Knowledge base management – Support Perxona

Beyond its default interaction modes, Perxona allows you to upload and manage company documents to create a custom knowledge base. This empowers your AI assistant to instantly reference accurate, up-to-date information, providing more professional and precise responses. This FAQ will guide you on how to upload documents, check their processing status, and offer best practices to help you effectively leverage the knowledge base feature.

Q1. How do I upload company knowledge for my AI avatar to use?

You can upload documents through the Console’s Knowledge section and connect them to your AI avatar for enhanced response quality.

Perxona offers an integrated Knowledge Management system that enables you to upload and organize your organization’s most relevant documents—such as support manuals, sales guides, or internal policies—so your AI Agents can reference accurate, up-to-date information during conversations.

To upload and activate a document for AI avatar use:

Navigate to the Knowledge page and click “Add” to begin.
Drag and drop your document into the upload dialog.
Provide a name and description to help identify the file, then click “Confirm.”
Once uploaded, the file status will show as “Processing” and then “Active,” indicating that the content is being parsed and embedded into the organizational database.

To connect the knowledge to your AI avatar:

Go to the Storyboard section and open an existing storyboard.
In the storyboard Settings, switch to the “Global Data” tab.
Select and attach your Active knowledge files from the list.

Your AI Agent will now be able to reference the uploaded knowledge during conversations, enhancing its accuracy and relevance.

Q2. How do I upload a knowledge file?

Navigate to the Knowledge page from the main navigation menu.
Click Add to open the upload dialog
Drag and drop your document into the dialog box or click to browse files from your computer.

Supported file types: PDF, DOC, TXT, CSV

Provide a name and description to help identify the content later.
Click Confirm to begin processing.
Document size

You’ll see the document status as Processing Active, which means it’s being converted into a structured format usable by the AI agent.

Q3. How do I check the status of a knowledge file?

The knowledge file status is:

Processing: the document is uploaded, and the system is processing
Not in use: the document is processed, but no storyboard refers to it
In use: a storyboard has been created and refers to this document
Error: the document cannot be processed. To resolve the issue, you can try reuploading the file or using an alternative file format.

You can edit, download, and update the knowledge file on this page.

!Note: Unlike avatar, scene, and storyboard, the knowledge file is not an asset, and any processed update will immediately apply to the published AI avatar. This ensures that end users always have up-to-date knowledge. Be sure you are using the correct knowledge file.

Q4. How should I structure reusable content in knowledge files?

The Perxona system parses and analyzes the document in combination with various strategies, such as semantic paragraphing, which ensures that the end user receives the most accurate answer through up-to-date files. But to keep the system simple and make sure the knowledge will be easy to find and constantly refreshed, we recommend that:

Keep documents concise: Split large manuals into smaller, focused documents for better chunking and retrieval.

Use descriptive titles: Name files clearly to make them easier to manage and assign.

Update regularly: Replace outdated files using the Update Version option to maintain content accuracy.

Prevent complex layout: AI agents perform better with clean and structured content.

Avoid:

Multi-column layouts (e.g., newspaper-style)
Text wrapped around images
Tables that mix multiple concepts or rows with merged cells

Prefer:

Left-aligned, single-column text
Clear section headers (e.g., ## Shipping Policy)
Simple, clean tables with consistent formatting

Use Clear Headings and Hierarchies: Use headings (e.g., H1, H2) to structure your content into meaningful sections.

e.g.,

## Returns & Refunds
document content…..

Represent Reusable Queries as Q&A Blocks: Use a clear Q&A structure for frequently asked questions.

e.g.,

Q: What is your return policy?

A: You may return any item within 30 days for a full refund.

Use Bullet Points for Lists

e.g.,

Features of Product X:

- Long battery life (up to 12 hours)

- Lightweight (1.2kg)

- Supports Wi-Fi and Bluetooth

Remove Irrelevant Metadata: Avoid including footers, legal disclaimers, and timestamps in each page unless essential. These can clutter the knowledge and confuse the language model during retrieval.

Maintain Consistent Language: Ensure consistent terminology throughout the document. For example, if you refer to your users as “clients,” don’t switch to “customers” mid-way. This helps the AI avatar stay semantically aligned with your brand language.