DeskPilot.psd1

@{
    RootModule        = 'DeskPilot.psm1'
    ModuleVersion     = '0.5.0'
    GUID              = 'b8f3a2d1-7c4e-4a9b-9f1d-2e6c5a0b3d77'
    Author            = 'Raimund Andree'
    CompanyName       = 'Raimund Andree'
    Copyright         = '(c) Raimund Andree. MIT licensed.'
    Description       = 'DeskPilot is a local, desktop-style web UI that fronts the ShellPilot engine to give non-technical users the full GitHub Copilot agent toolset (browse, read/write files, run commands, skills, instructions) with visible permissions and honest cost - no terminal or IDE required. The web UI is bundled in the module and served on loopback; ShellPilot and a Copilot-enabled GitHub account are required.'
    PowerShellVersion = '7.0'
    FunctionsToExport = 'Start-DeskPilot'
    CmdletsToExport   = @()
    VariablesToExport = @()
    AliasesToExport   = @()
    PrivateData       = @{
        PSData = @{
            Tags         = @('Copilot', 'GitHubCopilot', 'ShellPilot', 'Agent', 'AI', 'GUI', 'AgenticOperatingModel', 'PSEdition_Core', 'Windows', 'Linux', 'macOS')
            LicenseUri   = 'https://github.com/raandree/DeskPilot/blob/main/LICENSE'
            ProjectUri   = 'https://github.com/raandree/DeskPilot'
            IconUri      = 'https://raw.githubusercontent.com/raandree/DeskPilot/main/source/web/assets/logo-mark.png'
            ReleaseNotes = '## [0.5.0-preview0013] - 2026-08-19

### Added

- **Automatic response retries.** A Turn no longer stops at the first transient
  Engine failure or successful-but-empty response before work has begun. The new
  **Response retries** Setting controls how many extra attempts DeskPilot makes
  after the first (`0–100`, default `2`), with each wait capped at five seconds;
  `0` turns the behavior off, and Stop remains responsive during the wait.
  Retries stop as soon as any answer or Tool Activity has appeared, so DeskPilot
  never repeats a command or write behind the user''s back. The Setting also says
  plainly that failed attempts can still consume time and Copilot credits, and
  the final token and cost Usage includes every attempt the Engine recorded
  rather than only the successful one.

- **MCP servers.** DeskPilot can now attach Model Context Protocol servers, so
  the agent gains tools from programs you choose — an issue tracker, a database,
  a document store — beside its own. A new **MCP servers** tab in Settings takes
  either a command DeskPilot runs or the path to an `mcp.json` you already keep
  for another editor (both the VS Code `servers` and the Claude `mcpServers`
  shapes are read). Each server reports what it is actually doing: whether it is
  running, which protocol version it negotiated, and exactly which tools it
  contributes. Nothing is ever discovered on its own — a file that can start a
  program has to be named by you — and a server''s tools appear in the Activity
  panel as they are called, marked apart from DeskPilot''s own so you can always
  tell whose code just ran.

  Three deliberate limits, because an MCP server is somebody else''s program
  running with your privileges. **Secrets are never stored:** a server that needs
  a token names the environment variable, and DeskPilot reads the value from its
  own environment when it starts the server, so a settings backup cannot leak a
  key. **Reach can be narrowed:** *Only offer these tools* attaches a server for
  the two tools you want and keeps the rest away from the agent entirely. And a
  new **MCP servers** permission withholds every attached server''s tools for a
  job without stopping the servers — while the panel states plainly that the
  other permissions limit DeskPilot''s own tools and do not limit an attached
  server, which can bring file and shell tools of its own.

  Requires ShellPilot 0.4.0-preview0007 or later; on an older engine the panel
  says so and everything else works as before.

- **You can now watch the agent work.** The Activity panel only ever appeared
  once a job had finished, as an unordered list of what it had touched — so while
  the agent was working, the window said nothing beyond a spinner unless *Show
  the model''s thinking* was on, and the only files it named live were the ones it
  wrote. Every tool the agent uses now appears in the panel the moment it is used,
  in order: the files it reads and writes, the folders it lists, the commands it
  runs, the **pages it fetches**, the searches it makes. A run of the same kind of
  action folds into one line — *Read 6 files*, *Fetched 2 pages* — which is open
  while the job runs and closed when it ends, leaving the whole panel as a single
  line the reader can open again. The account is kept with the message, so it
  survives a reload, and a job that was stopped or ran out of its step budget now
  keeps it too — those never receive a result, so until now they showed no
  activity at all. A fetched page is also finally named by its address rather than
  by the raw instruction that requested it.

- **One way to throw away a whole review.** Reviewing a set of changed files
  offered exactly one decision — *Undo this file* — so putting a whole change set
  back meant walking every file and undoing each one. The review footer now
  carries **Discard all changes**, which puts every file listed in the review
  back the way it was at the last save and deletes the files that were never
  saved. It appears only when there is more than one file to discard, sits at the
  opposite end of the footer from **Close** so a mis-click cannot reach it, and
  always asks first — naming how many files it is about to take, listing them,
  and saying plainly that it cannot be undone.

'

            Prerelease   = 'preview0013'
        }
    }
}