# Collect into knowledge

This is where you gather the material for building requests. What you pile up here is used to generate requirement cards and as the evidence (the cited source) behind AI assistant answers.

![PM on Rails knowledge. Meeting notes, documents and folders in one list. Before: evidence lived in mail and chat. After: the starting point for requirement cards stays on this screen.](https://pmonrails.com/docs-img/knowledge.png)

## Collecting

- **Create documents in Markdown** — register meeting notes and working memos. Manage specifications under Design docs. Headings, bold, lists, quotes, code, links, dividers, tables, and images go in from the toolbar or by typing / at the start of a line.
- **Write with AI** — on the document creation screen, press **"Write with AI"** to open the side AI chat (**"Ask or instruct AI"** in the header does the same). Write an instruction and pick a template; it generates a draft **by referring to the requirement cards, meeting notes and other material in the project**. The result lands in the body editor, so you can check and edit it before saving (nothing is saved automatically).
- **Rewrite with AI** — on a saved document, press **"Rewrite with AI"** to get a rewrite **based on the body you have open** in the editor. A body you are still editing works as the base too, and in "Ask instead" mode the AI reads your unsaved body before answering. The saved document stays unchanged until you save.
- **Word / Excel import** — upload a `.docx` / `.xlsx` and it is converted to Markdown into the body (20MB limit; the original file is not kept).
- **Bulk import from a ZIP archive** — upload a `.zip` and text-oriented files inside (Markdown, Word, Excel and the like) are **expanded into separate knowledge documents**, one file each. You do not have to upload a folder of material one file at a time.
- **Automatic transcription of audio and video** — upload audio or video from "Import video/audio" and it is transcribed automatically. The backend can be switched between Whisper (no speaker separation) and Google Cloud Speech (with speaker separation).

![PM on Rails import. Pick video or audio and start transcription. Before: you replayed recordings and typed the points. After: the transcript becomes knowledge and material for requirement cards.](https://pmonrails.com/docs-img/collect.png)

- **Segmenting what was said** — the transcription is split up by topic, and key points, pain points, decisions and proper nouns are extracted and tagged.

## Organising

- **Folders** — when creating a document or importing a file, you can leave the destination empty and place it at the top level of Knowledge. When you need to organise it, use folders in a tree structure up to 16 levels deep.
- **Version history** — every time you edit the body, the state before the edit is saved automatically as a past version (conflicts from simultaneous edits are detected too).
- **Version diffs and rollback** — compare any two versions line by line, and restore a past version.
- **Review / approval** — the flow is request → in review → approve / send back. Approving and sending back are done by a **workspace administrator (owner / admin)** on the "Review" screen. If something is sent back, the requester can fix it and request again. **A history of who requested, approved or sent back what and when (the approval audit) is kept.**
- **Document types** — organise meeting notes and design memos by type. Keep specifications and API / DB designs in [Design docs](https://pmonrails.com/en/docs/basic-design.md).
- **Search across everything from the search bar** — the knowledge home has a **permanent search bar** (focus it with `⌘K` / `Ctrl+K`) for filtering documents, meeting notes and folders by title. Typing switches the view to results right away (the dedicated search page and folder pages were retired and consolidated into the home). Results are filtered automatically based on your view permissions.
- **List / tree view** — the home switches between two tabs, "List" and "Tree". Opening a folder shows its contents in the list.
- **Discard confirmation while editing** — if you switch projects part-way through creating or editing a document, you are asked "discard your edits?", which stops work disappearing by accident.

## Generating requirement cards in bulk

From "Generate requirement cards" at the top of Knowledge, you can generate [requirement cards](https://pmonrails.com/en/docs/demands.md) in bulk from the documents and meeting notes you collected, either **per folder** or **across the whole project** (processed in the background). Targeting the whole project also picks up meeting notes that are not filed in a folder. Import the results by choosing "new / update existing / skip" in the review panel.

> Bulk generation across a whole project is **limited to maintainers (project administrators)**, because it can involve information you don't want shown widely. Per-folder generation can be run by editors.

## Connections form automatically

What you pile up is not simply stored.

- **The origin of a request is kept** — when you save a requirement card, it is automatically linked to **the meeting notes and documents that were its evidence**. Later you can trace "which meeting and which document did this request come from".
- **The content of discussions is picked up and tied together** — from the meeting notes and documents you collected, AI automatically picks out "topics", "what was decided", "what to do" and "concerns", and ties them to related requests and stories. That way, when you later ask the [AI assistant](https://pmonrails.com/en/docs/ai-assistant.md) a question, it can answer with **connected information** rather than scattered documents.

> Lists the AI produces, such as acceptance criteria, are shown as plain bullets (not as checkboxes you can tick).

## Next step

- [Create requirement cards](https://pmonrails.com/en/docs/demands.md)

---

https://pmonrails.com/en/docs/knowledge