feat: sync ! command autocomplete with upstream MindRoom bot commands#3008
Closed
basnijholt wants to merge 18 commits into
Closed
feat: sync ! command autocomplete with upstream MindRoom bot commands#3008basnijholt wants to merge 18 commits into
basnijholt wants to merge 18 commits into
Conversation
* fix(message-extras): allow safe inline svg Permit inline SVG primitives in MindRoom message extras while explicitly stripping script-bearing SVG vectors, external references, animation tags, style attributes, and event handlers. * fix(message-extras): harden svg sanitizer review gaps * fix(message-extras): preserve svg aspect ratio Allow preserveAspectRatio through the message extras SVG sanitizer with explicit camelCase restoration so safe inline charts can control viewport scaling without reopening URL or script surfaces. * docs: record safe svg rebase validation
Voice recordings in the Cinny PWA were uploading at the browser's default MediaRecorder bitrate, which on iOS Safari produced AAC at roughly 188 kbps and ~767 KB for a 32-second clip. On slow mobile connections that file size is annoyingly large for what is supposed to be a quick, casual voice note. The fix sets `audioBitsPerSecond: 32000` on the MediaRecorder constructor and requests constrained `getUserMedia` input before recording: mono capture, a 24 kHz speech-oriented sample rate, plus echo cancellation, noise suppression, and auto-gain control. If a browser rejects those mic constraints with OverconstrainedError, recording falls back to unconstrained audio capture so the user is never blocked. The fallback path also re-checks the recording session id so a hook that unmounted during the initial permission prompt does not trigger a second mic prompt on a dead component. The bitrate value is Signal-equivalent for voice and keeps Opus essentially transparent on Chromium browsers while still materially shrinking AAC output on iOS. The mono channel-count constraint is the load-bearing iOS win, since WebKit historically has been inconsistent about respecting `audioBitsPerSecond` for the AAC encoder; mono input compresses more efficiently regardless of whether the bitrate hint is honored.
* Use WebGL background for splash screens * fix(splash): address WebGL background review * refactor(splash): rely on default particle background
* Fix config retry click handling * Address config retry review feedback
* Send attachments before the caption text (CINNY-108) Invert the room-input send-session ordering so attachments send first in selection order and the caption text sends last. MindRoom's backend coalesces a media batch into one agent turn and closes it on the trailing text event (mindroom#1255), so the caption must arrive after the uploads. - Drop the auto-thread-text-root mode: text plus attachments now uses the upload-root shape (first attachment is the room-level root, remaining attachments and the caption thread under it). - The caption threads under the upload root; reply metadata stays on the root upload. - Reset the composer when the session snapshots the caption instead of when the text event sends, since the caption can now wait behind slow uploads. - Mock components/editor/utils in RoomInput.test.ts: the controller imports the reset helpers from there, and its real resetEditor call crashed against the stubbed slate Editor (previously swallowed by the send-step catch). * Restore a failed caption into the composer (CINNY-108) With media-first ordering the caption usually fails only after every upload has sent and left the board, so the board's Send-again affordance (the only way to resume a blockedRoot session) no longer existed: the caption was silently lost, and the zombie session would flush the stale caption into a later unrelated send while swallowing that send's input. - The send-text failure path no longer sets blockedRoot; it restores the composer from a snapshot of editor.children taken at session start (cloned, since slate mutates children in place), clears textPending, and lets the session finish. Upload retries stay resumable; the session never resends a restored caption. - New restoreEditorContent helper in components/editor/utils prepends the snapshot, drops the placeholder paragraph when the composer was empty, and keeps anything typed since the snapshot. - Controller tests cover composer restore plus session completion without stale resends, and upload-retry resume after a failed caption. - Wrap the over-printWidth lines this branch introduced and shorten two test titles (repo prettier is pinned at printWidth 100).
Mirror the upstream MindRoom command set (mindroom/commands/parsing.py) in the frontend autocomplete list: - Add missing commands: !model, !thread_mode, !reload-plugins - Remove stale entries: !skill (removed upstream, parses as unknown) and !widget (does not exist upstream or client-side) - Order entries to match the order !help lists them - Add a test guarding that every syntax string starts with its name
|
I have read the CLA Document and I hereby sign the CLA You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot. |
Author
|
Opened against the wrong repository by mistake — this belongs on the mindroom-cinny fork. Apologies for the noise. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Makes the MindRoom
!command autocomplete in the composer mirror the actual command set of the upstream MindRoom bot (mindroom/src/mindroom/commands/parsing.py,CommandType+_COMMAND_DOCS), so every command the bot understands is autocompletable and no dead commands are suggested.!model [name|list|reset],!thread_mode [room|thread|reset|show],!reload-plugins!skill— explicitly removed upstream; parses asUNKNOWN(see upstreamtest_removed_skill_command_is_unknown)!widget— does not exist upstream and has no client-side handler!helplists them (upstreamCommandTypeenum order)syntaxstring starts with!<name>, guarding that autocomplete-inserted text stays validAdmin-only commands (
!reload-plugins,!config,!thread_mode) are kept in the list since the frontend has no admin-status signal and the bot replies with a clear permission error.Validation
npm run typecheck✅npm run build✅npx eslint src/app/mindroom/commands/✅npx vitest run src/app/mindroom/commands/✅ (6 tests)reload-plugins) matches the upstream regexes aftermessage.strip(), and no listed command parses asUNKNOWN.