# Write From Your Own Material, Not From Scratch: Drag Knowledge Base Items Into the Chat

> In a Beav chat, press ⌘K to open the right sidebar and drag cards from Knowledge, Assets or Manuscripts onto the input box. Reference cards with type labels appear above it and go along only when you send. The notes, images, videos and old drafts you've saved don't need uploading again, and you can ask the AI to mark which saved item each point comes from. A prompt, plus how it compares with re-uploading files to a general chatbot and building a notebook in a source-based research tool.

- Source: https://beav.pro/en/blog/features/write-with-your-own-library
- Author: JambaHailar (https://x.com/JambaHailar), Beav
- Published: 2026-11-22
- Last updated: 2026-11-22

> **Key takeaways**
>
> * Open the right workspace in a chat and drag cards from **Knowledge**, **Assets** or **Manuscripts** onto the input box. When you release, a reference card with a type label appears above the input. References are sent only when you click send, and you can remove them one by one before that.
> * With the input box focused, press `⌘K` (`Ctrl+K` on Windows / Linux) to open or close the right sidebar. Linked Obsidian vaults and Eagle libraries are listed there by name too.
> * This is where building a library pays off: no more re-uploading files or pasting links into a chatbot, because you're citing things you actually saved. Ask the AI to mark which item each point comes from, and you can check every one afterwards.

Most people start an AI writing session the same way: dig up a few links and paste them in, find a couple of images on their computer and upload them, then copy a chunk of something they wrote last time. Next topic, same routine. The chatbot meets your material from scratch every time.

If you've already saved that material in Beav, there's nothing to carry over. Your knowledge base holds the posts, articles and videos you've saved; the asset library holds images, videos, copy and decomposition outputs; Manuscripts holds your past drafts. Drag them into the chat when you write, and the AI works from material you actually collected instead of whatever the internet generally says. Saving, organizing and decomposing build the library. Citing it as you write is where it pays off.

## Who it is for

* Creators who have saved a pile of benchmark content and their own notes, but still copy and paste piece by piece when writing.
* Personal brands with a product or service, whose product sheets, customer feedback and old posts are already in the library and should back up what the AI writes.
* Content leads handing a team's collected material to the AI for a first draft, who want to know where each line came from.

## How to use it

The library needs something in it first. The most direct way to save content is to click &#x2A;*+** beside Knowledge Base, choose **Add link**, and paste the link or share text; **Add link** goes through the capture service and uses quota according to its rules. See [one box for every platform](https://beav.pro/en/blog/features/save-any-link-to-knowledge-base).

![Four steps to write from your own library: open the right sidebar with the shortcut, drag cards from Knowledge, Assets or Manuscripts onto the input box, set the brief and ask the AI to mark sources, check it and save it as a draft](https://beav.pro/blog/features/write-with-your-own-library-steps-en.webp)

*What you drag in is what you saved. It goes along when you send, and the marked sources let you check the result.*

1. **Start a new chat and open the right sidebar.** With the input box focused, press `⌘K` (`Ctrl+K&#x60; on Windows / Linux) to expand the right workspace, or click the sidebar button on the right of the chat title bar. It uses tabs like a browser; click &#x2A;*+** to open **Knowledge**, **Assets** or **Manuscripts**.
2. **Drag cards onto the input box.** Browse **Knowledge** or **Assets** and drag the cards you need onto the input box. When you release, a reference card with a type label appears above the input. Linked Obsidian vaults and Eagle libraries are listed by name as well, and manuscript cards also have an **Add to chat** button at the bottom. Since 2.8.13, typing `#` no longer opens a picker; it's just a character, so references are always dragged in.
3. **Set the brief, then ask for sources.** Give the platform, audience, key point and length, and ask the AI to note which item each point uses. For example:

```text
The references above are 5 items I saved: 3 Xiaohongshu posts from other creators in my niche,
1 summary of our shop's customer feedback, and 1 post I wrote last month.
Write a Xiaohongshu post on "what to prepare before bathing your cat for the first time",
for first-time cat owners, about 600 words.
Take points and examples only from these references. Learn how the other creators explain things, but don't copy their sentences.
After each point, put the reference it comes from in brackets. Where the references give no support, write "to be added" instead of inventing numbers.
```

4. **Check the references before you send.** References go along only when you click send, so remove any you dragged in by mistake. If you need to step away mid-draft, unsent text and reference cards are saved on this computer and waiting when you come back; drafts don't sync across devices.
5. **Check against the sources, then save it as a draft.** Use the sources the AI marked to go back to the original material: does the point mean what the source says, are the numbers really in the original? Once it's final, save it as a draft. Next time you write something similar, drag that draft in as a reference too.

![Cards from Knowledge, Assets, Manuscripts and linked Obsidian and Eagle libraries dragged into one chat, producing a draft with marked sources](https://beav.pro/blog/features/write-with-your-own-library-sources-en.webp)

*Illustration. Material collected in several places, used in one chat.*

**What about an image you just generated?** AI-generated images, videos and voiceovers don't go straight into the asset library; they land in **Generated**. Beav's AI and external AI tools search only the asset library by default, so to use the one you just made, say something like "use the image I just generated". Star the ones you'll reuse to save them to the library.

**Pasted attachments don't join the library.** Images and files you paste or drop in directly are saved only as attachments to that message. If you'll need one again, save it explicitly, and next time you can drag it in.

**Want it in another AI tool?** While Beav is running, every card on the local library web page has **Add to chat**; copy it and paste into ChatGPT, Claude or a similar tool.

**How do you keep it sounding like you?** References cover what to say; style covers how to say it. See [getting AI to remember your writing style](https://beav.pro/en/blog/features/ai-memory-and-writing-styles).

**What uses quota?** Saving with **Add link** goes through the capture service and uses quota according to its rules. The chat runs on the chat model and uses quota according to the usage rules.

## How it compares with the usual ways

|                               | Re-uploading files and pasting links into a general chatbot | Building a notebook in NotebookLM or a similar tool        | Beav: dragging knowledge base items into the chat                                            |
| ----------------------------- | ----------------------------------------------------------- | ---------------------------------------------------------- | -------------------------------------------------------------------------------------------- |
| Strength                      | Always at hand, many models, fastest for a one-off question | Q\&A grounded in a set of uploaded sources, with citations | Saved notes, images, videos, decompositions and old drafts sit in one library, ready to drag |
| Where the material comes from | You find and upload it each time                            | Uploaded per notebook                                      | What you saved with **Add link** and the files in your asset library                         |
| Upload again next time?       | Usually                                                     | Not within the same notebook; yes for a new one            | Save once, drag into any chat afterwards                                                     |
| Can you check the sources?    | You have to remember what you gave it                       | Citations jump to the source text                          | Ask the AI to mark sources; the originals stay in your library with their links              |
| Where the result lives        | Usually in that one chat                                    | In the notebook                                            | Saved as a draft you can drag in as a reference later                                        |

For a quick one-off question, a general chatbot is the fastest. For research across a batch of documents with strict citation links, NotebookLM and tools like it are purpose-built. Beav suits creators who keep saving content: material doesn't need uploading again, finished drafts go back into the library, and the more you write, the more you have to hand.

## Limits and notes

* **Check the sources the AI marks.** They make comparison quick, but they don't guarantee every line is accurate. Always check figures and quotes against the original.
* **Learn from other creators' delivery only.** You cite their posts to learn structure and angle. Get permission before using their sentences, images or video.
* **References go along only when you send.** Cards you drag in aren't handed to the AI until you click send, and you can remove them first.
* **Drafts stay on this computer.** Unsent text and reference cards are saved locally and don't sync across devices.
* **Beav doesn't post for you.** It helps you write and keep drafts. Publishing is yours.

## Next steps

* For why it's worth building up your own material, read the series overview, ["Accumulation Is the Moat"](https://beav.pro/en/blog/features/accumulation-is-the-moat).
* Write your own script from a decomposed talking-head video: [using a decomposed transcript and structure template](https://beav.pro/en/blog/features/rewrite-script-from-breakdown).
* Full reference: [give the AI material](https://beav.pro/en/docs/workspace/chat#give-the-ai-material), [the right workspace](https://beav.pro/en/docs/workspace/chat#the-right-workspace) and [Generated](https://beav.pro/en/docs/workflows/media#generated).

*Applies to Beav desktop 2.10 and later.*
