration.
4. **Knowledge Managers select the optimal chunking strategy** (default: chunk by page with overlap) to ensure effective knowledge retrieval.
5. **Knowledge Managers assign Advisors** to their Knowledges.
6. **Advisors use the chat interface** to support customers by retrieving knowledge-based answers.
7. **Feedback from Advisors and users is utilized** to refine the knowledge base, ensuring continuous improvement.

### How to Improve the Ingested Knowledge <a href="#how-to-improve-the-ingested-knowledge" id="how-to-improve-the-ingested-knowledge"></a>

* As a Knowledge Manager, you are responsible for the quality of the knowledge, so user feedback should be monitored regularly in the reporting section.
* Update the chunks if something wasn’t processed correctly. Remove unrelated data (noise). Complete chunks if they were truncated and the answers are not correct for the relevant question.
* Choose the most suitable chunking strategy:
  * For short answers that can be directly identified in the document, use **split by page with small overlap** (default).
  * For documents containing troubleshooting steps that require long, detailed answers, use a **chunking strategy with as much text as possible** (e.g., split by 800 words with a 400-word overlap).
* Remove unnecessary images from files before uploading them.
* Use advanced scraping methods if the extracted content is not good.
* Set the **formality, expertise level, and response length** appropriately for better answers.
* Utilize **client-specific instructions** to refine Knowledge’s responses (e.g., "answer briefly without reasoning").

Related content<br>


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.ixhello.com/knowledge-bot/knowledge-user-guide.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.