0.2.1

• New app icon — Sift now has its own teal icon in the taskbar and Start menu.
• Download page — the README links a one-click "latest" installer; the app keeps updating itself after that.
• Copy a video's URL with one click from its detail page (with a "Copied" confirmation).
• The sidebar Library button now returns to the list from a video's detail page.
• Settings is now organized into tabs: General, Transcription, AI, System.
• Queue is clearer: labeled fields, "Video & Audio" / "Best available" wording, and tag suggestions as you type.
• Slides: a clear Remove slide region button when a region is set.

0.2.0

• AI-polished documents from the Slides tab: turn a talk into a dense knowledge document, with slides dropped in where they belong. Pick how it's written — no AI (raw), a local model (Ollama), or an external one.
• Use your Claude subscription — a new Claude Code CLI provider runs polish and summaries through the claude command you're already logged into, no API key. Pick a default AI provider in Settings.
• Files tab on every video: one place listing the documents, transcripts, summaries, and prompt runs created for it, each with an Open to reveal the file.
• Transcripts and summaries are now saved to disk automatically (a .txt / .md beside your downloads), not just kept in the app.
• Save slides… exports the selected frames as image files at full resolution.
• Prompt playground in Settings to tune a prompt against a transcript before using it.
• Whisper transcription now shows a progress bar.
• Home tags keep the capitalization you type; the platform badge is tidier.

0.1.0

• New Slides tab on any downloaded video: pull out the data-bearing frames (slides, charts, on-screen text) automatically, or grab the current frame while watching.
• Curate what you keep — click to include/exclude, mark a slide region to focus on the projected screen, and skip the wide room/camera shots.
• Selected slides fold into the AI summary, and you can export a document (Markdown or PDF) that lays the transcript out with each slide dropped in at the moment it appears — no AI required.

0.0.15

• Sift now installs the tools it needs (yt-dlp, ffmpeg, Deno) automatically on first run — no manual setup step.
• Those tools now keep themselves up to date: on launch, an outdated tool updates silently in the background. A toggle in Settings → Binaries switches this to notify-only if you'd rather approve each update.
• A small toast shows when a tool is installing/updating or when an update is available.
• Fixed ffmpeg update detection — it previously could never tell that a newer ffmpeg build existed. (After this update, ffmpeg may refresh itself once as it moves to the new build-date versioning — expected, not a problem.)

0.0.14

• Dropdown menus now render dark with readable text (the previous fix didn't cover the open menu list on Windows).

0.0.13

• Video thumbnails now load reliably: YouTube thumbnails are cached locally (fetched once) instead of all hitting the network at once, which could fail to load a whole page of thumbnails in a large library. Off-screen thumbnails also load lazily as you scroll.

0.0.12

• Fixed dropdown menus (format, provider, filters) and date pickers rendering with a white background and unreadable text — they now match the dark theme.

0.0.11

• Library pagination controls now stay pinned at the bottom, with no empty gap beneath them on shorter pages. The list scrolls on its own while the search and filters stay in view.

0.0.10

• When you paste a URL that's already in your library, the preview now shows what you already have — badges for the downloaded video, its transcript, and any summaries.
• Downloading an already-captured video now asks for confirmation first, so you don't overwrite an existing file by accident.

0.0.9

• Tag box on Home now suggests your existing tags as you type.
• New setting: "Get transcript after download" (on by default) — a toggle to control whether transcribing runs automatically once a video finishes downloading. Leaving it on means a caption-less video already has its file ready for local Whisper transcription.