> ## Documentation Index
> Fetch the complete documentation index at: https://docs.valiancehealth.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Workspaces

> Organize files and authored documents in an AI workspace, and control exactly what the workspace chat assistant uses by promoting documents to chat.

## Overview

A **workspace** is a place to keep the context you want to work with alongside the
Healthproximate AI chat assistant. Each workspace holds two kinds of content — uploaded
**files** and authored **documents** — and gives you precise control over which of that
content the chat assistant actually uses.

The key idea: adding content to a workspace does **not** automatically change how the
assistant behaves. You decide what the assistant honors by **promoting** specific
documents **to chat**. Everything else stays as reference material you can browse, but
that the assistant won't treat as instructions.

Open workspaces from **Dashboard → AI → Workspaces**.

<Note>
  Workspaces are currently available to superusers.
</Note>

## When to use it

* You want the chat assistant to follow consistent preferences every time you work in a
  workspace (tone, format, house conventions).
* You have reusable how-to content or standing facts you want the assistant to apply
  without re-explaining them each session.
* You want to keep reference PDFs and documents in one place that the assistant can look
  through when you ask.

## Files vs. documents

A workspace separates content you bring in from content you author in the app.

|                             | **Files**                                                       | **Documents**                                               |
| --------------------------- | --------------------------------------------------------------- | ----------------------------------------------------------- |
| What they are               | Uploaded PDFs and documents                                     | Markdown you author in the app                              |
| How they get there          | Upload                                                          | **+ New document**, then edit and save                      |
| How the assistant uses them | Reference the assistant can list, read, and search when you ask | Reference only — until you **promote** the document to chat |

Files are always available to the assistant as readable context: it can list them, open
them, and search across them in response to your questions. Documents are different —
authored documents are inert by default, and only start influencing the assistant once
you promote them (see [Promoting a document to chat](#promoting-a-document-to-chat)).

## Creating and editing documents

<Steps>
  <Step title="Create the document">
    In a workspace, choose **+ New document** and pick a type (see
    [The five document types](#the-five-document-types)).
  </Step>

  <Step title="Write and preview">
    Author the content in markdown. The editor has an edit view and a preview view so
    you can check how it will read.
  </Step>

  <Step title="Save">
    Click **Save** to store your changes. Saving is explicit — if you try to leave with
    unsaved edits, you'll be warned so you don't lose work.
  </Step>
</Steps>

Documents live in the workspace and can be edited, re-typed, or archived at any time.

## The five document types

When you create a document you choose one of five types. The type determines what the
document is for and, importantly, whether it can influence the chat assistant.

| Type                       | What it's for                                            | Can be promoted to chat? |
| -------------------------- | -------------------------------------------------------- | ------------------------ |
| **Note**                   | Free-form notes and working scratch content              | No                       |
| **Runbook**                | Step-by-step procedures for people to follow             | No                       |
| **Memory note**            | Standing facts and context for the assistant to consider | Yes                      |
| **Skill template**         | Reusable how-to / skill content the assistant applies    | Yes                      |
| **Workspace instructions** | Workspace-wide behavior preferences for the assistant    | Yes                      |

Notes and Runbooks are for you and your team — they never affect the assistant. The
other three types are the ones you can promote.

## Promoting a document to chat

Promotion is how you tell the assistant to honor a document. To promote, open the
document's **⋯** menu and choose **Promote to chat**. A promoted document shows an
**"In chat"** badge and a lightning icon in the sidebar, and the chat assistant in that
workspace will start honoring it.

Three types can be promoted, and each carries a different kind of standing:

* **Workspace instructions** — workspace-wide behavior preferences. These have the
  highest standing of the three. **Only one Workspace instructions document can be
  active per workspace.**
* **Skill template** — reusable how-to or skill content the assistant applies when it's
  relevant.
* **Memory note** — standing facts and context the assistant considers. Treat these as
  context, not commands — they inform the assistant rather than direct it.

<Warning>
  **Only one active Workspace instructions document per workspace.** If a Workspace
  instructions document is already promoted, promoting a second one is rejected. Remove
  the first from chat before promoting the replacement.
</Warning>

### Editing a promoted document removes it from chat

Promotion is tied to the exact content you reviewed. To keep "what you approved is what
runs," Healthproximate automatically removes a document from chat when it changes:

* **Editing the content** of a promoted document removes it from chat. You'll see a
  warning when you save. Review the change, then re-promote it.
* **Changing its type** removes it from chat.
* **Archiving it** removes it from chat.

<Warning>
  After you edit a promoted document, it is no longer active in chat until you promote
  it again. Re-promote once you've confirmed the new content is what you want the
  assistant to use.
</Warning>

### Removing a document from chat

To deactivate a promotion at any time, open the document's **⋯** menu and choose
**Remove from chat**. The document stays in the workspace as reference — it just no
longer influences the assistant. The "In chat" badge and lightning icon go away.

### The `skills.md` file

A workspace can also carry an uploaded file named exactly `skills.md` at the workspace
root. It behaves like Workspace instructions. For most cases, an authored **Workspace
instructions** document is the recommended in-app replacement — it's easier to edit,
preview, and manage through the promotion lifecycle above.

## Limits and scope

* **Organization scope.** Everything in a workspace — files, documents, and promotions —
  is scoped to the workspace's organization.
* **Session size limits.** There's a limit on how much promoted content the assistant
  uses in a single chat session. An oversized document is **skipped whole**, with a
  notice, rather than being partially truncated — so the assistant never acts on half a
  document. If you hit this, split large content into focused documents.

## Troubleshooting

<AccordionGroup>
  <Accordion title="Why doesn't the chat assistant see my document?">
    The most common reason is that the document **isn't promoted**. Authored documents
    are reference-only until you promote them — open the document's **⋯** menu and
    choose **Promote to chat**, then look for the **"In chat"** badge. Note that Notes
    and Runbooks can't be promoted at all; only Memory notes, Skill templates, and
    Workspace instructions can.
  </Accordion>

  <Accordion title="It worked before, but the assistant stopped following my document.">
    Editing a promoted document's content removes it from chat automatically — the same
    happens if its type was changed or it was archived. Open the document, confirm the
    content is right, and **Promote to chat** again.
  </Accordion>

  <Accordion title="I can't promote a second Workspace instructions document.">
    Only one Workspace instructions document can be active per workspace. Remove the
    currently active one from chat first (its **⋯** menu → **Remove from chat**), then
    promote the new one.
  </Accordion>

  <Accordion title="My document is promoted but the assistant seems to ignore part of it.">
    Check whether it exceeds the per-session size limit. Oversized promoted documents
    are skipped whole rather than truncated, and you'll see a notice. Split long content
    into smaller, focused documents.
  </Accordion>
</AccordionGroup>

## Support

Questions about workspaces or promoting documents to chat? See [Support](/support) or
email **[admin@valiancehealth.ai](mailto:admin@valiancehealth.ai)**.