Fono

First release in which the streaming live-dictation pipeline is
actually reachable from the shipped binary. v0.2.1 and earlier
release artefacts were built without the interactive feature, so
record --live, the live overlay, test-overlay, and
[interactive].enabled were all stubbed out at build time. v0.2.2
flips interactive into the default feature set.

Also bundles the Wave 2 close-out: typed ModelCapabilities API for
fono-bench, per-asset .sha256 sidecar verification in
fono update, --bin-dir CLI flag, real-fixture CI bench gate
against the committed baseline, and the doc-reconciliation pass
(closed three superseded plans, backfilled ADRs 0005–0008 +
0010–0014, added 0017/0018/0019).

Highlights

• Live streaming dictation now ships in the binary. First-time
  users on v0.2.2 will see record --live actually work; v0.2.1
  users upgrading get the feature for the first time.
• Self-update supply-chain hardening. Each release now publishes
  a per-asset <asset>.sha256 sidecar alongside the aggregate
  SHA256SUMS; the next-version client verifies streamed bytes
  against the sidecar and refuses on mismatch.
• fono update --bin-dir <path> to override the install
  directory (matches install-script BIN_DIR semantics).
• Real-fixture CI gate. Every PR runs
  fono-bench equivalence --stt local --model tiny.en and diffs
  against docs/bench/baseline-comfortable-tiny-en.json.

See CHANGELOG.md for the full breakdown.

Install

One-liner (downloads the bare static binary, verifies SHA-256, drops
into ~/.local/bin):

curl -fsSL https://raw.githubusercontent.com/bogdanr/fono/main/packaging/install.sh | sh

Or grab a distro package directly from the assets below.