Spaces — workspaces for related chats

A Space is a named workspace that bundles everything a related body of work needs in one place: a default model, optional pinned files and folders, optional pinned prompts (some of which can be locked so they’re always attached and can’t be detached per chat), and an optional Space-scoped memory file. New chats created inside a Space inherit all of those automatically — so you stop re-attaching the same folder to every “novel” chat, stop re-pinning the same framing prompt to every “therapy notes” chat, and have one place where every conversation on a project lives.

Why use Spaces

Use a Space when you have a coherent body of work that crosses multiple chats:

• Writing a novel — pin your style guide, character sheets, and an outline folder; lock a “noir framing” prompt so every new chat opens with that voice baked in; every new chat in this Space starts with all that context.
• Job hunt — pin your resume and a folder of job descriptions; pre-attach your “cover letter writer” prompt; pick a vision-capable default model so screenshots of postings work.
• Therapy / journaling — set up a private Space-scoped memory file the model sees on every send; lock a “tone reframer” prompt so it can’t be removed by accident; pick a small, fast local model.
• Project planning — pin a folder of design docs; pre-attach your “brainstorm” and “devil’s advocate” prompts; set a Space memory file with project constraints.

Chats that don’t belong to any Space still work exactly as before — Spaces are opt-in. The default All chats view shows everything, in or out of a Space.

Creating a Space

Look at the top of the sidebar. The Spaces section shows the All chats pseudo-row and a dashed + New Space button. Click it. A small modal asks for a name and a color — type the name (any text), pick a swatch (or the ∅ to skip color), and click Create.

Your new Space appears in the sidebar and becomes active immediately. The chat list narrows to just chats in this Space (empty for a fresh Space — every chat you’ve ever had still exists, they’re just under All chats).

Filtering by Space

Click any Space row in the sidebar to activate that Space:

• The chat list narrows to chats filed into this Space.
• The Space row gets a left-edge accent bar in its color.
• + New chat creates a chat inside this Space.
• The lock icon (private chat) is hidden — private chats and Space context don’t mix because the Space’s whole point is persistent context.

Click All chats to deactivate and see your full history. Ekorbia remembers your active Space across launches, so you’ll land back where you left off.

Editing a Space

Hover any Space row to reveal a ⋯ overflow button. Click it:

• Edit settings… — opens the full Space-settings dialog (see below).
• Rename Space — change the display name (the URL-safe slug stays pinned so disk paths don’t move out from under you).
• Change color… — pick a new palette swatch.
• Delete Space — removes the Space row. Chats stay — they just move back to All chats.

The Space settings dialog

The big one. Five sections, all editable in one place:

1. Name + Color

Top of the dialog. Rename freely; color picks from the seven-swatch palette (Amber / Yellow / Green / Teal / Blue / Purple / Red) or ∅ for no color.

2. Default model

A text input. Type a model id (e.g. gemma4:latest) and every new chat in this Space will start with that model selected. Leave it empty to inherit your global default-model preference.

3. Memory file

Each Space can optionally have its own memory.md file, separate from the global one. Both are injected as system messages on every send (not just the first), so you can edit them mid-conversation and the next turn reflects the change.

Four buttons:

• Browse… — picks any .md file you want to use. Suggested path is ~/Documents/Ekorbia/Spaces/<slug>/memory.md.
• Edit — opens the file in your OS default editor. If the file doesn’t exist yet, Ekorbia creates it (with a small starter template) at the suggested path. Sets the path on the row so you don’t lose it on cancel.
• Reveal — shows the file in Finder / Explorer / your file manager.
• Clear — removes the path from the Space (the file on disk stays untouched; you can re-pick it later).

The Space memory injects after the global memory file, so it overlays project-specific context on top of your stable user facts. Inside the prompt, it’s wrapped as:

<user_memory>
… your global memory …
</user_memory>
<space_memory>
… this Space's memory …
</space_memory>

Note: in compare mode, neither global memory nor Space memory is currently injected — compare-mode sends are simpler by design. That parity may change in a future release.

4. Pinned prompts

A chip strip showing every prompt in your library. Click a chip to toggle pinning. Pinned prompts get the amber tint. Whatever’s pinned here will be auto-attached to every new chat you create in this Space — saving you the slash-trigger step in the composer.

Locking a pinned prompt

Pinned chips show a small lock icon at their right edge. Click the lock to lock the pin:

• A locked pin is always attached to new chats in this Space (same as an unlocked pin), AND
• In the composer’s prompt-chip strip on every chat in this Space, the chip’s × detach button is suppressed — the lock glyph replaces it. Locked pins can’t be removed per-chat; the user has to come back to Space settings and unlock them.

Lock the prompts you genuinely want enforced for the project’s framing — the kind of context where “removing it just for this chat” would mean the chat stops belonging to the Space’s purpose. Leave the rest unlocked so they’re still attached by default but the user can detach them on a per-chat basis when something doesn’t fit.

Locked pins still surface as chips on the user’s own messages in the transcript — so anyone reading the chat back later sees exactly which prompts were in force at send time. (For now, the chip carries the prompt name; the body isn’t snapshotted into the chat itself.)

“+ New prompt for this Space”

Below the chip strip is a dashed + New prompt for this Space button. Click it to inline-author a new prompt right from the dialog:

• A small form opens with Name (pre-filled to "<Space name> framing") and Body fields.
• Click Save & pin to write the prompt to your library and auto-pin and auto-lock it for this Space in one shot.
• Click Cancel to discard.

This is the fast path for “I want to write a custom framing prompt and force it on every chat in this project.” The prompt lives in the regular library (it’s a normal .md file under your Prompts folder) — it’s just bootstrapped with the Space already pre-locked. You can later edit, rename, or delete it like any other library prompt.

You can still attach or detach unlocked prompts per-chat via the composer’s slash-picker; the Space’s pinned set is just the starting set for new chats. Removing an unlocked pinned prompt from a chat doesn’t unpin it from the Space.

5. Pinned attachments

Add file… and Add folder… buttons open the usual file pickers; what you pick gets pinned to the Space. Every new chat in this Space will have those attachments instantiated automatically — small files inlined, large files chunked, folders walked + indexed with the embedding model — using the exact same pipeline as the composer’s paperclip / folder buttons.

The list shows each pinned attachment with its kind (FILE / FOLDER) and path. Click the × on a row to remove a pin. Removing a pin doesn’t touch chats that were already created — they keep their copies.

Click Save to apply everything; Cancel discards every change. Both buttons close the dialog.

Moving an existing chat into a Space

Right-click any chat in the sidebar. The context menu now has a Move to Space → submenu listing every Space you’ve created plus a (none) option to remove a chat from its Space. Pick where you want it; the chat moves immediately and (if the new Space is active) appears in the filtered view.

You can also move a chat out of a Space by activating the destination Space (or All chats), right-clicking, and picking (none) — same flow.

Where things live on disk

• The list of Spaces, the pinned-attachment list, and the pinned-prompt list all live in the same SQLite database as your chats (ekorbia.db in your app data directory).
• Pinned file paths are stored as-is — Ekorbia doesn’t copy or move the files themselves. If you move a pinned file off disk, new chats in the Space will toast a warning at instantiation; existing chats keep working.
• Pinned prompt slugs reference your prompt library by filename. Deleting the underlying .md from your prompts folder silently drops the pin (the row stays in the DB but read-time filtering skips it).
• The Space memory file is wherever you set it via Browse / Edit. The default location is ~/Documents/Ekorbia/Spaces/<slug>/memory.md, but it can be anywhere you choose.

What’s deferred

Two things in the Space model are kept simple in this release:

• Pinned watches — a Space can’t yet own watches the way it owns prompts and attachments. Plan to add this once the use-case shape clarifies; for now, watches are global.
• Drag-reorder pinned prompts — pin order is the order you toggled the chips on. Reordering manually is on the roadmap.