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-preview0017] - 2026-08-24 ### Added - **Image previews instead of "there is no text to compare."** A screenshot the agent saved, a logo you dropped into the Project, an icon that changed — the Diff viewer and the file viewer now *show the picture* for the formats a browser can draw (PNG, JPEG, GIF, WebP, BMP, ICO, AVIF), captioned with what happened to the file, on a checkerboard so a transparent image reads on either theme. Anything else still says plainly that it has no text to compare. There is no before-and-after for a picture: the preview is the file as it stands now. Backed by a new `GET /api/fs/image`, confined to the selected Project exactly like the file tree. It is the file''s **own signature bytes** that decide the media type it is served under, never its extension, so a file named `.png` that is really a web page is refused rather than handed to the browser. SVG is deliberately not previewed — it is script-capable markup, and being text it is already readable as text. Every response now also carries `X-Content-Type-Options: nosniff`. - **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. ### Changed - **A Turn is now laid out in the order it happened, with the answer last.** The reasoning trace used to be printed as one block above the reply, so the answer you were waiting for was buried under thousands of lines explaining it, and the box stayed open forever once the job was done. DeskPilot now shows the Turn the way GitHub Copilot Chat does: each run of thinking gets its own box, placed at the point it happened, and whatever the model said before that run stays above it rather than being collected somewhere else — so neither the reasoning nor the reply piles up on one side of the other. Each box folds to a single line the moment its run ends, labelled with how long it took, and any of them can be opened again afterwards. A run streams open while it is happening, so nothing is hidden while you wait for it, and a box you opened yourself stays open rather than being shut under you when the run finishes. The complete answer is always rendered in full at the end, below everything else. ### Fixed - **The end of a long answer is reachable again.** The thread could stop scrolling with a bar apparently already at the end while more of the answer was still below — reachable only by dragging a text selection downwards. The bar at the end belonged to the *thinking box*, not the conversation: while a run of thinking streams, that box sat in the middle of the message with the answer and the Activity panel below it, and it scrolled on its own — so a mouse wheel over it moved the box and never reached the conversation behind it. A run that is still streaming no longer scrolls separately; it shows its newest lines and passes the wheel straight through, and it is shorter, so it no longer fills the window on its own. A finished run still scrolls when you open it, because nothing is moving then and you asked for it. Two smaller causes of the same thing were fixed with it. The window follows the newest output by re-aiming at the bottom on every streamed frame, and because that scroll was animated, an animation restarted that often never arrived — it trailed the content instead of landing on it. And the checks that run once a job finishes, filling in the checkpoint markers and auto-compaction rebuilding the conversation, make the thread taller *after* the last scroll. Following is now immediate, and the window follows once more after everything that finishes a job. The one scroll that should glide, jumping to the thinking box from the status line beside the spinner, still does. - **Attaching a photo no longer fails the whole message.** A camera-sized image attached to a Conversation ended the Turn at once with *Request Entity Too Large* and nothing in it to act on, while the same picture worked in Copilot Chat. DeskPilot was sending the original file at full resolution, and an image is written into the request encoded — about a third larger again — so one modern phone photo could overflow the request on its own, and attaching the same one twice made it certain. An image too large to send is now scaled down to the resolution the Models work at before it is uploaded, and DeskPilot says how many it resized rather than changing your file silently. An image that is already small enough is never touched, a PNG stays a PNG so screenshot text is not smeared, and an animated image — GIF, animated WebP or APNG — is left alone rather than being flattened to a single frame. If an oversized image still arrives — over Intercom, which has no browser to resize it — it is now refused up front with a message naming the file and its actual size, instead of costing a round trip and coming back as an unreadable error. A Turn refused this way hands your prompt and your Attachments back to the composer instead of discarding them. - **Pasting a file into the chat is no longer slow.** Attaching a screenshot or a file with Ctrl+V left the composer waiting seconds before the attachment chip appeared, and the delay grew with the size of the file. Two things inside the upload handler were responsible: the request body was scanned for its part boundaries one byte at a time, and the file''s own bytes were then copied into a structure that stored every single byte as a separate boxed object — several million of them for one pasted screenshot. Both are gone; a paste that took well over a second of server time now takes a few milliseconds. A binary upload is also no longer decoded as text on arrival, which was a second full pass over every byte for a string nothing ever read. - **Sta' Prerelease = 'preview0017' } } } |