Edit: Label ▸ 16 colours + Select Label Group, Paste Attributes… (⌥⌘V, with
Scale Attribute Times), Remove Attributes…, Select All Matching, Remove
Unused, Consolidate Duplicates. File: New ▸ Sequence From Clip, Bin From
Selection (⇧B), Offline File…; Close Project (⇧⌘W, asks to save), Save All.
Clip: Make Subclip… (⌘U, Restrict Trims), Edit Subclip…, Modify ▸ Audio
Channels… (⇧G) / Timecode…, Video Options ▸ Frame Hold Options…, Add Frame
Hold (now splits at the playhead), Insert Frame Hold Segment, Field
Options…, Time Interpolation (Frame Sampling / Frame Blending / Optical
Flow, blending rendered on CPU; optical flow falls back to blending),
Fit to frame / Fill frame; Audio Options ▸ Breakout to Mono, Extract Audio
(WAV next to the project, imported); Replace With Clip ▸ From Source
Monitor / Match Frame / From Bin. Hold Filters: effects animate through a
hold unless set. Menus support nested submenus (egui and native) and follow
Premiere's order. Dialogs in panels::clip_dialogs with automation ids and
ui.set {clipDialog}.
Project schema v9 (no-op migration v8→v9, v8-subclip fixture): clip
time_interpolation / hold_filters / field_options / source_channels,
Interpretation.audio_channels, Subclip.restrict_trims.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- crates/speech (L2): Transcriber trait, FixedTranscriber, Whisper model catalogue (pinned,
SHA-256 verified) with optional downloader (feature `download`: ureq + rustls + RustCrypto),
pure-Rust Whisper inference on candle with word timestamps (feature `whisper`), log-mel, VAD
word tightening, MFCC speaker clustering. Both features off by default and for wasm.
- project: per-media-item transcripts (Project::transcripts); schema v9 with a no-op v8->v9
migration and a v8 fixture.
- edit::transcript: sequence transcript derived from clip transcripts, search, paragraphs,
word ranges, pause/filler detection, ripple delete of many ranges, caption layout.
- engine: transcript.* commands (generate, set, delete, inspect, search, select, extract, lift,
renameSpeaker, removeFillers, removePauses, createCaptions, models, downloadModel);
Session::transcriber for hosts/tests; features `whisper` / `speech-download` (also on the
desktop and CLI apps). Without whisper, transcript.generate fails with a clear error.
- ui: Text panel Transcript tab (Transcribe button, speaker paragraphs of clickable words,
Shift+click selection -> In/Out, extract/lift/remove fillers/create captions, search) with
text.transcript.* automation ids; headless UI test.
- docs/transcripts.md, crates/speech/README.md, architecture and project-files updates.
whisper-tiny measures 12.2% WER over 797 words on local samples.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MonitorView state is serde and settable with ui.set {program|source}; view.* UI commands with
menu checkmarks; wrench and zoom menus with automation ids; headless kittest tests (view_ui.rs).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Vertical text layout in the text engine (one line per character, columns right to left).
All edits are undoable engine commands; engine tests cover each menu group.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- eframe web runner on WebGPU (WebGL2 fallback; GPU compositor on WebGPU, CPU on WebGL2).
- fs: virtual file table of File/Blob handles, 1 MiB chunked async range reads with an
LRU cache and read-ahead (pending + retry instead of blocking), downloads for written
files, /opfs/ paths written to OPFS.
- import: File System Access picker / <input type=file>, drag-and-drop anywhere,
container index prefetch (moov wherever it is), retried imports; project open/save.
- recovery: OPFS auto-save snapshot while dirty, media copies, reopened on next visit.
- audio: AudioWorklet output fed ahead from the UI thread; played-frame audio clock.
- webcodecs: H.264/HEVC/VP9/AV1 MP4/MOV decoded by the browser VideoDecoder via a
reader opener (fallback to our decoders on any error), probed with isConfigSupported.
- api: window.filmcraft (execute/request/commands/screenshot/importFiles/importUrl/
openProject/files/readFile/info).
- cargo xtask web [--dev] [--serve PORT] (wasm-bindgen CLI); xtask wasm also checks
filmcraft-ui-egui and filmcraft-web; layer L6 for the web app.
- Original web icon (favicon.svg) with attribution sidecar; docs/web.md; headless-Chrome
smoke test (tests/smoke.mjs).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Agents' ui.* requests used ViewportCommand::Focus, so the user's typing in another app landed in FilmCraft as shortcuts (stray Extract, Mark In/Out, markers…). Verified live: the frontmost app stays frontmost through launch, ui.click, ui.key and ui.screenshot, and the requests still take effect.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Av1Decoder in filmcraft-codecs: feeds the av1C configuration OBUs before the
first sample (and after every reset), converts pictures to 8/16-bit planar
YUV with the stream's colour description, and reports random-access points
from the first frame header of each temporal unit (shown key frame) so GOP
seeking ignores bogus container sync flags. Matroska V_AV1 tracks map to an
av01 sample entry; MP4 labels show the AV1 profile.
All-intra 8-bit MP4 and 10-bit WebM files seek sample-exactly against
libdav1d. Inter frames still return "unsupported" until inter prediction
lands.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
captions.setStyle {margin: 60} (pixels, as one would guess) was silently clamped to 0.45 and
put subtitles mid-frame. The parameter doc and docs/captions.md now give the unit and range.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Replaces the Night of the Living Dead trailer screenshot with a real job: a 3:10
documentary cut of NASA's Apollo 11 film (images.nasa.gov, US Government works) built
through the control channel — 49 picture cuts with dissolves, B-roll, lower-third titles
set by the text engine, subtitles from the NASA air-to-ground transcript, mission audio,
a CC0 Chopin bed ducked by Essential Sound, a Lumetri grade with keyframes, markers and a
rendered section. Sidecar lists every source and the reproduction steps; NASA does not
endorse this project.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
masks.track {direction, frames?, method?, wait?}: progress + cancel via jobs.*, keyframes written live as one undo step per run. Synthetic known-motion tests: <0.75 px drift over 10 frames (render), 0.29 px max vertex error over -8..+12 frames end to end (engine).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New L0 crate filmcraft-dnx, written from SMPTE ST 2019-1:2016 + Amd 1:2023
(tables extracted from the standard's text by tools/extract_tables.py).
Decoder: every compression ID (DNxHD 1235-1260 incl. field-coded and
macroblock-adaptive interlace, thin rasters, RGB 4:4:4; DNxHR 1270-1274
LB/SQ/HQ/HQX/444), 8/10/12-bit, 4:2:0/4:2:2/4:4:4, DCT and lossless alpha,
macroblock rows in parallel. Matches ffmpeg within IDCT precision (max 2 LSB,
<=50 per million beyond 1, 87-99% identical) on 22 ffmpeg fixtures; follows
ffmpeg where it departs from the standard's text (reconstruction offset,
CID 1271 weights, 4:4:4 signalling) - documented in the README.
Encoder: DNxHR LB/SQ/HQ/HQX(10/12-bit)/444, CBR at the eq. 7.1 frame sizes,
uniform-qsf binary search + per-macroblock refinement; ffmpeg decodes every
profile with -xerror (HQ luma PSNR 58.8-66.3 dB).
Wiring: MOV AVdn/AVdh import (codec label from the CID), export
Format::DnxHr (MOV AVdh, dnxProfile lb|sq|hq|hqx) in the export dialog and
file.exportMedia.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- file.projectManager {destination, mode: collect|consolidate, sequences?,
excludeUnused, handles, preset, includeProxies, includePreviews, dryRun, wait}.
- Collect copies every file the chosen sequences use (nested sequences and subclips
followed), optionally proxies and render previews; consolidate writes only the used
media range plus handles with a transcode preset and re-bases media time so edits,
keyframes and markers stay put. Unused clips can be excluded.
- Dry run reports original / resulting disk space and the file plan; the copy runs
as a job and writes the project last with the new paths and fingerprints.
- UI: Project Manager dialog (sequences, mode, preset, options, destination,
Calculate → sizes) with automation ids pm.*.
- Tests: collected project opens with its media moved away and renders the same
frames; consolidate trims to used range + handles and matches (>38 dB) the original.
- ROADMAP: M11 done.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- media.createProxies: background job (jobs.list progress) transcoding with our own
encoders — ProRes 422 Proxy ¼/½, ProRes LT ½, H.264 ¼/½ (audio included) — into
<media dir>/Proxies or a chosen folder; attached when the job finishes
(Session::poll_persistence applies finished media jobs).
- media.attachProxies (duration ±1 frame and frame rate must match; size/aspect may
differ), media.detachProxies, media.reconnectFullRes, media.toggleProxies
(Preferences ▸ Media ▸ enableProxies; bumps the view revision without dirtying the
project), media.proxyPresets.
- Proxy frames are smaller; ProxySource reports the original size and the compositor
derives its pixel scale from the frame, so Motion and effects render the same
picture (44 dB PSNR vs full res with Motion + Gaussian Blur, 48 dB at ½).
Export always renders full resolution.
- ExportSettings::prores_profile (proxy / lt / standard / hq).
- project.ingestSettings + file.import: copy (fingerprint-verified), transcode,
create proxies, copy and create proxies.
- UI: Toggle Proxies button on both monitors (lit when on), Create Proxies dialog,
Project panel Proxy submenu and P badges.
- Measured (one core, 4K H.264 testsrc2): CPU composite 201 → 35 ms/frame at ¼ with
ProRes Proxy ¼, 277 → 131 ms at ½ with H.264 ¼; decode 9.2 → 0.7 ms with H.264 ¼.
- Tests: create+attach+toggle same composite modulo scale, attach checks, export
ignores proxies, ingest copy + proxies, ignored 4K benchmark, headless UI test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- render::color_match: non-ML statistics matching — Oklab means per tonal range (smoothstep
shadow/midtone/highlight weights) plus mean chroma; "Face Detection" down-weights skin-tone
pixels (Oklab hue 20–75°) in the colour statistics and then limits the skin hue shift to 8°;
the grade is solved in Lumetri's own parameters (3 wheels, their lightness, saturation) by
damped Gauss–Newton through the real Lumetri implementation, so the result stays editable
- render::render_clip: one clip's layer at a time (the shot as Lumetri sees it)
- lumetri.applyMatch {clip?, referenceTime|Frame|Timecode, faceDetection} sets the grade in one
undo step and reports the statistics distance before/after; refuses a reference inside the
clip itself
- Lumetri panel: Color Match row (reference timecode, Face Detection, Apply Match; automation
ids lumetri.match.*)
- tests: Oklab reference values, self-match is a no-op, a known grade is recovered (residual
−86 %), skin protection holds the skin hue shift to ≤ 8° (51° without), engine and headless
UI tests
- docs: control protocol (colour commands and ids), testing; ROADMAP M8 row and log
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Missing media no longer fails: opening a project scans for missing files
(`media.findMissing`, `file.open` → missingMedia). Missing clips render our own
offline slate (render::offline: red striped field, warning triangle, "MEDIA NOT
FOUND", file name, hint) with silent audio, and the Link Media dialog opens.
- Media identity: imports record the size and a fast content fingerprint (hash of
size + first/last MiB) in MediaClip::identity (optional field, older builds
ignore it, no schema bump).
- media.relink checks file name / extension / fingerprint ("Clip ID") / duration /
media start / frame size+rate+channels, refuses a mismatched fingerprint unless
forced, supports Align Timecode (Project::shift_media_time), and relinks the
others automatically by the derived folder remap (Windows/macOS separators).
- media.autoRelink {from,to} | {folder}, media.search, media.replaceFootage,
media.makeOffline {deleteFiles?}, media.offlineAll, media.status, media.linkMedia.
- Media pool caches sources per reference (path, offline flag), so relink, Make
Offline and undo take effect at once; proxy-aware resolution and a full-res
provider that export uses.
- codecs::gop no longer holds its lock while decoding: rayon decoders inside a
parallel export could steal another frame of the same source and deadlock.
- UI: Link Media dialog (rows, match options, folder search with preview, Locate…,
Link, Offline, Offline All, Cancel), Make Offline dialog, Project panel offline
markers and context menu, media thumbnails keyed by media reference.
- Tests: moved media opens offline (slate rendered) and relinks others by folder,
prefix remap and folder search, fingerprint mismatch refused, Make Offline,
headless UI test of the dialog.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- model: SequenceSettings::color (ColorPipeline: working space, wide gamut, Auto Tone Map Media;
serde default Rec. 709) and Interpretation::color_space (override; None = metadata)
- render::colorman: source colour space from the override or VUI/colr/MKV metadata, decode tables
(camera log with super-whites, PQ, HLG + OOTF), gamut matrix, tone mapping into SDR working
spaces, gamut mapping; working → SDR monitor transform at the top level
(RenderOptions::working_output keeps working values for HDR export and scopes). Plain Rec. 709
media in a Rec. 709 sequence keeps the old fast path and stays on the GPU; managed layers and
HDR/wide-gamut sequences composite on the CPU
- export: H.264 and ProRes of a PQ/HLG sequence encode BT.2020 PQ/HLG and signal it: VUI / ProRes
header, colr (nclx/nclc), and for PQ mdcv + clli boxes and H.264 SEI (payloadTypes 137/144);
`sdr: true` (and render previews) export tone-mapped SDR. isobmff reads/writes mdcv and clli
- commands: sequence.colorSettings (Sequence ▸ Color Management…), clip.interpretFootage
(Clip ▸ Modify ▸ Interpret Footage…, Project panel context menu), color.spaces,
media.colorInfo; dialogs with automation ids colorDialog.*
- scopes: HDR sequences show a cd/m² waveform on a PQ scale (203 marked) and a BT.2020
vectorscope from the working-space render (scopes.hdrWaveform)
- display-referred wide-gamut input is desaturated into gamut keeping in-gamut colours exactly;
the soft compression is reserved for camera gamuts
- tests: colorman unit tests, engine (interpret/undo, PQ working space values and display),
goldens log_to_rec709 and hdr_tone_map (own renders + sidecars), export PQ/HLG × ProRes/H.264
read back by our demuxer and ffprobe (color_transfer/primaries/space, mastering display and
content light side data), headless UI (dialogs, HDR scopes)
- docs: architecture §5.2 (pipeline, HDR export, display colour management: no macOS EDR yet)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- crates/ui-egui/src/panels/essential_sound.rs: per-type sections (Loudness with Auto-Match,
Repair, Clarity, Creative, Ducking with Generate Keyframes, Duration note, Pan), preset dropdown
with save/delete, code-drawn type glyphs, automation ids for every control; every control runs an
essentialSound.* command and a slider drag is one undo step
- headless kittest test (essential_sound_ui.rs) with optional wgpu snapshots
- docs: architecture §5.2, control-protocol ids, testing, ROADMAP
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Play waits for the first PREROLL_FRAMES (6) frames, at most 0.5 s, before starting the clock
and audio: the first frames after Play were due before any decoder could produce them.
- A new on-screen request for a view drops the queued and cancels the running request it
replaces (scrubbing used to decode every visited position oldest-first).
- playback_plan: when per-frame CPU work (effects; decode excluded, it is sequential anyway)
exceeds what the workers can do in real time, start only frames that can still be on time and
space them evenly, instead of every frame arriving late. Cost estimate updated from finished
and (as a lower bound) cancelled prefetch jobs. No change while the workers keep up.
- PlaybackMeter::resync: frames while the window is hidden (occluded/minimized) are not drops.
- Scopes render after the Program prefetch while playing (prio 40, was 2).
- ui.inspect playback.preroll; docs/architecture.md frame scheduling.
bench, same-session A/B, median of 3 (GPU path, Full, 8 s, load avg 22–64):
h264-1080 190/2 → 192/0 shown/dropped (3/3 runs 0 dropped)
stack3 189/3 → 192/0; h264-2160 173/19 → 192/0; demo 186/6 → 189/3
seek-storm: 10 → 15 of 40 seeks shown within 150 ms, 1581 → 1300 ms CPU per seek
after-preview live effects play (median of 2): 118/74 → 134/58, cancelled jobs 74 → 48;
plays from previews 192/0 both
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Program monitor overlay for graphic clips: Type tool click creates a text layer (in the
selected graphic under the playhead or a new clip) and edits it in place with a caret,
selection highlight, arrows/word/line navigation, Shift-select, Cmd+A/C/X/V, Return,
Backspace/Delete, Esc; typing is one undo step. Selection tool selects, moves and scales
layers (bounding box, handles, anchor); double-click edits text. Rectangle/Ellipse tools
drag out shapes; Pen tool places a closed path. Automation ids program.layer.*.
- Properties (when a graphic is selected) and Essential Graphics: Layers list, Align and
Transform (align / distribute), Text (source text, font family + style incl. system fonts,
size, paragraph alignment, tracking, leading, baseline shift, box width, faux bold/italic,
caps, underline, kerning/ligatures), Shape, Appearance (fill, 2 strokes, background,
shadow). Automation ids graphics.*.
- Headless UI test (egui_kittest) drives the whole flow through the control channel; with
FILMCRAFT_UI_SHOTS it renders PNGs via wgpu (works with a locked screen).
- docs/graphics.md; architecture, ROADMAP, ATTRIBUTION (code-drawn glyphs) updated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Engine shortcut set (crates/engine/src/shortcuts.rs): bindings as data
(command, canonical chord, optional panel), panel shortcuts override app ones
while the panel has focus; per-OS modifiers (Cmd = Cmd/Ctrl, Ctrl = macOS
Control, folded into Ctrl elsewhere and reported as conflicts); conflict
detection, reassignment, undo/redo, Cancel restores; frontend commands
registered so tools/transport/panels are bindable too.
- Presets: FilmCraft Default, Premiere Pro / Final Cut Pro / Avid Media
Composer Compatible (key tables mapped onto FilmCraft command ids); custom
presets as JSON in <data dir>/Keyboard Shortcuts/, active set persisted,
Save As / Delete / Export / Import (Windows-notation files converted).
- shortcuts.* commands for agents/MCP (list/get/set/clear/undo/redo/conflicts/
forKey/resolve/presets/loadPreset/savePreset/deletePreset/export/import/audit).
- Premiere-default audit: commands without a shortcut get Premiere's key when
free (Shift+D Apply Default Transitions, Shift+J/L slow shuttle, Shift+K Play
Around, Project > Shift+O Open in Source Monitor); Audio Clip/Track Mixer
window shortcuts corrected (Shift+9 / Shift+6).
- Edit > Keyboard Shortcuts... (Cmd+Alt+K): keyboard drawn in code with
app/panel/both colouring and modifier filter, searchable command list with a
record-on-click Shortcut column and +, Key: detail list, conflict warnings,
Undo/Redo/Clear/Cancel/OK; automation ids shortcuts.*.
- Menus (in-app and native macOS key equivalents), command.list and the input
loop follow the active bindings live.
- Tests: 7 engine tests; scripted kittest test (record keys, live menu update,
reassignment, undo/redo, cancel, OK, panel-specific Left in History).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- New L1 crate filmcraft-text: fonts (bundled Inter Regular/Medium/SemiBold/Bold/Italic,
JetBrains Mono, Noto Serif; DirectorySource scans system font folders reading only name/OS2
tables; family/style resolution with synthetic bold/italic; per-char fallback), layout
(harfrust kerning/ligatures, unicode-bidi, unicode-linebreak, alignment incl. justify,
leading, tracking, baseline shift, all/small caps, underline, carets/hit-test/selection,
cached), raster (moved from captions; active-edge list, shapes, emboldening), render
(outline + sub-pixel glyph-mask caches; transformed-outline path for rotation/skew),
mask (distance-transform strokes, blur).
- Captions burn-in now shapes and kerns via the engine (font.rs/raster.rs removed).
- Timecode / Clip Name effects draw real text (full clip names, not seven-segment digits);
text_burnin golden re-blessed.
- UI takes its font bytes from the text crate (no duplicate copies in the binary).
- New fonts: Inter Bold, Inter Italic (OFL-1.1, rsms/inter v4.1), Noto Serif Regular
(OFL-1.1, notofonts) with sidecars, ATTRIBUTION.md rows and licence text.
- Perf: 3-line 96 px title at 1080p ~1.6 ms cold / 0.15 ms warm (release).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Root cause: not a lock. When the window is hidden or behind others (always when the app is
driven over the control channel), macOS App Nap drops every thread of the process to background
priority 4 and defers its timers. On a loaded machine (load avg ~150) a priority-4 process gets no
CPU: after Render In to Out + play the UI thread stayed runnable inside the run loop's timer
callout, every thread at PRI 4 (`ps -M`), 0 % CPU, control channel silent. Relaunched with App Nap
off the same script ran through: render 2–13 s instead of 15 s, every reply immediate (2/2 runs).
- apps/filmcraft/src/app_nap.rs: hold a user-initiated, latency-critical NSProcessInfo activity
for the life of the process (= NSAppSleepDisabled for a bundle; idle system sleep allowed).
- ui.screenshot: a hidden window or sleeping display never presents, so the screenshot event
never came and the UI repainted at full speed forever (100 % CPU). It now fails after 10 s with
"no frame was presented (window hidden or display asleep)" and polls at 60 Hz meanwhile.
Regression test: app_nap::tests (the activity is taken on macOS). The starvation itself needs a
hidden window on a loaded machine and is checked with the control-channel script in the commit
notes (captions.newTrack/add, markIn/Out, Full, renderInToOut wait, stop, playhead.set,
renderBar, ui.screenshot): all replies within 10 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- crates/ui-egui/examples/bench_playback.rs drives the Program monitor's own FrameServer,
prefetch (now FrameServer::schedule_playback, shared with the monitor) and PlaybackMeter at a
simulated 60 Hz display, optionally compositing on a headless wgpu device. Scenarios:
h264-1080, stack3 (3 × 1080p, scaled/rotated/opacity), h264-2160, demo, after-preview
(live play → Render Effects In to Out → 2 plays), seek-storm. Reports latency percentiles,
on-time ratio, worker/decode wall and thread-CPU time, UI-thread present time, process CPU per
frame, decoder seeks/samples (codecs::gop_stats) and wasted jobs; --json for machine output.
Fixtures are generated with ffmpeg into target/fixtures/playback.
- FrameServer records per-job timings (thread CPU via rustix clocks) while profiling is on.
- Dropped/shown were counted per UI refresh: "shown" counted every refresh that had the frame
(322 "shown" in an 8 s play of 192 frames) and "dropped" counted a frame that arrived a few ms
after its first refresh. PlaybackMeter counts timeline frames: shown when the exact picture was
on screen while due, dropped otherwise, including frames passed over without a refresh.
Baseline (GPU path, 8 s, load avg 104–168 on 14 cores; CPU columns are load-robust):
h264-1080 Full 191/1 shown/dropped, 46 ms CPU/frame; Half 188/4
stack3 Full 189/3, 102 ms CPU/frame; Half 157/35 (3 extra decoder seeks)
h264-2160 Full 8/184, 313 ms CPU/frame, 10 seeks; demo Full 114/78, Half 191/1
after-preview: live 0/192; 1st play after render 124/67, 2nd 119/73 (UI-thread present
p50 21 ms / p95 45 ms); seek-storm: 0 of 40 seeks shown within 150 ms
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
