chore: version packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and publish to npm yourself or setup this action to publish automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@sapiom/harness@0.11.0

Minor Changes

• fc231c9: Studio: creating an agent in a project is now a form that creates it, not a
  message asking your coding agent to.

  Every create door ended the same way — a session started and an English
  sentence was typed into the terminal asking the coding agent to please call a
  scaffold tool. Studio never created anything, so a failure arrived as a
  confused model rather than an error, and "did it work?" could only be answered
  by reading a terminal. On a project that already held agents it simply could
  not work: the scaffold was aimed at the project folder, which is not empty, and
  the reply was a paragraph asking you which subdirectory you meant.

  • Create an agent in {project} opens a small dialog: a name, a starter, and
    the project it lands in — stated, not asked again, because you clicked that
    row. Submit and Studio creates the agent itself.
  • Creation completes before the chat starts. The agent is on disk and in
    your rail before a session opens on it, and the rail scrolls it into view. A
    first instruction is optional, and the session opens on that instead of on a
    request to scaffold.
  • A refusal is a sentence in the dialog. A name already taken in that
    project, a name that is not a folder name, a folder Studio does not show as a
    project — each is refused with a reason you can act on, and nothing
    half-created is left behind if the scaffold itself fails.
  • The bundled starters in the template gallery take the same path, so the
    two ways of starting from a starter cannot drift. Cloning a published
    template still goes through your coding agent, which is a different operation
    with a different failure mode.
  • The empty-project row's Create the first agent here now responds to a
    click; it had been unclickable.
  • Starting from an idea on the home screen is unchanged: no project, no name,
    and a folder that does not exist yet, so it stays a conversation.

  New endpoint POST /api/agents/scaffold — { root, name, template? } → the
  created agent's path. It runs the same scaffold the CLI does and refuses on its
  own findings rather than on the caller's word.

• e72d172: Studio: selecting a project fills the workbench instead of replacing it — your
  conversation stays, and the project's map draws beside it.

  Picking a project used to collapse the shell to one column and take both panes
  away: the map arrived as a full-width destination and the chat you were mid-
  sentence in vanished. Browsing a template gallery is a detour; looking at your
  project's shape while you talk to it is not. Now the conversation keeps the
  centre and the map fills the right pane.

  • A project is somewhere you work. Selecting one makes it the conversation's
    subject: the tab strip shows that project's live sessions — including the ones
    bound to its agents, which were invisible before — and a project with nothing
    running is given a session at its root.
  • Selecting an agent inside the project changes only the right pane. The
    session already reaches every agent in its project, so the chat does not move.
    Crossing to a different project still hands the conversation over.
  • The map and an agent's board are one canvas at two heights. Clicking an
    agent on the map drills into its board, and a control at the head of the tab
    row cuts back up to the map — it used to be a one-way door. The rail selection
    and the canvas always agree, driven from either side.
  • Tabs are now Canvas and Steps. Steps is disabled at map altitude with the
    reason, since a whole project has no step list. The Code tab is gone; its
    "Trigger from your code" snippets moved to the Steps surface, under the deploy
    banner — the moment that question is actually asked.
  • The active session's tab is kept in view when the strip's own width changes,
    so a newly created tab can no longer be stranded off-screen.

  An agent that is linked to Sapiom but has no ready cloud build yet gets the
  snippets section with the reason there is nothing to copy, rather than nothing
  at all. Snippet clicks are now attributed to the canvas surface rather than a
  Code-tab surface that no longer exists; object=snippet is unchanged.

  Membership stays derived from where an agent lives on disk. Nothing is stamped
  onto a session record, so moving an agent or removing a project cannot leave a
  second, staler answer behind.

• 5745707: Studio: the project map draws your groups as named containers, instead of one
  endless column of agents.

  A project map used to draw every agent it contained as one flat set, ignoring
  the structure the rail was showing right beside it. Open a folder holding
  several systems and you got a single vertical column — a folder holding a few
  dozen agents came out thousands of pixels tall and one card wide, which no
  amount of zooming out makes readable.

  • One labelled container per group. The map now reads the same groups the
    rail does, so a system you named in the rail is a system you can see on the
    map, under exactly that name and in the same order. Rename or regroup in the
    rail and the map follows immediately.
  • Ungrouped is a container too. "No connections detected" is a real answer
    about a project, not an absence, so it gets a labelled box rather than being
    scattered loose. A project with one group renders as one group.
  • Agents with no connections wrap instead of stacking. Inside a container,
    unconnected agents fill the width and wrap, which is what replaces the column.
  • A connector between two containers is still drawn. Splitting a system
    across two groups does not make the link between the halves disappear; it is
    drawn dotted, as a link between systems rather than wiring inside one.
  • Container names stay readable as you zoom out, which is the moment you are
    looking at the whole project and need to know which system is which.

  Nothing about how groups are detected, edited, or stored has changed. The map is
  a new reader of the arrangement your project already has; it never writes one.

• e5af82b: Project rows in the rail now carry one action control instead of two glyphs that acted on different things. The + created an agent inside the project while the × removed the project itself — same size, side by side, so + read as "add project". Both now live behind the row's ⋮, where each action names its own subject ("Create an agent in ", "Remove from the rail").

  Double-clicking a project or folder label now folds and unfolds it, the way the chevron always has. On a project row the single click still selects, so a double-click leaves the project both selected and folded.

  A one-time card explains what projects and agents are and what happens to your rows on upgrade. It is dismissible, shown once, and re-openable from the account menu.

Patch Changes

• 1d2ae02: Stop an agent that can never be named from holding a whole project's graph at
  "Graph may be incomplete".

  An agent that does not declare a name has its source read to recover one. If
  even one agent in a project could not be read, the projection refused to cache,
  so the project reopened as degraded under the "Graph may be incomplete" banner
  every time. A companion package with no agent export in it has nothing to find
  no matter how often it is re-read, and a single such directory was enough to
  mark everything around it incomplete.

  The cache now depends on whether the name lookup can still produce a different
  answer, not on whether it found a name. Only one case is treated as final, the
  one that can be proven: a project with no TypeScript in it has nothing for an
  agent name to be declared in, and no install or re-run will invent one. Every
  other failure still blocks caching, so that project keeps its banner and its
  Retry button — the thing that clears it once dependencies are installed, or
  after a check that ran out of time succeeds on a second attempt.

  The affected agents keep their warnings either way. The graph stops calling
  itself incomplete over agents it was never going to resolve; it does not go
  quiet about what it could not resolve.

  Also stops one agent being registered twice when Studio reaches the same
  directory by two different paths — once as given and once with symlinks
  resolved. The duplicate pair made every reference between agents ambiguous,
  silently dropping those connections from the graph. This prevents new
  duplicates; a workspace that already contains a pair from an earlier version
  still needs them removed by hand.

@sapiom/agent-studio@0.1.20

Patch Changes

• Updated dependencies [fc231c9]
• Updated dependencies [1d2ae02]
• Updated dependencies [e72d172]
• Updated dependencies [5745707]
• Updated dependencies [e5af82b]
  • @sapiom/harness@0.11.0

@sapiom/cli@10.0.0

Patch Changes

• Updated dependencies [fc231c9]
• Updated dependencies [1d2ae02]
• Updated dependencies [e72d172]
• Updated dependencies [5745707]
• Updated dependencies [e5af82b]
  • @sapiom/harness@0.11.0

@sapiom/harness-desktop@0.4.1

Patch Changes

• Updated dependencies [fc231c9]
• Updated dependencies [1d2ae02]
• Updated dependencies [e72d172]
• Updated dependencies [5745707]
• Updated dependencies [e5af82b]
  • @sapiom/harness@0.11.0