Upload documents, text, and images

The IperChat agent knows your site (it read it during the demo) plus anything you add to the knowledge base. When a visitor asks a question, the system searches for the most relevant information across your materials and passes it to the AI model as context. This is called RAG (Retrieval-Augmented Generation).

The more material you upload, the more specific the agent becomes. Upload:

• Price lists, catalogs, product sheets
• Internal FAQs (even those not on the site)
• Manuals, procedures, terms of sale
• Q&A history from your real customers

Everything is managed from the ops panel, “Knowledge Base” section (side menu → Agents group).

Document types

The toolbar at the top has four buttons — Import Website, Import Q&A (Excel), Add image, Add text — plus a drag-and-drop area for PDFs/DOCX.

PDF and DOCX

The most common case. Max file size depends on the plan (see Plan limits).

1. At the top of the “Knowledge Base” page you’ll find the “Drag a PDF or Word file here or select from your computer” area.
2. Drag the file or click select from your computer.
3. Wait for the “Ready” status (typically 30-60 seconds for a medium document).

What happens under the hood: the file is uploaded, split into chunks, converted to vectors (embeddings), and stored in a semantic search engine. When a relevant question arrives, the most relevant chunks land in the model’s context.

Free text

For when you don’t have a document but want to enter information directly:

1. Click “Add text” in the toolbar.
2. Give a Title (e.g. “2026 Returns policy”) and Content.
3. Save.

The agent treats it like a mini-document. Great for:

• Operational notes (“Closed on Tuesday evenings”)
• Specific FAQs (“Cancelling an appointment requires 24-hour notice”)
• Quick updates (“New price list starting January 2026”)

Images

Upload photos of products, price lists, infographics, floor plans (JPEG, PNG, or WebP, max 10 MB):

1. Click “Add image” in the toolbar.
2. Select the file and give a detailed description: this is the text the agent reads to understand what the image shows.
3. Save.

When the agent decides an image is relevant to the answer, it displays it directly in chat. Great for catalogs, menus, room layouts.

Q&A import (Excel)

If you have a history of questions/answers from your customers (exported from your CRM, Helpscout, an email folder), upload an Excel file with a question column and an answer column:

1. Click “Import Q&A (Excel)” in the toolbar.
2. Upload the .xlsx file (max 10 MB, minimum 10 rows, maximum 5,000).
3. The system analyzes the Q&As, groups by topic, synthesizes a canonical answer for each topic, and creates several automatic text entries (child documents).

The in-app help module (“How to prepare the Excel file”) explains recognized column headers and limits.

Import website

To feed the knowledge base directly from your site’s content, click “Import Website”: enter the URL and the maximum number of pages to read. The system extracts the page text and adds it as documents.

Document status

Each document in the table shows a Status:

• Uploading… — upload in progress.
• Processing… — the system is extracting text, chunking, and generating embeddings.
• Ready — indexed, the agent can use it.
• Error — something went wrong. Open the document or check the tooltip to see the details (password-protected PDF, corrupt file, unsupported format).

Table columns also include Type (PDF, DOCX, Text, Image, Chat Import, Site Import, Auto), Chunks, Size, and Date / Uploaded / Modified.

When the agent uses a document

The agent automatically decides which documents to consult based on the question. If you ask “How much does the consultation cost?” it looks in price lists and pricing sheets. If you ask “How do I book?” it looks in booking procedures.

No need to tell it when to use what: matching is semantic.

Updating a document

PDF/DOCX documents cannot be edited in place: they’re replaced (delete the old one via the actions menu → Delete, then upload the new one). Alternatively, use the Re-process action to re-ingest the same file without re-uploading.

Free text documents can be edited directly: open them from the list, use Edit, save. Re-indexing starts automatically.

Limits

Vary by plan (default values — can be customized by the platform administrator):

When you reach the document limit, you can’t upload new ones until you delete one. The agent keeps working with those already uploaded. See Plans and limits for the rest of the plan details.