Capture work by describing it, with screenshots and recordings
Triage capture is now a single natural-language composer instead of a title field and a note. Describe the issue, the idea, or the work in your own words, and attach whatever evidence you already have. You can paste a screenshot straight from the clipboard, paste files copied in the file manager, drag and drop images, videos, logs, and documents, or pick them from a file dialog. Every attachment shows its name, type, size, and upload state, with a progress bar for large files, plus a retry for a failed upload and a cancel for one still in flight. Videos are first-class evidence: MP4, WebM, and MOV are accepted up to 200 MB each, streamed to storage rather than held in memory, and served with range support so a recording plays and seeks in place while you review it.
If the voice add-on is enabled, a microphone sits inside the composer and reuses the same recorder, transcription settings, and permissions as the existing voice capture, so there is one voice path rather than two. The transcript is inserted at your cursor for review and never submits anything on its own, and the global voice hotkey dictates into an open composer instead of starting a competing capture. Pekan records whether a report was typed, dictated, or both, so the Inbox shows how it was made. Recordings exist only long enough to be transcribed: audio is never stored or attached.
Nothing is lost when a capture is interrupted. Evidence is saved before anything interprets it, so closing the panel, restarting the app, or a failed AI run all leave the report and its uploads recoverable, and an abandoned draft is cleaned up along with its files. Submitting is idempotent, so a double-click, a retry, or a resubmitted restored draft produces exactly one Triage item.
Drafting with AI now reads the evidence. Pekan hands the selected provider only the files it can genuinely open, names that provider and lists those files before anything is shared, and lets you decline. The proposal reports what was actually read: a run that claims to have watched a video it could not open is corrected, and unreadable or unsupported media is surfaced as a coverage gap rather than quietly implied to be understood. The original report and its screenshots stay on screen beside the proposal, every generated field remains editable, and you can answer a clarifying question or add an instruction to regenerate while keeping what you already have. Accepting moves the same record onto the board, so attachments and capture provenance stay intact.
Captures still work without AI. With no provider available, or after a failed run, the original report and all of its evidence are retained for manual review or a later retry.
Highlights
- One natural-language field replaces the title-and-note capture form.
- Paste screenshots, paste copied files, drag and drop, or pick files.
- MP4, WebM, and MOV up to 200 MB, streamed and playable in review.
- Dictate into the composer and edit the transcript before capturing.
- Drafts survive a close, a restart, and a failed AI run.
- The provider and the exact files are disclosed before anything is shared.
- AI proposals report what was actually read instead of implying more.