rust-toolchain.toml pins to stable 1.94.1 so contributors and CI runners
share the exact rustc / rustfmt / clippy versions. Without the pin, every
machine surfaces a different lint set depending on its local install — six
pre-existing lints showed up on 1.94.1 that 1.93-era HANDOVER reported clean.
Clippy fixes (all pre-existing, not introduced by feature work):
- crates/storage/src/database.rs: std::iter::repeat().take() -> repeat_n()
- crates/llm/src/lib.rs (docs): "+ frontends" was parsed as a markdown bullet
continuation by rustdoc, breaking doc-lazy-continuation. Reworded to "and".
- crates/llm/src/lib.rs (loop): while-let-on-iterator -> for-loop.
- src-tauri/src/commands/security.rs: .iter().any(|a| *a == x) -> .contains(&x).
- src-tauri/src/lib.rs: io::Error::new(Other, e) -> io::Error::other(e).
- src-tauri/src/tauri_app_data_migration.rs: drop function-tail `return`s
inside cfg blocks; each platform's block now ends with a tail expression.
cargo fmt sweep across the workspace. Mechanical layout-only changes;
no semantics affected.
Workspace gates after this commit:
- cargo fmt --check: clean
- cargo clippy --workspace --all-targets -- -D warnings: clean
- cargo test --workspace: 405/0 (will become 409/0 with Phase A.1+A.2)
The trait's default implementation of transcribe_sync_with_abort fell
through to plain transcribe_sync, silently dropping the abort flag for
any backend that did not explicitly override it. That re-introduced the
5725836 wedge for SpeechModelAdapter (Parakeet today, any future
cloud STT or transformer adapter tomorrow): the live session's
drain_inference timeout would set the flag, the backend would ignore
it, the orphan inference thread would keep the engine Mutex held, and
the next start/stop would deadlock on it.
Removing the default impl makes the method required, so any backend
that does not implement it fails to compile. Compile-time enforcement
of cancellation completeness.
SpeechModelAdapter now implements the method explicitly with a
pre-decode short-circuit: if the abort flag is already set before we
dispatch into transcribe-rs (which owns the Parakeet decoder behind an
opaque call that has no cancellation hook), return an error
immediately. The in-flight decode itself is still uncancellable, but
that is documented with a SAFETY comment and bounded by the live
session's drain timeout dropping the receiver — the orphan exits the
instant it tries to send.
Regression tests:
- transcriber::tests::transcribe_sync_with_abort_is_required_and_flag_is_observed —
fake backend records the abort flag at dispatch time; proves the
trait method is required (file would not compile without it) and the
flag is actually piped through.
- local_engine::tests::speech_model_adapter_short_circuits_when_abort_set_pre_dispatch —
SpeechModelAdapter must NOT call the underlying decoder when the
abort flag was set before dispatch.
- local_engine::tests::speech_model_adapter_dispatches_decoder_when_abort_clear —
companion: clear flag must still dispatch (we did not kill
transcription entirely).
cargo test -p lumotia-transcription --lib: 64 passed, 0 failed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two reversibility kills in the model-download path both followed the
same pattern: SHA mismatch on an existing file triggered
`remove_file(&dest)` BEFORE the network round-trip. A network blip /
power loss between the unlink and the eventual `rename(.part, dest)`
left users with neither the old (corrupt-but-readable) model nor a
fresh one — 1.5-20 GB redownload from scratch with no fallback.
Rev-1 (crates/llm/src/model_manager.rs):
- Extract the existing-file decision into `download_to`, drop the
pre-emptive unlink. `download_impl` already writes via `.part`
and atomically renames; the rename overwrites on success and
leaves dest untouched on failure.
- Regression test `download_failure_preserves_existing_file` plants
a sentinel "OLD" file at dest, points at a 500-returning server,
and asserts dest still exists with original contents after the
failed download.
Rev-5 (crates/transcription/src/model_manager.rs):
- Drop the pre-emptive unlink in the outer `download()` SHA-mismatch
branch. Same atomic rename via `download_file`.
- Make `write_verified_manifest` atomic: write to `.tmp`, fsync,
rename. Previous direct `fs::write` truncates-then-writes, so a
crash mid-write left an empty/torn manifest and triggered a full
GB-sized redownload on next boot.
- `download_file_failure_preserves_existing_dest_file` and
`manifest_write_is_atomic` regression tests added.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three findings chain: thread::spawn discarded the JoinHandle and gave
no cancellation route into the whisper backend; drain_inference busy-
polled with no timeout; stop_live_transcription_session held the
lifecycle AsyncMutex across the await of a non-cancellable
spawn_blocking handle. Together: a single wedged inference (ggml
deadlock, GPU stall) bricked every future start/stop until app restart.
Fix: each inference task carries an Arc<AtomicBool> abort_flag; the
flag is wired into whisper-rs::FullParams::set_abort_callback_safe so
the spawned blocking thread checks it and exits cleanly. drain_inference
is bounded by a deadline (3 x chunk_duration, min 2s); on expiry the
flag is set, the receiver is dropped, and a typed Error::InferenceTimeout
status is surfaced. Drop for InferenceTask asserts the abort flag so
any '?'-propagation or panic unwind closes the cancellation route
without relying on the explicit drain path. stop_live_transcription_session
restructured so the lifecycle guard is dropped BEFORE the JoinHandle
is awaited; start_live_transcription_session releases the guard
explicitly after installing the RunningLiveSession.
Unit test added: dropping_inference_task_sets_abort_flag covers the
Race-A regression directly. A real Race-B drain-timeout test would
require a wedged whisper-rs, which is hard to fixture; that path is
covered by the SAFETY comment and cargo build + manual smoke.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Codex independent review found 11 blockers post-cascade. All addressed.
CRITICAL (data-loss / crash):
10. crates/core/src/paths.rs — migrate_legacy_data_dir_inner used
fs::rename which fails with EXDEV when source + target are on
different filesystems (encrypted-home, bind mounts, separate
$XDG_DATA_HOME partition). Combined with the Phase 5 QC fix that
made migration errors fatal, this would crash on first launch
for any user whose data dir spans filesystems. Added
rename_or_copy_tree() that falls back to copy_dir_recursive +
remove_dir_all on CrossesDevices / errno 18 (EXDEV). Symlinks
preserved verbatim. Same fallback applied to magnotia.db ->
lumotia.db inside the dir.
11. Added 4 unit tests: copy_dir_recursive preserves nested
structure, rename_or_copy_tree same-filesystem happy path,
is_cross_device classifies CrossesDevices kind + raw errno 18.
Doc residuals (blockers 1-9):
1. crates/cloud-providers/Cargo.toml — "Wyrdnote pending rebrand"
description.
2. crates/cloud-providers/src/provider.rs — module docs + test
fixture Wyrdnote refs.
3. crates/transcription/src/orchestrator.rs — module docs + test
fixture Wyrdnote refs.
4. docs/roadmap/2026-05-10-pkm-phase-tooling-shortlist.md — phase
name + outputs/wyrdnote path refs.
5. docs/architecture-map/04-llm-formatting-mcp/llm-tests.md —
MAGNOTIA_LLM_TEST_MODEL env var.
6. .../cloud-providers-stubs.md — MAGNOTIA_API_KEY_*.
7. docs/architecture-map/03-audio-transcription/tests-and-fixtures.md
— MAGNOTIA_WHISPER_*.
8. docs/gpu-tuning/plan.md — MAGNOTIA_BENCH_RUN.
9. docs/superpowers/specs/2026-05-09-battery-gpu-aware-thread-tuning-
design.md + corresponding plan — MAGNOTIA_POWER_STATE_OVERRIDE,
MAGNOTIA_INFERENCE_THREADS.
cargo test --workspace: 343 pass / 0 fail (up from 339; +4 EXDEV
fallback tests).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 3 of the rebrand cascade per locked decision D4. MagnotiaError ->
Error in crates/core/src/error.rs (the crate name already qualifies it).
92 usages across 14 .rs files renamed via word-boundary sed.
One collision required disambiguation: lumotia_storage already had its
own local Error type (introduced by the slop-pass Area A residuals work).
crates/storage/src/error.rs aliases the imported core error as CoreError
on import; the From<Error> for CoreError boundary impl and the
CoreError::Storage construction site use the alias.
cargo build --workspace passes. cargo test --workspace: 330 pass, 0 fail.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Introduces the architectural spine for the Wyrdnote engine layer per
the spec at outputs/wyrdnote/2026-05-10-engine-architecture-spec.md
in the CORBEL-Main workspace, Codex-reviewed 2026/05/10.
Clean-room derivative of VoiceInk's TranscriptionService /
TranscriptionServiceRegistry / TranscriptionPipeline shape. Built from
the architectural pattern only; no GPL-3.0 code, comments, or
identifiers lifted. Wyrdnote names, control flow, and trait signatures
are original.
What this lands:
crates/cloud-providers/src/provider.rs (new, 184 lines)
TranscriptionProvider async trait. Object-safe via async_trait so
Arc<dyn TranscriptionProvider> is the canonical shape. Lives in
cloud-providers (not transcription) per D7 so an OEM licensee
implementing the trait depends only on this crate plus magnotia-core
(no transcription internals leaked through the trait surface).
Surrounding types: ProviderId (lower-kebab-case), ProviderKind,
NetworkRequirement, CostClass, ProviderCapabilities, EngineProfile,
ProviderTranscript. Compile-time object-safety witness in tests.
crates/cloud-providers/Cargo.toml (modified)
Adds async-trait + serde dependencies.
crates/cloud-providers/src/lib.rs (modified)
Re-exports the provider surface.
crates/transcription/src/registry.rs (new, 183 lines)
EngineRegistry: catalogue of providers keyed by ProviderId. Push-style
registration, default-provider id captured at construction. Read-only
after boot. Full unit-test coverage: empty default, register/get
round-trip, default-resolves-after-registration, re-register replaces,
ids enumeration.
crates/transcription/src/orchestrator.rs (new, 286 lines)
LocalProviderAdapter wraps Arc<LocalEngine>, presents async
TranscriptionProvider upward via tokio::task::spawn_blocking. Per D7
the adapter lives in the orchestrator, NOT as impl TranscriptionProvider
for LocalEngine — this keeps the dependency edge one-directional and
avoids leaking async-runtime requirements onto the synchronous
Transcriber trait.
Orchestrator: single transcribe() entry point; resolves a provider
from the registry, derives TranscriptionOptions from the EngineProfile,
dispatches. Returns a clear error when an unregistered provider is
named. Unit tests use a mock CannedProvider to validate dispatch,
error handling, and option routing without booting a real model.
crates/transcription/Cargo.toml (modified)
Depends on magnotia-cloud-providers + async-trait. Dev-dep tokio
gains macros + rt-multi-thread features for #[tokio::test].
crates/transcription/src/lib.rs (modified)
Exports Orchestrator, EngineRegistry, LocalProviderAdapter, and
re-exports the provider trait surface so downstream crates depend
only on magnotia-transcription for both local engines and the trait.
KNOWN-ISSUES.md (modified)
Adds KI-06: existing dictation, live, and meeting commands still call
LocalEngine::transcribe_sync directly via pick_engine. The orchestrator
path is dormant until a Phase A.1 follow-up commit migrates the call
sites. Cloud providers (Phase G) cannot be exercised end-to-end until
the rewire lands. Phasing rationale: this commit lands the abstractions
cleanly with full test coverage; the rewire is a separate diff so the
chunking + post-processing logic in transcribe_file moves without
inflating this commit beyond review fidelity.
Test results: 9 new tests pass (orchestrator: dispatches, errors-on-
unregistered, routes-initial-prompt, object-safe; registry: empty-default,
round-trip, default-resolves, re-register, ids-list). Pre-existing
59 transcription tests + 5 cloud-providers tests still green.
cargo check --workspace clean.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Extends the JFK thread-count sweep to print four labelled panels
(AC+CPU, AC+GPU, battery+CPU, battery+GPU) driven by
MAGNOTIA_POWER_STATE_OVERRIDE. Each panel shows the helper's
predicted thread count for its (power, gpu) combination alongside
the empirical RTF table for n_threads = [1, 2, 4, physical, logical,
maybe 8].
The actual whisper context is initialised once (Vulkan if compiled
in and resolvable, CPU otherwise), so the RTF rows themselves are
produced by the same backend across panels. The CPU vs GPU axis
controls only the helper-pick column, which is the empirical
question we want answered: is the LLM=2 / Whisper=4 GPU floor right?
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wires the whisper backend through the new tuning helper. gpu_offloaded
combines a compile-time feature check (whisper-vulkan, in default
features) with a runtime libvulkan probe via the hardware module. If
libvulkan1 is missing on Linux, whisper-rs's vulkan backend silently
falls back to CPU, so we should not reduce threads in that case.
The old magnotia_core::constants::inference_thread_count import is
replaced by tuning::{inference_thread_count, Workload}. The constants
module's helper is removed in Task 6.1 once the LLM call site has
also migrated.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Bundled work from a low-end laptop (Ryzen 5 4650U / Vega 6 / Linux Mint
22.2 / X11) profiling pass.
perf: WEBKIT_DISABLE_DMABUF_RENDERER=1 default on all Linux
Previously only set on Wayland sessions. Empirically it's a
significant idle-cost win on integrated GPUs in either session type:
env-var matrix (release binary, 75s settle, 10s jiffies CPU sample)
showed magnotia idle CPU 12.30% → 2.80% of one core and idle GPU
17% → 10% on this hardware. Users can opt back in by exporting
WEBKIT_DISABLE_DMABUF_RENDERER=0. The Wayland-only XWayland
fallback (GDK_BACKEND=x11, WINIT_UNIX_BACKEND=x11) is unchanged.
fix: secondary-windows ACL — allow set-always-on-top
The float window's pin toggle calls setAlwaysOnTop() but the
secondary-windows capability didn't permit it, so the popout was
stuck always-on-top regardless of the pin state. Adds the
core:window:allow-set-always-on-top permission. Narrow scope.
fix: guard registerGlobalHotkey against non-main webviews
Cross-window settings sync via localStorage can re-fire the
$effect(() => settings.globalHotkey) callback inside popout webviews
where the main layout's registerGlobalHotkey is reachable. Adds an
early-return when the current window label is not "main", so the
popout doesn't trigger an ACL-denied register/unregister and the
user no longer sees a spurious "Hotkey not registered" toast when
popouts are open. Keeps the global-shortcut perm scoped to main.
build: pin @tauri-apps/api 2.10.1 + @tauri-apps/plugin-dialog 2.7.1
Match the Rust crate versions tauri-cli's version-mismatch check
enforces during release builds. Without this, `npm run tauri build`
exits 0 silently while emitting an Error and never producing
binaries.
test: add crates/transcription/tests/jfk_bench.rs
Reproducible RTF regression fixture. Env-gated on
MAGNOTIA_WHISPER_TEST_MODEL + MAGNOTIA_WHISPER_TEST_AUDIO so it
never runs in CI without setup. Loads the JFK WAV inline (no hound
dep), times model load + cold + warm transcribe, prints SUMMARY.
Baselines on this hardware:
--release --features whisper: cold RTF 0.054, warm 0.050, RSS 125 MB
--release --features whisper,whisper-vulkan: cold RTF 0.029, warm 0.028, RSS 125 MB
Vulkan on RADV/Vega 6 nearly halves transcription latency for
Whisper Tiny — useful baseline for Phase 10 hardware-recommendation
scoring.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replace all instances of the legacy product names "Kon" and "Corbie" with
"Magnotia" across user-facing copy, code identifiers, package names, bundle
ids, file paths, and documentation. Preserves the unrelated "konsole" (KDE
terminal) reference and the parent CORBEL company name.
- Renames 10 Rust crates (kon-* → magnotia-*) and the tauri binary
- Updates package.json, tauri.conf.json (productName + identifier)
- Renames CSS classes (kon-rh-* → magnotia-rh-*) and animations
- Renames brand and roadmap docs
- Regenerates Cargo.lock and package-lock.json
Verified: svelte-check passes; pure-rust crates compile under new names.
Two issues in flush_is_idempotent_and_leaves_clean_state from
581a098:
1. silence_close_samples and max_chunk_samples were cast `as u64`
but with_thresholds takes usize — wouldn't compile.
2. enter_threshold was 0.005 and exit_threshold 0.01, which
violates the hysteresis invariant (enter must be >= exit) and
panics in debug_assert at runtime. Swap to 0.01 / 0.005 so the
test actually runs.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Both `kon-transcription` and `kon-llm` previously hardcoded their native
acceleration features in Cargo.toml — `whisper-rs` with `vulkan`,
`llama-cpp-2` with `openmp` + `vulkan`. That worked everywhere desktop
ships (Linux/macOS/Windows all have Vulkan via MoltenVK on Mac), but it
made an Android build structurally impossible: NDK builds against drivers
that vary wildly across SoCs (Adreno OK, Mali patchy, PowerVR worse), and
some older devices have no Vulkan at all.
Roadmap step 0 from the Android plan: make the GPU acceleration
opt-in so a CPU-only target compiles. Reuses the existing pattern that
README's "future Windows non-AVX2 build" comment hinted at.
- kon-transcription: new `whisper-vulkan` feature gates `whisper-rs/vulkan`
via the optional-syntax `whisper-rs?/vulkan`. Default features stay as
`["whisper", "whisper-vulkan"]` so desktop is unchanged.
- kon-llm: new `gpu-vulkan` and `openmp` features each gate the matching
`llama-cpp-2` feature. Default stays `["gpu-vulkan", "openmp"]`. They are
independent so an Android Vulkan build can opt into vulkan without
openmp (NDK OpenMP linking has known cross-version fragility).
CPU-only build invocations:
cargo build -p kon-transcription --no-default-features --features whisper
cargo build -p kon-llm --no-default-features
Verified: all 91 tests in the buildable-in-sandbox crates still pass.
The two crates whose Cargo.toml changed (kon-transcription, kon-llm)
can't be compiled in this sandbox (ort-sys CDN + cmake-built llama.cpp);
CI's Linux/macOS/Windows builders will exercise the default-feature path
exactly as before.
https://claude.ai/code/session_0189xUb6ie6t9qHkzatGZ9Rb
The earlier audit noted that `flush()` had three exit paths but only two
of them explicitly cleared all state-machine fields:
1. InSpeech with non-empty active_chunk: emit_active_chunk_and_close()
handled it.
2. InSpeech with empty active_chunk (hit_max-mid-flush): handled inline.
3. Idle (no padded frame, or padded frame closed cleanly): no explicit
reset — silent_tail_samples / pending_onset_frames / onset_buffer
could carry stale values from `consume_frame` calls inside the same
flush.
In the worst case, the first feed of a fresh recording could see leftover
onset bookkeeping and produce a chunk start that doesn't match the new
session's audio. Reusing the same `RmsVadChunker` across stop/start is
the main path that would hit this.
Add a single defence-in-depth reset block at the end of flush — every
exit path lands the chunker in the same fields a fresh chunker has,
except `next_sample_index` (the running total-samples counter, intent-
ionally preserved). Test asserts: a second flush after a full speech →
silence → partial-pending sequence emits zero chunks, and a subsequent
silent feed also emits zero, proving no stale state leaked.
(kon-transcription doesn't build in the audit sandbox because ort-sys's
build script can't reach pyke's CDN; CI cross-platform compiles it.)
https://claude.ai/code/session_0189xUb6ie6t9qHkzatGZ9Rb
Auto-applied cargo clippy --fix across 11 files — needless return,
unnecessary cast, map_or simplification, repeat().take() → repeat_n(),
iter().any() → contains(), manual char comparison, lifetime elision,
push_str single-char, reference immediately dereferenced.
Also fixed three lints on file_storage.rs manually: two doc-list-item
overindentations, plus the same needless-return. Baseline main was
not clippy-clean with -D warnings before; after this pass one
needless_range_loop warning remains (live.rs:1089) that clippy's
suggested rewrite would make less readable — left for a dedicated
refactor session.
Build + workspace tests remain green (245 passing, 0 failing, 1
ignored).
MAJOR from the 2026-04-22 review (model_manager.rs:161-262): reqwest
does not return Err on 4xx/5xx by default. The resume branch
validated 206/200 and errored on anything else, but the non-resume
branch skipped the status check entirely — a 404 or 500 body was
streamed into .part and atomically renamed over the destination as
if it were the model file. For models without a sha256 declared,
this failure mode was silent and catastrophic (the engine would
crash loading an HTML error page as GGML on next launch).
Adds an is_success() check in the non-resume branch: any non-2xx
returns KonError::DownloadFailed with the HTTP status in the
message, and (importantly) we return before File::create so no
.part file is left behind.
Test: spawn_500_server that responds 500 to any request; a fresh
(no .part, no sha256) download must Err with "HTTP 500" and leave
neither .part nor dest on disk.
CRITICAL from the 2026-04-22 code review: RmsVadChunker::flush() was
calling consume_frame() on a zero-padded final frame via `let _ = ...`,
discarding any VadChunk the call emitted. If the padded frame triggered
end-of-utterance (silent tail + padding zeros push past
silence_close_samples) or max_chunk_samples (buffered speech + padding
push past the cap), the emitted chunk was lost; the outer state check
then either returned None or an empty closing chunk.
Changes the VadChunker trait flush signature from Option<VadChunk> to
Vec<VadChunk> so both the mid-flush emission (from consume_frame) and
the closing emission (from emit_active_chunk_and_close) can be
surfaced. Updates RmsVadChunker::flush to collect from both sites
and skip a zero-length closing emit when the hit_max continue variant
already cleared active_chunk.
Two regression tests land alongside:
- flush_preserves_hit_max_chunk_from_padded_final_frame: tight
max_chunk, sub-frame speech tail; pre-fix dropped the chunk, post-fix
emitted samples cover the full active-speech region.
- flush_preserves_end_of_utterance_chunk_from_padded_final_frame:
silent tail near silence_close; padded zero frame closes the
utterance inside consume_frame; pre-fix returned None.
No production callers yet — the VadChunker wiring in live.rs is a
deferred item from A.3. API change is clean within the repo.
Review feedback (MINOR): the original <= 0.0 clamp caught negatives
and zero but not non-finite inputs. Rust's saturating float-to-int
cast turns f64::INFINITY into u64::MAX, which would park the capture
buffer origin beyond any reachable sample index and trim the whole
buffer forever if a future end_secs source ever produces infinity
(clock glitch, overflow upstream, corrupted timestamp in a pass).
Adds is_finite() check. NaN, +infinity, -infinity, and zero all
return 0, which downstream trim_buffer_to_commit_point treats as
no-op. Test covers all three non-finite cases.
Review feedback (CRITICAL): when a chunk hit max_chunk_samples during
continuous speech, emit_active_chunk reset state to Idle. The next
1-2 loud frames of post-split continued speech went into onset_buffer
and were silently cleared if silence arrived before the 3-frame onset
threshold — 50-100ms of user audio lost at every max-chunk boundary
in long-continuous-speech scenarios.
Splits emit_active_chunk into two variants:
- emit_active_chunk_and_close: the existing behaviour. Used for
end-of-utterance closes and end-of-session flush. Truncates trailing
silence, resets state to Idle.
- emit_active_chunk_continue: mid-utterance split on max_chunk. Stays
in State::InSpeech, clears active_chunk for continued accumulation,
advances active_chunk_start by the emitted length so the next
chunk's start_sample is contiguous with this one's end. No
silence-trim (by definition still in speech — end-of-utterance
takes priority).
Adds max_chunk_split_preserves_audio_contiguity test: feeds 17 frames
of continuous speech into a chunker with a 4-frame cap, asserts
(a) chunk[i+1].start_sample == chunk[i].start_sample + chunk[i].samples.len()
across every pair, and (b) the final emitted region reaches the end
of the fed speech with no sample loss.
Review feedback (CRITICAL): LocalAgreement::push could panic with an
index OOB when a later pass arrived shorter than committed_count.
Concrete case: commit [a, b], next pass arrives [a] — lcp_len=1,
new_committed=max(1, 2)=2, then latest[2..] panicked because
latest.len()==1.
A Whisper re-transcription of an overlapping window can legitimately
collapse repeated segments, or the user can stop mid-utterance after
some tokens were already committed, both of which produce this
shape. The committed_count invariant still holds (non-shrinkage) —
it is the slicing that was unsafe.
Clamps every latest[..] slice against latest.len() before indexing.
committed_count stays at new_committed even when the pass is shorter:
non-shrinkage is relative to what we have already emitted, not to
the current pass length. newly_committed and tentative both return
empty when the shorter pass has nothing past the committed prefix.
Adds two regression tests:
- shorter_pass_after_commit_does_not_panic (commit 2, push 1)
- empty_pass_after_commit_does_not_panic (commit 1, push empty)
New streaming::buffer_trim module with two pure helpers:
- sample_index_for_seconds(end_secs, sample_rate) -> u64: converts
LocalAgreement::last_committed_end_secs() into an absolute sample
index. Defensive against negative end_secs (treats as 0) so a future
clock-skewed timestamp cannot wrap to a huge u64.
- trim_buffer_to_commit_point(buffer, buffer_start_sample,
commit_sample_index) -> new_buffer_start_sample: drains the prefix
of the capture buffer that falls below the commit point and returns
the new absolute-index origin.
Edge cases covered by tests:
- commit before buffer start → no drain
- commit equal to buffer start → no drain
- commit inside buffer → drain prefix, advance origin
- commit at buffer end → drain all, origin moves forward
- commit past buffer end → drain all, origin parks at commit (rare
edge after a committer reset)
- sample_index_for_seconds rounds nearest, negatives clamp to 0
- integration with LocalAgreement::last_committed_end_secs
trim_bounds_buffer_over_long_session is the acceptance fixture for
the ufal #120/#102 failure mode: 100 cycles of 16_000 captured samples
with a 200-sample tentative tail per cycle, and the buffer stays below
2× the tentative envelope instead of growing to 1.6M samples.
Integration into src-tauri/src/commands/live.rs deferred to the
dogfood session that wires VadChunker and LocalAgreement end-to-end —
the trim is a one-line replacement at the maybe_dispatch_chunk drain
site once the committer is feeding it commit points.
New streaming::commit_policy module implementing ufal's
LocalAgreement-n pattern: tokens emitted by the streaming ASR
pipeline stay tentative until N consecutive passes produce the same
prefix, at which point the agreed prefix commits.
Types:
- Token: text + absolute start/end seconds. PartialEq is text-only so
identical words from overlapping Whisper windows compare equal even
when their timestamps drift.
- CommitDecision { newly_committed, tentative }: the partition fed
back to the live-session worker after each pass.
- CommitPolicy enum with LocalAgreement { n }. Default is n=2 (ufal).
- LocalAgreement: stateful committer with push/flush/reset and a
last_committed_end_secs accessor. Brief item #25 uses the latter
to compute the sample-index drain target for aggressive buffer trim.
Invariants exercised by tests:
- first pass is all tentative (need 2 passes to commit under n=2)
- two matching passes commit their common prefix
- divergent second pass commits nothing
- extending agreement commits only the newly-agreed tokens
- tentative tail tracks latest pass only (no stale guesses)
- committed prefix never shrinks, even if later passes contradict
- flush emits any tentative-but-not-committed tail at session end
- flush on empty history is a no-op
- reset clears all commit state
- n=3 requires three matching passes before anything commits
- CommitPolicy::default() is LocalAgreement { n: 2 }
Integration into src-tauri/src/commands/live.rs deferred — the
tentative/committed split needs the B-side 'tentative: bool' field on
LiveResultMessage.segments (workstream-B #24 UI contract) and
validation against real streaming captures before going live.
New crates/transcription/src/streaming/ module with:
- VadChunker trait: Send-bound, object-safe, push/flush/reset/
next_sample_index. Same surface a future Silero backend will
present, so live.rs wiring does not change when Silero drops in.
- VadChunk type: (start_sample: u64, samples: Vec<f32>) for
commit-policy sample-offset bookkeeping in #24.
- RmsVadChunker: fallback backend the plan permits while the
ort 2.0.0-rc.10 vs rc.12 ecosystem conflict blocks silero-vad-rust
/ voice_activity_detector. Tuned to match the existing
evaluate_speech_gate behaviour (enter 0.003, exit 0.0014, 3-frame
onset, 500 ms silence close, 2 s max chunk).
Key behavioural properties, each backed by a test:
- pure silence emits nothing
- samples between exit and enter thresholds never trigger onset
- a single loud frame does not start a chunk (sustained speech only)
- sustained speech followed by silence emits exactly one chunk
- hysteresis: a dip between enter and exit does not split a chunk
- max_chunk_samples caps continuous speech (Whisper never fed > 2 s)
- flush surfaces in-flight speech
- flush on an idle chunker emits nothing
- reset restores a clean state
- emitted chunk start_sample includes the onset buffer (Whisper sees
the speech attack, not post-onset audio)
Open items tracked as follow-ups:
1. Silero backend via direct ort rc.12 bridge (Handy-style). Blocked
on either ecosystem ort alignment or dedicated bridge session.
2. Integration into src-tauri/src/commands/live.rs. Deferred so
threshold tuning can be validated against real microphone
captures rather than synthetic constant-signal fixtures.
New crates/transcription/src/transcriber.rs defines a Transcriber
trait (Send supertrait for spawn_blocking travel) with
TranscriberCapabilities (sample_rate, channels, supports_initial_prompt).
TranscriberCapabilities.sample_rate is load-bearing for the upcoming
progressive WAV writer (#19).
Concrete impls:
- SpeechModelAdapter wraps Box<dyn transcribe_rs::SpeechModel + Send>
for Parakeet (and any future transcribe-rs-backed engine).
- WhisperRsBackend moves its transcribe_sync body into the impl,
widening the signature from &self to &mut self so per-call
WhisperState can be created cleanly through the trait object.
LocalEngine now holds Box<dyn Transcriber + Send>; dispatch in
transcribe_sync collapses from a match to a direct call. Adds
LocalEngine::capabilities() for the WAV-writer.
Cargo feature flag "whisper" (default on) makes whisper-rs, num_cpus,
and the whole whisper_rs_backend module optional. cargo check
--no-default-features -p kon-transcription now builds without pulling
whisper-rs-sys — the escape hatch brief item #6 / #13 called for on
Windows / non-AVX2 / cloud-only builds. load_whisper is cfg-gated
behind the same feature.
src-tauri/src/commands/models.rs load_model_from_disk returns
Box<dyn Transcriber + Send> instead of SpeechBackend; caller chain
(ensure_model_loaded, prewarm_default_model) is unchanged.
transcriber_trait_is_object_safe test lands alongside the trait as a
compile-time witness against future Self-returning / generic-method
additions.
Review feedback: the previous test would pass even if the Range-header
logic in download_file were deleted entirely, because File::create
truncates the stale .part regardless of which branch set
actually_resuming to false.
Tightens spawn_no_range_server to return HTTP 400 when the request
carries no Range header, and only return 200 + full body when Range is
present. A regression that stops sending Range now surfaces as a
download failure (empty body from 400, bytes != body assertion)
instead of silently passing through the truncation path.
The downloader already handles servers that return 200 to a Range
request by falling through to a truncating File::create on the .part
path, discarding stale partial bytes. That branch had no dedicated
fixture test — the SHA mismatch and Range-honouring resume cases were
covered but the mirror / proxy that strips Range support was not.
Adds spawn_no_range_server (always 200, full body regardless of Range
header) and download_file_restarts_when_server_ignores_range. Writes 12
bytes of stale content to .part, kicks off a download, asserts the
final file matches the fresh body exactly (not stale-bytes-prefixed)
and the .part file is cleaned up.
Brings the transcription downloader to test-coverage parity with
crates/llm/src/model_manager.rs per brief item #8 ("test coverage
parity" acceptance).
Two new Settings → AI knobs that compose cleanly with what already
shipped (aiTier, LLM model, translator prompt framing).
**B.1 #15 — Named cleanup presets.** LlmPromptPreset enum
(Default / Email / Notes / Code) appends a short context hint onto
the CLEANUP_PROMPT just before generation. Presets shape tone and
structure ("email paragraph", "bulleted meeting notes", "preserve
technical terms") without licensing the content-editing the
translator-not-editor framing forbids. cleanup_transcript_text_cmd
now takes `preset: Option<String>` which runs through the new
LlmPromptPreset::parse (normalises aliases like "meeting-notes",
collapses unknown values to Default).
**A.1 #28 — Sequential-GPU guard.** New LocalEngine::unload drops
the backend + model_id so a subsequent load actually reclaims VRAM.
load_llm_model, load_model, and load_parakeet_model Tauri commands
grow an optional `concurrent: bool` argument. When concurrent is
Some(false), loading LLM first unloads whisper+parakeet, and vice
versa — prevents VRAM OOM on tight-VRAM setups. Default is the
previous parallel behaviour so nothing changes for multi-GB cards.
Transcribe-in-progress paths (transcribe_pcm, transcribe_file, live)
pass None, so mid-dictation model loads don't accidentally tear
down the LLM.
Settings UI (AI section):
- Cleanup preset segmented button + descriptive copy for each option.
- GPU concurrency segmented button with explicit trade-off text
("faster transitions vs fits in tight VRAM").
Frontend wiring:
- settings.llmPromptPreset flows from DictationPage's
cleanupTranscriptIfEnabled into the Tauri command.
- settings.aiGpuConcurrency flows from both DictationPage (auto-load
on record) and SettingsPage (manual load/unload buttons) as
`concurrent: "parallel" === true` to the load commands.
Tests: three new preset cases in crates/ai-formatting/src/llm_client.rs
(parse aliases, suffix non-empty for non-default, default suffix
empty). All 139 existing lib tests still pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ports the kon-llm model_manager resume pattern the rest of the way
into kon-transcription::model_manager:
- download() now validates an existing complete file against its
sha256 before skipping; a hash mismatch removes the file and
re-fetches, instead of serving a corrupt file to whisper.cpp.
- download_file() now distinguishes 206 Partial Content, 200 OK
(resume silently ignored by server), and other statuses, rather
than treating any non-206 as 'just use it as a fresh start'.
200-on-Range is handled by discarding the partial and starting
over cleanly.
- New tests: download_file_resumes_from_partial_and_verifies_sha
(TcpListener fixture, same shape as kon-llm's), and
download_file_fails_on_sha_mismatch_and_cleans_part_file.
- sha256_of_file helper + unit test for the existing-file guard.
Dev-deps: tempfile + tokio(net/io-util/macros). Total workspace
lib-test count: 116 → 123.
Co-authored-by: jars <jakejars@users.noreply.github.com>
Adds a build.rs guard that parses Cargo.lock and panics on Windows
if the tokenizers crate ever appears in the workspace dependency
tree, mirroring the MSVC C-runtime conflict that broke Whispering
v7.11.0 when they linked whisper-rs-sys + tokenizers in the same
binary.
On non-Windows hosts the guard downgrades to a cargo:warning so
cross-compilation or CI from Linux surfaces the issue before a
Windows build attempt actually panics.
No tokenizers crate is in the tree today; the guard is preemptive.
If we ever legitimately need HF tokenizers on Windows, the escape
hatch is an out-of-process sidecar (separate CRT).
Co-authored-by: jars <jakejars@users.noreply.github.com>
Major quality pass on top of Phase 2. Five substantive changes plus
cross-cutting touches across audio, hotkey, transcription, and Tauri
command layers.
Transcription quality
- Long-audio chunking in commands/transcription.rs: Parakeet and large
file transcription now chunk-and-recompose with overlap trimming, so
the live-path chunking advantage extends to file-based workflows.
- Stateful live speech gate in commands/live.rs on top of the earlier
duplicate-boundary filtering — distinguishes start-of-speech from
mid-speech and holds state across chunks.
Auto-learning corrections
- New crates/ai-formatting/src/correction_learning.rs: extracts user
text corrections from viewer edits and proposes additions to the
active profile's vocabulary.
- src-tauri/src/commands/profiles.rs bridge for frontend-driven
confirmation of learned terms.
- src/routes/viewer/+page.svelte hooks the learning path into the
segment-edit flow so corrections feed profile_terms without a
separate 'train this profile' UX.
Transcript profile provenance
- Migration v8 (crates/storage/src/migrations.rs) adds profile_id to
transcripts, defaulting to DEFAULT_PROFILE_ID so existing rows stay
valid.
- crates/storage/src/database.rs: TranscriptRow + CRUD carry profile_id.
- src-tauri/src/commands/transcripts.rs: add_transcript accepts and
persists profile_id.
- DictationPage.svelte + FilesPage.svelte send activeProfileId on
capture so learned corrections are attributed to the right profile.
Cleanup prompt contract
- crates/ai-formatting/src/llm_client.rs hardened: the CLEANUP_PROMPT
now specifies concrete do/do-not rules, ready for a real model-backed
cleanup pass. The llm_client is still a stub — kon-llm remains unwired
— but the prompt shape is final.
Cross-cutting polish
- Minor touches in audio (capture/decode/resample), hotkey (lib/linux/stub),
core, transcription (concurrency/model_manager/local_engine/whisper_rs),
and the rest of src-tauri/src/commands/*: error-path tightening, log
clarity, TS-migration follow-ups (@ts-nocheck additions for incremental
typing).
Verified locally: npm run check, cargo test -p kon-ai-formatting,
cargo test -p kon-storage, cargo test -p kon --lib commands::live::tests,
cargo check — all green.
Scope boundary: kon-llm crate is still a stub; task extraction remains
rule-based. Bundled local-LLM runtime is the next clean step and is not
in this commit.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
transcribe-rs 0.3.10's ParakeetModel::transcribe_raw ignores its
options argument and calls self.infer(samples, &TimestampGranularity::default())
where default is TimestampGranularity::Token — per-subword segments.
That surfaces in Kon as output like 'T Est Ing One , Two , Three . W Ow ,
This Is T Ri Ble .' because DictationPage joins segment texts with ' '.
Introduce a thin ParakeetWordGranularity wrapper that implements
SpeechModel and overrides transcribe_raw to call the concrete
ParakeetModel::transcribe_with() with ParakeetParams { timestamp_granularity:
Some(Word) }. Pre-existing bug unrelated to Phase 2 work — surfaced during
Group 1 dogfooding because Parakeet was being tested for the first time.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
crates/storage/src/migrations.rs:
Replace naive sql.split(';') with split_statements() that tracks
BEGIN...END depth, so migrations containing trigger definitions
execute correctly instead of being split mid-block.
crates/transcription/src/lib.rs:
Re-export transcribe_rs::SpeechModel so callers in src-tauri can
reference it without adding transcribe-rs as a direct dependency.
src-tauri/src/commands/models.rs:
Use Box<dyn SpeechModel + Send> as the load_model_from_disk return
type, matching the trait object that transcription crates produce.
src-tauri/src/commands/transcripts.rs:
Remove the stale .map(|v| v as i32) casts on sample_rate and
audio_channels. InsertTranscriptParams now stores these as i64
(ebf449b), matching the i64 fields on CreateTranscriptRequest;
casting to i32 first would silently truncate large sample rates.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Upstream transcribe-rs 0.3.10 added required fields to TranscribeOptions.
Set both to None (use engine defaults).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
