Knowledge Files (Drive Sync)
Uploaded knowledge files are documents you control directly: price lists, policies, FAQs, brochures. They live in the project's Google Drive folder; a watcher notices your changes and keeps the project's vector store in step automatically. This is one of the two ways content enters the knowledge base; the other is Website Scraping.
The Drive folder structure
Every project gets its own root folder in Google Drive, created during onboarding (see Onboard AI Bot). Inside that root are three subfolders, each with a clear job:
| Subfolder | What goes here |
|---|---|
| knowledge-base | The documents the bot answers from. Files here are synced into the project's vector store. |
| reports | The project's evaluation report sheet (the Generic Response Evaluation Report), where answers are logged for review. |
| instructions | The system-prompt document for the project (a clone of the Generic System Prompt doc). Editing it updates how the assistant behaves. |
The Drive-watcher
You never touch the vector store directly. Instead, a small Apps Script (a script that runs on the Google side, attached to the project's Drive folder) acts as a watcher. Its only job is to notice a change and raise a signal:
- A file was created in the knowledge-base subfolder.
- A file was updated (replaced or edited).
- A file was deleted.
When the watcher sees one of these events, it triggers the AI File Management workflow (one of the Q-AI Bot's background automation workflows) and hands off the rest of the work.
The watcher runs on a schedule: about every 10 minutes during the day (roughly 7 AM to 10 PM), pausing overnight. So a file you add or edit is usually synced within a few minutes, while a change made late at night is picked up the next morning.
What AI File Management does
The AI File Management workflow reacts to the watcher's signal and maps each kind of change to one action:
| Drive event | What the workflow does |
|---|---|
| Create | Uploads the new file and attaches it to the project's vector store, then records it in the knowledge-base index. |
| Update | Replaces the file in the vector store with the new version. |
| Delete | Removes the file from the vector store. |
The result: the project's knowledge base always reflects what is currently in the Drive folder. Drop in a new policy and the bot can answer from it; delete an outdated brochure and the bot stops answering from it.
The system prompt lives in Drive too
The assistant's system prompt (the instructions that set its persona, rules, and tone) is not hand-edited in the managed configuration database. Instead, it lives in the instructions subfolder document described above.
When you edit that document, AI File Management picks up the change and writes the new instructions into the project's configuration, keeping prompt edits in one friendly place (a Google Doc) instead of the database. The system prompt is the only assistant setting changed this way; for which settings are edited where, see Per-Project Settings.
Where to go next
- Website Scraping: the other way to fill the knowledge base, for content that lives on the web.
- Knowledge Base & Retrieval: how the bot searches everything at answer time.
- Onboard AI Bot: provisions the Drive folder tree and the vector store for a new project.