Knowledge Base

When you register and log in for the first time, an empty knowledge base is automatically generated. The next step is to add knowledge sources.

We currently support three types of knowledge sources:

Files – Unstructured data from uploaded files.
Websites – Extracting and crawling all pages from a website.
GitHub Repositories – Indexing content from a GitHub repository.

Processing Files

EpisBase processes files in the background in three steps, ensuring optimal data extraction and usability.

Step 1: Textual Processing

In this initial stage, EpisBase extracts plain text from the files, making the knowledge source immediately available for use.

The processing time depends on the number of files and the content size, ranging from a few minutes to up to an hour.

The Content tab of the knowledge source displays the following statuses:

Pending for Processing – The file has been acknowledged and is waiting in the queue for processing.
Processing – The file is currently being processed.
Processed Successfully – The file is ready to be used in chat applications or quizzes.
Failed to Process – The file could not be processed. The reason for the failure will be displayed in a tooltip on the status.

Handling Processing Failures

If a file fails to process, an error message will be displayed. In such cases:

A “Process Failed Content” button will appear, allowing you to retry processing.
If the issue persists, you can either delete the file or contact us at hello@episbase.com for further assistance.

Step 2: Visual Processing

This secondary processing phase enhances the extracted data by analyzing visual elements within the document. This includes:

Extracting text from images and scanned documents.
Recognizing tables, charts, and diagrams, converting them into structured formats.
Generating descriptions for complex visuals, improving their accessibility for search and retrieval.

Unlike textual processing, no additional status updates will be displayed for visual processing. However, once completed, the content will receive a “Visually Processed” badge.

Step 3: Evolution – Learning & Summarization

To ensure the knowledge source remains effective and improves over time, EpisBase continuously learns from usage patterns:

When the knowledge source is used in a chat app, quiz, or API, the AI observes cases where it fails to retrieve correct answers, even though relevant content exists.
Over time, EpisBase optimizes the content by:
- Summarizing data for better context retrieval.
- Adjusting content chunks to improve relevance and accuracy.

Once this learning process is complete, the content will receive an “Evolved” tag, indicating that it has been enhanced for better performance.