Backlog

Next up

In Progress

Done

Main Roadmap

Again with the idea of ‘scanning’ documents (mainly in PDF format) to find out exactly what they contain and then rename them accordingly. Option: also retrieve the name of the establishment, the date, etc. if it concerns invoices, for example.

In Review

Rename files according to their content

Thierry Weber

It would be great to support some sort of traits or persistence memory. For example I would like to specify some common useful directories or commands or scripts. I could write smth like this ~/code - it’s the directory of my git repos ~/pet-proejcts - it’s the directory of my personal projects ~/job - it’s the working directory So the Substage could understand some commands like “go to my working directory and open the main project in VSCode”

Traits support

Daniil Voidilov

If you have an item on the desktop selected, using the Substage keyboard shortcut could show a floating panel at the bottom of the screen

Allow using directly on Desktop

joethephish

Currently to use local LLMs via LM Studio or Ollama. While this works and is relatively straightforward, some people would prefer to have a more streamlined “batteries included” approach. The design would be that you need to download an optional extra model via the settings (we would recommend a model so it’s one click), and then it would “just work”.The compromise is that it would take up rather a lot of your computer’s memory for a supposedly lightweight “simple” task.

Built-in local LLM support

When I use shortcut keys to toggle prompt window, the default folder that open is, /Users/username/Is it possible to add a feature to select default open folder, for example /Users/username/Downloadsor /Application/DropboxThank you

default folder feature

Pikachu

for the touch typists here, it would be really great to be able to see view the full command output, the info, etc viewable with keyboard shortcuts. Every part of substage should be accessible via shortcuts. This would overall make it safer as well as I could more conveniently check things before taking actions on files. Also could help with more detailed analysis type of prompts where I want to read a longer summary that was output. Should be as easy as triggering a shortcut.

customizable shortcuts for EVERYTHING

MPH Sound

Currently the only way that LLMs are used, is to convert natural language prompts into Terminal commands, and then to summarise the output of that command.However, if we add support for arbitrary API calls (assuming user is using their own API key), then we could allow stuff like “transcribe this audio”, using OpenAI’s Whisper API. Behind the scenes it would make a <code>curl</code> call, and make use of the API key that the user has already supplied in Substage’s settings.

Add direct support for AI provider API calls

From a user: “It doesn’t play well with a very popular app I use (also on Setapp), Default Folder X (DFX). The critical thing is DFX’s “drawer” utility which, when enabled, pops out at the bottom of the Finder window. With Substage running the DFX drawer can’t pop out when I try to enable it.”

Integration with "Default Folder X"

Integrate with alternative file browsers and apps

Some people would prefer a different presentation to the docked bar, for example if they feel it takes up too much space, or would prefer to interact by right-clicking on files and choosing a contextual menu option.We could consider the ability to dock and undock, so you could drag it on/off a Finder window and resize it.

Alternative presentation to the full width docked bar

Basically a built in version of “use cases” from the <a target="" rel="noopener noreferrer nofollow" class="link" href="https://selkie.design/substage/">bottom of the main website</a>, but allows people explore what’s possible directly from the prompt barSuggested by a commenter <a target="" rel="noopener noreferrer nofollow" class="link" href="https://www.reddit.com/r/macapps/comments/1jl4kz0/comment/mk0rgiv/">on Reddit</a>

Planned

Add a dropdown to select common use cases and demonstrate what's possible

Currently there’s no built-in command line tool on macOS for reading plain text from PDFs (for summarisation etc), as far as I know. Textutil can be used for reading from RTFs, Word Docs, HTML and more, but not PDF. We’d have to embed some helper binary in the same way that we do for ffmpeg.

Add support for reading plain text from PDFs

<img data-width="100%" data-align="left" data-featurebase-content-key="67ab1e0362151453820f2c42/post/681b75945094044ac40025c8/0196ab37-1165-7d52-92d5-4d5bc29724ee/b64u-aW1hZ2UucG5n.png" src="https://67ab1e0362151453820f2c42.featurebase-attachments.com/c/post/681b75945094044ac40025c8/0196ab37-1165-7d52-92d5-4d5bc29724ee/b64u-aW1hZ2UucG5n.png?X-Amz-Expires=3600&amp;X-Amz-Date=20250712T080000Z&amp;X-Amz-Algorithm=AWS4-HMAC-SHA256&amp;X-Amz-Credential=DO0098AADP7QLUJN4ZLF%2F20250712%2Ffra1%2Fs3%2Faws4_request&amp;X-Amz-SignedHeaders=host&amp;X-Amz-Signature=c1b660333295d6f47a6214df606e75dc3eb67093dd44a8d49c9609cbd6203ced" data-featurebase-content-filename="image.png">It has 12 possible states, and yet the descriptions only state 6. <ol><li>Automatically runs high risk commands (e.g. deletions, potential data loss). You feelin' lucky, pun</li><li>Automatically runs moderate risk commands where data loss is possible if mistakes are made (e.g. file modification</li><li>Automatically runs certain modifications (e.g. renaming files)</li><li>Only automatically runs benign, informational or basic modification commands (default)</li><li>Only automatically runs benign or informational commands</li><li>Substage will always ask for confirmation</li></ol> I believe a checklist could be an alternative. (exact appearance idrk) Ex: ☑️ Always keep the original file (Any changes prompted to the file will result in a new file that has the modifications☑️ Always run with confirmation (If this is checked, all other boxes and text will be greyed outAutomatically run the following commands:<ul><li>☑️ Benign or informational commands</li><li>☑️ Basic modification commands</li><li>☑️ Renaming of files</li><li>☑️ Moderate risk commands where data loss is possible (e.g. file modification)</li><li>☑️ High risk commands including deletions</li></ul>

Confusing risk level slider - revamp (possibly into checklist?)

lionheo9

If we bundle this and add training examples, it should work well: <a target="_blank" rel="noopener noreferrer nofollow" class="link" href="https://github.com/jdberry/tag">https://github.com/jdberry/tag</a>

Add support for Finder tags

So the way I want and plan to use substage is to toggle the command bar regularly in order to make new actions or prompt new queries on selected files or folders WHEN I need them. The animation of it dropping and bouncing down is very very cool, but after a while I really don’t find it necessary, and would be happier with it being instantaneous. By all means keep it but would very much like an option to disable it.

Option to disable the toggle animation.

Sometimes summarisation can take some time, particularly if using a local LLM and if it has to summarise a large amount of content. So it would be good if we made the command output icon visible as soon as soon as the command starts generating output. If more than ~1-2 seconds pass while waiting for summarisation, we could even show a (truncated) version of the actual output before it gets replaced with the summarised version…

Make command output available immediately, instead of only after summarisation

As suggested/reported by Joe F (mergesort):<ul><li>He’s not keen on the overall alignment and organisation of all buttons/inputs (it’s true that in certain states there’s a lot of icons, they have no hover states, and perhaps they’re not grouped effectively enough).</li><li>Joe would prefer a more standard left aligned prompt field, he doesn’t like the way it needs to slide while you type. This is fair enough, we should implement this at the very least as an option.</li></ul>One small thought of my own: could group settings button with model picker, since it’s settings-y. (I don’t think we should remove the settings button entirely since many people don’t even realise the menu bar icon exists).

Refine UI layout

As reported by Joe F / Mergesort<ul><li>“There seems to be a bug where I wouldn't be allowed to ⌘Tab to apps on another space when Substage was active, it would just shoot me back to the space I was on. My exact configuration is: 4 spaces, 2 full screen apps, and Finder is set active to all spaces.”</li><li>Substage window not moving between finder windows (sometimes? when spaces active?)</li></ul>

Buggy when used with Spaces

Currently the official binaries of ffmpeg and ffprobe are intel only so to use it you need to have Rosetta enabled.There are static binaries here:<a target="_blank" rel="noopener noreferrer nofollow" class="link" href="https://gitlab.com/martinr92/ffmpeg">https://gitlab.com/martinr92/ffmpeg</a><a target="_blank" rel="noopener noreferrer nofollow" class="link" href="https://ffmpeg.martin-riedl.de/">https://ffmpeg.martin-riedl.de/</a>And you can create universal binaries like this:lipo -create path/to/x86_64/executable path/to/arm64/executable -output path/to/universal/executable

Include universal binaries of ffmpeg and ffprobe

It would be great if you could allow an optional (large!) download of an open source LLM model. We’d have to investigate the performance of such a model, both in terms of latency and accuracy. Potentially there could also be a cheaper subscription plan if you only use offline model(s) since Selkie wouldn’t have any per-user monthly overhead.

Completed

Support offline LLMs

I'd like to request a lifetime purchase option for the app. I, and likely multiple other users, are already managing many subscriptions, and a one-time payment would be a very welcome alternative.

Lifetime Purchase Option

Usman

Very very often the prompt returns with a follow-up question for more clarification, no way to answer it currently. Only thing I can do is rewrite the original prompt with hopefully enough clarification for it to put together the action.

Must-Have: Follow-up on Prompt

As with the idea to support offline LLMs, we could potentially offer a different pricing option if Selkie doesn’t have any ongoing overheads.

Allow usage of user's own OpenAI etc API keys

Currently the LLM-generated commands cannot be edited (or even selected!). It would be great if it was possible for a user who knows what they’re doing to tweak or correct a command before it gets run.And/or, assuming we create a bookmark feature for reusable commands, it should be possible to edit them before saving them, of course.

Editable generated commands

Something useful to me would be to be able to type something like “select all the CSV files in this folder” or “select all the files I created yesterday” or “select all the files with ‘quote’ in the filename” - currently, it reports a count and/or has an icon that you hover over to see a list of files.With all the files selected, I could then issue a command like “move the selected files to the folder named ‘Import Files’”…or “drag to the Trash” (with warning based on danger level of the command, of course).

Select/Perform actions on files

Julian Daniel

Currently you can: <ul><li>Use keyboard shortcut to show/hide Substage. This also tries to be smart, so if you’re in another app it’ll first switch to Finder+Substage, and only on second press (if Substage is fully front/active) will it actually hide.</li><li>Press Return to confirm, Escape to cancel/clear.</li><li>Press up/down to cycle through command history</li></ul> Areas for improvement:<ul><li>BUG: Sometimes when activating Substage, the prompt field isn’t selected?</li><li>Add a way to jump back and forth between Substage and its attached Finder window.</li></ul>

Improve keyboard support

This could be a bad idea but how about LLM memory/history? As in you can chain requests without having to specify every time. Pressing enter would work as usual, but something like option enter / control enter / etc would let you chain requests in the same 'chat'. For example, File A and File B exist Selecting File A, requesting "make into MP4, 1080p" followed by selecting file B, pressing enter, and requesting "this too" will result in a "No request provided" However, Selecting File A, requesting "make into MP4, 1080p" followed by selecting file B, pressing dedicated hotkey (or a continue button next to the return button on the app), and requesting "this too" will result in file B also getting turned to a 1080p MP4 video. Not sure how this will work with online models and API keys, but for local ones, while the model does get unloaded like how it works rn, request history will shortly be kept on cache until you press enter(which resets to default) An additional option which defaults to keeping conversational data (unless a dedicated hotkey is pressed that resets to default) could be interesting.

LLM (conditional) conversation history awareness

The LLM doesn’t currently know when selected items are in fact folders. We could be smart about the way we prompt by:<ul><li>referring to selected items as folders when they are (when they’re exclusively folders?)</li><li>Putting trailing slashes on folder paths to make it even clearer</li></ul>

Distinguish selected files from selected folders

Currently there’s a 20 file limit because of the way that Substage functions, passing in the names of all the individual files to a command - it would force the LLM to create a potentially extremely long Terminal command with a large number of files, and is likely to break down at some point. What if you have 10,000 files selected in a directory of 11,000 files, and want to zip them? Probably Substage isn’t the right tool for the job! But is there anything we can do to improve behaviour?