HivemindOS manual
Documents And Brain Drop
HivemindOS reads common documents on your computer and turns them into text an agent can use. The bundled native reader works locally and does not require a separate converter or API key.
Fast local reading is the default, with no cloud upload for conversion. If you then send the text to a cloud-backed agent or model in Chat, it becomes part of that request like anything you type. Use a local runtime when a document must stay entirely on your device.
Where you can use it
| Workflow | What you do | What you get |
|---|---|---|
| Chat | Attach a document to a message. | Text for the current request only, labeled as untrusted source material. |
| Feed the brain | Open Brain > Hive Vault, then select files or folders with the lower-right control, or drag them onto the vault. | Markdown notes in the Shared Brain’s Imported Sources area, with the source name, content hash, and import time. |
| Company data room | Choose a local data-room folder while importing a Zero Human Company. | A preview, then a Sources library in the company cockpit linked to Shared Brain notes. |
A long document fed to the brain, or imported with a company data room, is saved as a short overview note plus a folder of part notes, a few pages each, so a search can find any page of it. Your original files stay where they are. Unchanged files are not converted again. Attaching a file in Chat never turns it into Shared Brain memory on its own.
Supported file types
The reader accepts 16 supported file extensions:
| Extension | Format | Result |
|---|---|---|
.md, .markdown |
Markdown | The Markdown text. |
.txt |
Plain text | Readable text. |
.csv |
Comma-separated values | A Markdown table. |
.json |
JSON | Formatted JSON in a code block. |
.xml |
XML | Readable XML in a code block. |
.html, .htm |
HTML | Page content as Markdown. |
.pdf |
Text from readable pages. | |
.docx |
Microsoft Word | Document text. |
.pptx |
Microsoft PowerPoint | Text grouped by slide. |
.xlsx, .xls |
Microsoft Excel | Sheets as Markdown tables. |
.epub |
EPUB ebook | Chapter text. |
.msg |
Outlook message | Headers, body, and attachment names. |
.zip |
Document archive | All supported documents inside, grouped by path. |
Conversion keeps readable structure such as headings, rows, slides, and message fields. It does not reproduce layout, styling, charts, macros, or embedded media. The fast reader does not run OCR, so an image-only PDF or scan without a text layer may show a no-readable-content error.
Enhanced reading
Distill also offers Enhanced (Pro) reading for scanned pages, complex tables, formulas, images, and multi-column layouts. It needs an active HivemindOS Pro or Max plan and uses your shared HivemindOS credit balance.
- Enhanced reading is $0.03 per page with a $0.29 minimum.
- Distill shows the most a read can cost before it starts, then charges only for the pages read.
- Selected files are processed online in this mode. Choose Fast and private when a source must stay on your computer.
The normal reader needs no subscription and is the right choice for ordinary digital documents.
Limits
- Only local files; remote URLs and symbolic links are rejected.
- Up to 64 MiB per file and 1,000,000 characters of extracted text.
- A ZIP may hold at most 200 entries, with up to 64 MiB per expanded file and 128 MiB expanded in total.
- ZIPs are rejected if they contain nested ZIP files, links, unsupported types, paths that escape the archive, or extreme compression. One bad entry rejects the whole archive, so nothing is skipped silently.
- Empty, broken, or unsupported files fail with a clear error instead of reaching the agent.
What stays in your control
Every converted document is evidence, not authority. Text inside a file cannot become a system instruction, company directive, spending approval, credential, or permission. Chat wraps it as untrusted content, and saved imports keep their source hash for review.
Troubleshooting
- Unsupported file type: save a copy in one of the formats above.
- A PDF has pages but no text: it is probably image-only. In Distill, choose Enhanced (Pro), or run OCR in a trusted local tool and import the searchable copy.
- A large document is rejected: split it, or export only the sheets, slides, or pages you need.
- A ZIP is rejected: remove nested ZIPs, links, and unsupported files, and keep it to 200 supported documents or fewer.
- The reader is unavailable: install or repair the current HivemindOS desktop release. Installing Python or MarkItDown separately will not repair the bundled native reader.