# Changelog

## 2.0.21

- Add public Facebook videos, Reels and share/r or share/v links using the same
  exact-quality selection, measured-size estimates, live streaming and MP3 flow.
- Show Facebook metadata and provider-specific errors without attempting to
  bypass private, login-only, group, LIVE or photo restrictions.

## 2.0.20

- Display measured Instagram source sizes supplied by the worker instead of
  "Size unknown" when the CDN exposes a reliable byte count.
- Include audio in merged-video size estimates and mark the result with ≈;
  only a finished muxed/converted file has an exact final size.

## 2.0.19

- Add public Instagram Reels and single video posts through the hosted backend,
  with original quality selection, live streaming and MP3 conversion.
- Show Instagram creator/duration and clear errors for unavailable or restricted
  posts. Private posts, Stories, photos and carousels are not supported.

## 2.0.18

- Replace misleading printed prompt instructions with a real recovery flow:
  backgrounded query URLs pause, `fg` brings bdm forward, then an active prompt
  reads the complete pasted URL before any API request is made.
- Never display a paste-question example at an idle shell prompt.

## 2.0.17

- Stop before inspection when a backgrounded query URL may have been split by
  the shell, rather than automatically downloading with missing parameters.
- Guide users to the quote-free paste prompt: "Paste the full YouTube or public
  TikTok link, then press Enter (no quotes needed):".

## 2.0.16

- Warn immediately when an unquoted `&` backgrounds the command, skip terminal
  menus safely, and reject background terminal reads before contacting the API.
- Recheck terminal ownership after inspection and regression-test real Bash
  job control, quoted arguments and the quote-free paste prompt.
- Move the default API to `https://download-me.up.railway.app/api/v1`.
- Automatically migrate saved Vercel and temporary Railway API settings while
  preserving custom backends, output folders and quality preferences.

## 2.0.15

- Handle links pasted without quotes. An unquoted `&` no longer leaves the job
  half-stopped on the format menu; `bdm` detects a backgrounded run, explains the
  shell split, and picks the recommended format instead of blocking on a read.
- Run `bdm` alone to paste a link safely, pipe it with `bdm -`, or use
  `bdm --paste` to read the system clipboard — no quoting required.
- Clean stray quotes, backticks, angle brackets, markdown links and zero-width
  characters from a pasted URL without changing the address itself.

## 2.0.14

- Remove the CLI's fixed 2 GB transfer ceiling and default size-based format
  exclusions. The service now defaults to no duration or per-file size limit.
- Preserve verified HTTP byte-count checks, retry safety and partial cleanup.
- Long/large downloads still depend on available storage, source access and
  network conditions; "no fixed limit" is not a promise of infinite resources.

## 2.0.13

- Public TikTok video/share links use the same hosted streaming and resumable
  download pipeline as YouTube when enabled by the service.
- Show TikTok creator/duration and accurate portrait dimensions instead of
  misleading vertical-resolution labels.
- Explain unavailable, private or upstream-restricted TikTok posts clearly.
- No extra local tools required. Private posts, LIVE and photo slideshows are
  not supported; source quality and watermark removal are not guaranteed.

## 2.0.12

- Recover dropped live transfers using the same cached stream's ETag and the
  number of bytes actually saved to disk. Never append mismatching ranges.
- Honour retry delays while a merged file establishes its final byte count.
- Keep one live connection by default: parallel ranges did not reliably beat
  the measured source bottleneck and introduced additional dropped connections.
- Interrupted/expired transfers still clean up partial files; Ctrl+C cancels
  immediately rather than waiting for automatic retries.

## 2.0.11

- Download directly from live worker streams; no complete file/R2 preparation
  and no local Python, yt-dlp, ffmpeg or Node installation for standalone users.
- Keep exact source totals when available; show clearly marked size estimates
  for live muxes, never pretend an estimate is a verified HTTP length.
- Cancel interrupted live jobs and remove partial files before reporting failure.
- Embed Node 22 in standalone Windows x64, macOS Intel/Apple Silicon, and Linux
  x64/ARM64 builds. CI tests real downloads with external tools removed from PATH.

## 2.0.10

- Download directly from the worker when the service serves files itself,
  instead of only accepting object-storage redirects.

## 2.0.9

- Show honest preparation stages (`Merging video and audio`, `Uploading finished
  file`) and a clearer `Waiting for a download slot` message instead of hiding
  the wait behind `Finishing audio/video`.

## 2.0.8

- Read the live output limit from `GET /api/v1/limits` (2 GB fallback) instead of
  hardcoding 500 MB, so larger files are offered and the limit can change
  server-side without a CLI release.

## 2.0.7

- Accept signed Cloudflare R2 artifact redirects (`*.r2.cloudflarestorage.com`)
  in addition to Vercel Blob, so downloads keep working after the storage
  migration.
- Use the backend's advertised `X-Artifact-Bytes` header instead of a `HEAD`
  probe. A signed R2 GET rejects `HEAD`; without this the parallel range save
  silently fell back to a single connection.

## 2.0.6

- Save large files using eight parallel byte ranges by default; tune with
  `-x, --connections` (1–16). Failed ranges retry safely before stream fallback.
- Smooth coarse preparation progress without racing to 99% before upload ends.
- Show real backend preparation/upload progress and retain exact selected quality.

## 2.0.5

- Remove the remaining `save`-era compatibility shims: `SAVE_*` environment
  fallbacks, legacy package-name handling in the install/uninstall scripts, and
  the legacy CLI document. Only `BDM_*` variables and `bdm`/`bdm-cli` names are
  recognized.

## 2.0.4

- Rename the public npm package to `bdm-cli` and the installed command to `bdm`
  (the standalone archives, .deb and .pkg are now named `bdm-*` and install
  `bdm`). There is no `save` command alias in 2.0.4.
- Link npm users to the browser documentation at the Download Me website.
- Document the current 60-minute source-video limit.

## 2.0.3

- Remove the session/CSRF bootstrap. The backend now scopes anonymous jobs to the
  request IP, so the CLI is stateless and works across backend instances.
- Keep `save doctor` as a connectivity check against the public statistics route.

## 2.0.2

- Connect to the verified production backend by default
  (`https://download-me-api.vercel.app/api/v1`); `--backend`, `SAVE_BACKEND_URL`
  and saved `backendUrl` still override it.
- Follow the backend's artifact redirect to Vercel Blob, accepting only HTTPS
  `*.blob.vercel-storage.com` locations so files stream without a second hop
  through the API.
- Raise request timeouts and keep polling while a healthy job is still preparing
  so slow first-time preparation no longer reports a false timeout.

## 2.0.1

- Match the web picker: one compatible format per resolution, one audio source,
  and an MP3 action, with exact/estimated source sizes and no codec duplicates.
- Add animated terminal activity, elapsed time, preparation progress and separate
  file-saving bytes/speed/ETA. Piped logs remain plain and concise.
- Accept `save -v` as well as `save -V` and `save --version`.
- Poll job updates every 500 ms instead of 1 second.

## 2.0.0

### Breaking changes

- `save` is now an HTTP client for the Download Me backend. Inspection, media
  download, audio conversion and verification run remotely; no local media engine
  or automatic local fallback is used.
- The automatic development connection is `http://localhost:8080/api/v1`.
  A running compatible Download Me API and worker on that device are required.
  A publicly hosted default has not been configured yet.
- Supported media currently means single public YouTube videos. The former
  local engine's other platforms, playlists, cookies, proxy, subtitles, aria2,
  connection-count and timing options are not supported.
- Requested video resolution must exist exactly; unavailable qualities fail
  instead of silently downloading a lower quality.

### Added

- Session/CSRF-aware inspection, job polling, cancellation and streamed file saving.
- Atomic no-clobber file saving, temporary transfer cleanup and Unicode filenames.
- Short options for quality, audio, format, picker, listing, batch, overwrite,
  backend override and opt-in usage reporting.
- CLI source attribution for backend-verified completion statistics.
- Optional visit reporting, off by default, with disclosed IP/client metadata.
- Standalone packaging with an embedded runtime and no bundled media tools.
- User-facing installation and usage documentation, npm installation instructions,
  and backend-client regression tests.

### Installation

- npm: `npm install -g @opiobonniky/save@latest` (requires Node.js 18+).
- Standalone ZIP or platform installer: no installed Node.js required.
- Pick one installation method to avoid multiple `save` executables on PATH.
# 2.0.22

- Advertise the expanded public-source adapters: Vimeo, Dailymotion, Reddit,
  X/Twitter, Twitch, Pinterest and SoundCloud, alongside the original four.
- Show readable source names and provider-specific restriction messages.
- Preserve exact format selection, audio-only/MP3 downloads and native streaming.
- Add hosted-client regression coverage for every new platform.
