v2.0.0

Which archive do I want?

You run	Archive
macOS, Apple Silicon	aarch64-apple-darwin
macOS, Intel	x86_64-apple-darwin
Linux x86-64	x86_64-unknown-linux-gnu
Linux x86-64, static (Alpine, containers)	x86_64-unknown-linux-musl
Linux ARM64 (Graviton, Pi 64-bit)	aarch64-unknown-linux-gnu
Windows x64	x86_64-pc-windows-msvc

Each archive ships with a .sha256 next to it. The macOS binaries are
not signed: Gatekeeper blocks them on first launch, which you clear with
xattr -d com.apple.quarantine rustedin or right-click → Open.

Upgrading from 1.x

Every 1.x command moves under rustedin linkedin … — the
README carries the full
mapping table. Your rustedin.json is migrated automatically the first time
2.0 reads it; nothing to do by hand.

---

rustedin becomes multi-platform. It absorbs rustameta — a sibling CLI for
Facebook Pages and Instagram — into a single binary, a single config file and a
single OAuth story, and grows a broadcast command that publishes the same
content to all of them at once.

Added

• Facebook Pages: facebook post (text and/or link), facebook photo
  (single photo, or multi-photo via attached_media), facebook video, all
  with --schedule (validated against Meta's 10-minute / 75-day window) and
  --draft.
• Instagram: instagram post (image, video, or a 2–10 item carousel),
  instagram reel, instagram story, driven by the container → poll → publish
  handshake. Local images are relayed through the linked Facebook Page as
  unpublished photos so the container can ingest their CDN URL; local videos use
  the resumable upload protocol. instagram publish recovers a container whose
  processing outlasted the poll timeout, and instagram limit reports the
  24-hour publishing quota.
• Meta account management: meta setup, meta auth (one Facebook Login
  covers both platforms), meta token to adopt a Business system user token,
  meta pages --refresh, meta use to pin a default Page, meta get as a raw
  Graph API escape hatch. appsecret_proof is sent on every call.
• broadcast — the same --text / --image / --link fanned out to
  several platform:account[/page] targets in parallel, mapped onto what each
  platform actually accepts. Failures are per target, --dry-run prints the
  plan without publishing, and the process exits non-zero if any target failed.
• accounts and status at the root, aggregating every platform;
  --check validates Meta tokens against /debug_token.
• migrate --from <file> — folds another config file into this one. Written
  for the 1.x → 2.0 move: point it at a rustameta.json and its app credentials
  and accounts land under the meta key. Nothing already configured is
  overwritten unless --force says so.
• Cargo features linkedin and meta (both on by default) to build a
  smaller, single-platform binary. CI checks each one on its own.
• Retries with exponential backoff on LinkedIn calls, which previously had none.
• Actionable LinkedIn errors: the serviceErrorCode envelope is reduced to one
  line, with a hint naming the fix for 401, 403, 422, 426 and 429.
• linkedin auth --port to move the OAuth callback off a busy port 8765.

Changed

• Breaking — the CLI is grouped by platform. Every 1.x command moves under
  linkedin (aliases: li, fb, ig, for the other groups):

  1.x	2.0
  rustedin setup --app=…	rustedin linkedin setup --app=…
  rustedin auth --account=…	rustedin linkedin auth --account=…
  rustedin accounts	rustedin linkedin accounts (or root accounts for every platform)
  rustedin status	rustedin linkedin status (or root status)
  rustedin post	rustedin linkedin post
  rustedin reshare	rustedin linkedin reshare
  rustedin share	rustedin linkedin share
  rustedin get-post	rustedin linkedin get-post
  rustedin comments	rustedin linkedin comments
  rustedin profile	rustedin linkedin profile

• Breaking — the config file is namespaced per platform, so the same alias
  can name a LinkedIn company page and a Meta account without colliding:
  linkedInApp and the top-level accounts map move under linkedin.app and
  linkedin.accounts. 1.x files are upgraded automatically on load; nothing
  to do by hand. Unknown top-level keys now survive a rewrite.

• A malformed config file is an error instead of a silent reset — overwriting a
  corrupt file would have destroyed the only copy of the stored tokens.

• LinkedIn's commentary and title limits are counted in characters, not
  bytes. A 3000-character post full of accents or emoji was previously rejected.

• LinkedIn calls now go through the shared pooled HTTP client, with connect and
  request timeouts, instead of building a client per request.

• Percent-encoding of the LinkedIn authorization URL is complete, rather than a
  hand-rolled substitution of six characters.

• status reports LinkedIn expiry dates alongside the remaining days.

Fixed

• Publishing calls are never repeated after a 5xx or a timeout. Only
  failures that provably did not reach the platform — a refused connection, or
  an explicit "throttled, not processed" — are retried on a write. A retry on a
  timed-out publish could have posted twice.
• A concurrent reshare (and now broadcast) refreshes its tokens up front,
  sequentially. Two tasks renewing at the same time each rewrote the whole
  config file from its own stale copy, losing the other's new token.
• Downloading a remote --image no longer decodes the bytes as UTF-8, which
  corrupted every non-ASCII byte of the image.

Dependencies

• Bumped: rand 0.8 → 0.9 (migrated to the new rng() / random() API),
  tokio 1.52.3, clap 4.6.1, serde_json 1.0.150, open 5.3.6.
• Added, behind the meta feature: hmac 0.12, sha2 0.10 (for
  appsecret_proof), and reqwest/multipart.
• Pinned toolchain via rust-toolchain.toml so local and CI use the same Rust /
  clippy version.
• scripts/check.sh runs the exact CI checks locally in one command; the opt-in
  .githooks/pre-push hook blocks a push when they fail
  (git config core.hooksPath .githooks).
• CI: bumped actions/checkout to v7 and installs the pinned toolchain via
  rustup show.