GATHER — Connect your data
Upload your files and Tyrex turns them into a private, searchable knowledge base only your AI can access.
GATHER is where your AI gets its knowledge. You upload files, and Tyrex builds a private knowledge base your AI can search — no manual setup, no indexing config, no prompt engineering.
Where to do it
Upload from two places:
- Inside the Create wizard — the KUMPUL step asks if you want to feed your AI data as you build it.
- Data Studio (sidebar) — upload anytime, then attach the file to any persona.
- AI Library (Works → AI Library) — manage your datasets, attach them to multiple personas, or list them for sale.
Processing runs in the background — upload, walk away, and come back when the file shows as ready.
Supported formats
- Documents — PDF, Word (.docx)
- Spreadsheets — Excel (.xlsx, .csv)
- Structured data — JSON
- Plain text — notes, transcripts, markdown
Everything is processed automatically: the file is read, split into searchable pieces, and indexed. You'll see the file move through processing to ready — then it's live for your AI.
Files you upload become part of your AI's knowledge base. They're never used to train a shared model and never exposed to other users. If you later sell a dataset in the Marketplace, personally identifiable information is stripped automatically before listing.
Attaching data to an AI
Knowledge is scoped per AI (persona). When you attach a file or dataset to a persona, only that AI can retrieve it — your teacher AI doesn't see your business AI's files, even inside the same account.
In the app: open AI Library (Works → AI Library), select the file or dataset, and choose which personas can access it (data is managed there; persona pages show what's attached, read-only). The same dataset can be attached to several personas — one well-built knowledge pack can power multiple AIs.
You can attach the same dataset to multiple personas at once, so one well-built knowledge pack can power several AIs.
How your AI uses the data
When you chat, your AI searches its attached knowledge base for the most relevant pieces and grounds its answer in them — with sources you can check. That means:
- Answers are based on your documents, not general guesses.
- Long documents are handled by searching them, not cramming them into a prompt.
- Spreadsheets and structured data can be queried, summarized, and analyzed.
Storage limits
Knowledge counts against your storage quota:
| Plan | Storage |
|---|---|
| Free | 100 MB |
| Pro | 10 GB |
| Elite | 100 GB |
Storage is measured by the size of your indexed data. Cleaning up old files frees space immediately.
Optional: question generation
For files with lots of useful material, you can have Tyrex generate Q&A pairs from the content — a fast way to surface what your AI now knows, and a great starting point for testing it.
Practical example
A teacher uploads three chapter PDFs and last year's question bank. Their AI can now:
- Create practice questions from the chapters.
- Turn the question bank into fresh variants with reordered options.
- Explain concepts using the exact terminology of the textbook.
All from files that were never retyped or reformatted.
Next: SHAPE — assemble your AI.