Resources

Resources are the material your company did not write — articles, PDFs, ebooks, transcripts, a page you pasted the URL of. You ingest it once; your AI Employees search it and quote from it instead of guessing.

What counts as a Resource

Genosyn keeps three different kinds of knowledge apart on purpose, and picking the right one is most of getting good answers out of an AI Employee.

Resource
Someone else's words, ingested for study. A vendor's API guide, a competitor's pricing page, the ebook your founder keeps recommending. Read on demand, never automatically injected.
Note
A page your team authors together, in markdown. See workspace chat for how they get referenced in conversation.
Memory
A short durable fact about how your company works, injected into every prompt. Facts, not documents.

Add one

Go to Resources in the sidebar under Knowledge. Three tiles:

  • Paste a URL — Genosyn fetches the page and extracts its readable text. The text is a snapshot taken at that moment; it does not re-fetch later.
  • Paste text — a transcript, an email thread, anything you can copy. This one stays editable afterwards as a markdown document.
  • Upload a file — PDF, EPUB, Word, plain text, Markdown or HTML, up to 25 MB. Video files are stored and playable but not transcribed yet.

Whatever you add, the original is kept and the extracted text is what gets searched. A Resource can carry any number of tags.

How an AI Employee uses one

Resources are read through tools, on demand — they never sit in an employee's prompt. There are three read tools and they are designed to be used in order.

search_resources
Finds the passage. Every word of the query has to appear somewhere, in any order — “refund policy” matches a handbook that says “our policy for refunds”. Each hit comes back with a snippet of the matching text and the exact character offset it came from.
get_resource
Reads from that offset. A window at a time, with the offset to resume from, so an employee can work through a 300-page book instead of receiving the first chapter and a note saying the rest was discarded.
list_resources
Browses the shelf — titles, summaries and tags, newest first, paginated. For when the employee wants to know what exists rather than answer a question.

Employees can also curate: create_resource files a URL, a paste, or a file it already holds — an emailed contract, a form it downloaded. update_resource and delete_resource need a higher grant. export_resource renders the body as PDF, HTML, Markdown or plain text and attaches it to the reply.

Who can read it

Every Member of the company sees every Resource. AI Employees are granted access one row at a time, at three escalating levels — View only, Can edit, and Can delete. Open a Resource and use Share to change them.

You rarely have to. A new Resource is granted View only to every AI Employee in the company as it is created, and an employee hired later is granted the company's existing library as it is hired — so the order in which you fill the shelf and grow the team no longer decides who can read anything. An employee that files its own Resource gets full control of that row; teammates start at view-only.

Reading and exporting

The detail page is type-aware: PDFs open in the browser's viewer, EPUBs in a reader with a table of contents and saved position, text Resources as an editable markdown document, and a URL Resource shows a card linking the original. Everything can be exported as PDF, HTML, Markdown or plain text from the Download menu.

Originals you uploaded are served as downloads rather than rendered in the page, except for the three kinds Genosyn has a viewer for — PDF, EPUB and video. An uploaded HTML or SVG file is stored and downloadable but never rendered on the app's own origin.

Limits

  • 25 MB per uploaded file.
  • 1 MiB of extracted text per Resource. A longer ebook is stored whole and its text truncated at that point.
  • 8 MiB per export.
  • Retrieval is lexical — it matches the words you use. Embeddings and vector search are not shipped yet.