Studio

Knowledge Base for AI assistants

Prepare a shared document library for RAG assistants, product guides or virtual trainers. Select which files each character may use when answering questions.

The Knowledge Base is how a character stops improvising and starts answering from your actual material — the price list, the policy document, the product manual. Open it from the top bar: Knowledge › Knowledge Base ("Documents and pages your characters answer from"). Files live here at workspace level; each character is then given access to specific ones on its own Knowledge Base tab (Mind & knowledge › Knowledge Base in the character editor).

What you can upload

Drag files onto the drop zone or click it (Drag a file or click). You can drop several at once; each is processed on its own.

FormatNotes
PDFThe most common source. The text layer is read page by page; a scanned PDF with no text layer fails.
Word (.docx)Handled directly — no need to export to PDF first. The paragraph text is read, so put important answers in paragraphs rather than Word tables.
Text and Markdown (.txt, .md)Ideal, because you control exactly how it is written.
CSVGood for price lists and specification tables. It is read as plain text, so keep column headings meaningful.
Images — JPG, PNG, WebPThe text inside them is read automatically, so a photographed menu or a screenshot of a table works.

Maximum 50 MB per file. Other file types are refused, and so is a file whose contents do not match its extension.

Files count against your plan's knowledge storage — the Knowledge storage meter in the top bar's plan menu. When the storage is full, or your plan includes none, the upload is refused. Your plan's storage is listed on Plan & Quota.

What happens after upload

  1. PROCESSING — the file is read, split into short overlapping passages, and indexed so it can be searched by meaning rather than by keyword.
  2. READY — the file can now be used. The list (Files, with a count) refreshes itself, so you can watch it change. Each row shows the size, the number of chunks, the date and whether the file is tenant-wide.
  3. FAILED — the file could not be read. Usually a scanned PDF with no text, or an image with no readable text.
The workspace Knowledge Base page with a drag-and-drop upload box above a Files list of three sample files, each marked READY with a trash icon.
Knowledge › Knowledge Base in the sample workspace. Files dropped on Drag a file or click appear under Files; each row shows the size, chunk count, date and scope (here tenant-wide), a READY badge once indexing finishes, and the trash icon that deletes the file.
Uploading here does not give any character access. A file added on this page starts switched off for every character. Open each character's Knowledge Base tab and turn it on. To do both in one step, upload from inside the character instead — files added there are enabled for that character immediately. The Characters gallery shows how many knowledge sources each character has enabled.

Deleting

The trash icon on a file removes it from the workspace, together with its indexed passages and every character's access to it, immediately and without a confirmation prompt. There is no undo — re-upload if you delete the wrong one. To stop one character using a file without deleting it, switch it off on that character's Knowledge Base tab.

Writing documents that retrieve well

Retrieval finds passages, not documents. How you write has more effect on answer quality than which model you pay for.

DoInstead of
One file per subject — "Returns policy", "Delivery times", "Warranty"A single 200-page company handbook
Headings that read like customer questions — "How long does delivery take?""Section 4.2.1 — Logistics"
The answer immediately under the headingAn answer spread across three pages
Spelling out the subject in each section"As mentioned above, it is 14 days"
Plain sentencesTables of abbreviations with no context
That last one matters more than it looks. Passages are retrieved on their own, without the pages around them. A line that reads "14 days" is useless out of context; "Returns are accepted within 14 days of delivery" answers the question by itself.

Checking it works

Every character's Knowledge Base tab has a Search Test box. It searches only the files enabled for that character. Type the question a real customer would ask — not the keyword you think is in the document — and confirm the returned passages actually contain the answer. If they do not, the fix is in the document, not in the character.

Do this before blaming the AI. Most of the time a character "making things up" is a character that retrieved nothing useful, because the answer was buried in a table or phrased in language no customer would use.
The knowledge search runs in the standard conversation path. A character whose Core AI is Gemini Live (realtime) or OpenAI Realtime (voice) does not search these files.

Where to go next

Check your storage

Plans & usage →

IAMX AI STUDIO