Training the Chatbot with Your Content

You can train the chatbot to answer questions based on your own uploaded documents.

Steps:

Go to Dashboard → Resources
Upload up to 10 files (max 3 MB per file)
Supported formats: TXT, DOCX, PDF, MD
After uploading, click “Retrain Model” to apply changes

Only plain text is extracted from documents. Formatting like tables, bold, or font size is not preserved.
Page count is estimated assuming 1 page = 2,000 characters of plain text.

Knowledge Base Management

In Settings → Knowledge base, you can manage all content sources used to train the chatbot.

You can add two types of sources using the Add source button:

1/ Documents

Supported file types: TXT, DOCX, PDF, MD
Maximum of 10 files, up to 3 MB per file
Drag and drop files or select them manually

2/ Web

Add a website starting from a given URL
By default, the crawler will explore the entire site structure beginning from the entered URL
If you enable Specific link, only that single page will be indexed
You can exclude paths (e.g., admin/, login/) to prevent them from being crawled
Start the process with Run crawler to fetch and index the content

For each source (file or url), you can view: File name or link name, Upload date, Status, Number of extracted text pages, Actions (preview, delete). The total number of uploaded pages is shown in the right panel. Pages are estimated at 2,000 characters of plain text. After uploading documents or adding websites, click Retrain Model to apply the new sources.

Tip: You can edit files or view them in 'Hierarchy View', which helps you see the data structure exactly as the model sees it, allowing for better adjustments.