recipe.mjs and prefs.mjs printed recipe.json and preferences.json values
raw: names, workflows, approval dates, brief values, confirmed-in project
names, and the known-recipe list in an unknown-name error. Both files
travel with a project, so an escape sequence in them could rewrite or
hide earlier output, set the window title or write the clipboard in
terminals that allow it. The SFX provider's empty-search line repeats the
intent, and its Openverse fallback line repeats a network error.
Each of those lines now goes through terminalText, like the resolve and
adopt lines beside them. --json output is unchanged.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* test(core): svg attribute selectors rewrite in linear time over many fragment references
* perf(core): index a scope's svg reference aliases once by id and attribute name
* refactor(core): split svg attribute alias lookup out of the selector rewrite
* test(core): refresh page goldens and font mirror for the smaller inlined svg alias code
* test(core): repeated references to one svg symbol rewrite in linear time
* test(core): count alias checks so repeated svg references show quadratic work
* perf(core): index each repeated svg reference alias once
* fix(studio): a sealed preview's timeline shows its text clips' words and look
A sealed preview's timeline rows come from the clip list the film sends, which
carried no text, so its text clips drew as blank rows. The film now sends each
plain text layer's words (up to 200 characters) and look, read by the same rule
Studio uses for its own rows, and the row shows the words as plain text.
* fix(studio): check a sealed clip's text on arrival, cap only what the film sends
Studio now takes a clip's text from the film's message only when every field is
text within the film's limits; otherwise the clip keeps its row without words.
The 200-character limit applies to the film's message alone, so an open
preview's own rows keep their full words as before.
* test(parsers): a tab in a concat list path is kept
* fix(parsers): refuse only what ends a concat list line, so a tab in a folder name renders
* test(parsers): a real ffmpeg join keeps every clip under a tab-named folder
* test(parsers): end ffprobe options before the joined file in the tab join test
Layering walked the whole host data-variable-values map and the whole render
overrides map for every declarer and every composition id, so many declarers
that emit nothing were never charged to the size budget. Each declarer now
walks only its own declared ids, and a host's overlap with the overrides is
computed once per host. Output is unchanged except that names inherited from
Object.prototype are no longer treated as declared.
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
* test: a concat list refuses paths whose folder name holds a line break
* fix(engine): write every concat list through one writer that refuses control characters
* test: name the line-break folders without slashes; skip them on Windows
* Revert "fix(engine): write every concat list through one writer that refuses control characters"
This reverts commit 97d482340a.
* Reapply "fix(engine): write every concat list through one writer that refuses control characters"
This reverts commit 7d723fd276.
* test(engine): a chunk concat list ends with a newline
* fix(producer): raise the sub-composition expansion limit to 64 MB
A render may now inline up to 64 MB of composition HTML across its mounts,
four times the 16 MB per-file limit, so a project with several large
distinct scenes is no longer refused.
* test(producer): size the generated-output tests from the expansion limit
Four size-limit tests used fixed counts that only exceed the current
32 MB expansion limit. Compute each count from
MAX_EXPANDED_COMPOSITION_BYTES so the tests keep reaching the limit if
it changes.
* test(studio): a sealed preview keeps its loading overlay while media buffers
* fix(studio): a sealed preview's loading overlay waits for its media through the editing port
* fix(studio): Studio never calls a preview ready before its asset check has answered
* refactor(studio): the preview asset wait lives outside the Player component
* refactor(studio): the preview asset wait moves to its own file, out of the Player
* refactor(studio): one asset-wait state; a sealed frame's first check is its own value
* fix(studio): a sealed preview's missing asset answer is asked again, never taken as ready
* fix(studio): a closed preview frees its editing-port client
* style(studio): keep the client comment under the line limit
* fix(cli): a project .env provider key gives way only in a received project
Your own or a cloned project's .env provider keys are used as set again, as
the docs describe. Capture no longer drops a project .env vision key next to
a shell credential, and the audio scripts no longer drop an own project's
Gemini key next to the shell's other Gemini key. A received project's keys
still give way to your own credential, and the notes that name a project
key when it is used stay.
* test(cli): check the note for a cloned project's key; say when a received key is ignored
* fix(studio): a sealed preview keeps Studio's size, move and turn edits after a seek
* fix(studio): a sealed preview reapplies Studio's edits on each frame it seeks or plays to
* fix(studio): build the sealed reapply script from constants, never a caller's value
* fix(studio): a sealed preview reapplies Studio's edits after a redraw at the same time
* fix(studio): serve the seek reapply only into sealed preview documents
* chore(core): shorten the redraw count's doc line
Capture's image captioner can now authenticate to Vertex AI with
HYPERFRAMES_VERTEX_PROJECT_ID plus HYPERFRAMES_VERTEX_ACCESS_TOKEN, as an
alternative to a service-account JSON. A caller that mints a short-lived
token per run no longer has to hand the CLI a long-lived key. The token is
used as given and never refreshed; when both are set, the token is used.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(studio): a script edit re-runs in place in a sealed preview, without a reload
* fix(studio): a sealed timing edit takes the saved clip and film timing along with its soft reload
* test(studio): the sealed timing test checks a soft reload was sent before reading it
* refactor(studio): the sealed soft reload code passes the quality audit without suppressions
* chore(core): core's size budget makes room for the sealed preview's frame-side editing code
Main already measured 1,841,737 bytes packed against a 1,830,000 budget; the sealed soft reload brings it to 1,855,277. Budgets follow SIZE_BUDGET_HEADROOM: ceil(1,855,277 x 1.05) packed, ceil(4,665,659 x 1.05) unpacked.
* fix(core): a sealed timing edit is not skipped when a nested file's root repeats a clip's id
* test(engine): a capture page whose tab Chrome closed while opening it is replaced once or named
* fix(engine): reopen a render page whose tab Chrome closed while opening, or say why it can't
* style(engine): keep openPage's comment to two lines
* fix(studio): a sealed resize saves its anchor on the route GSAP moves it by
* style(studio): format the sealed resize test and drop the new comment lines
* fix(skills): the launcher's floor is 0.9.67
The skill launcher now refuses any HyperFrames release older than 0.9.67, the latest
release and the oldest one the HyperFrames Desktop app will accept. This applies to
--cli-version, to HYPERFRAMES_STANDALONE_CLI_VERSION, to a helper started with --script and
to a plugin manifest pin. A standalone install's default npx range starts at 0.9.67. Every
plugin manifest in this repo already pins 0.9.67.
The skills guide and the plugin installation reference name the new floor.
* docs(skills): the launcher floor stays at or above Desktop's oldest accepted release
Desktop currently accepts 0.9.66 and later, and a pending Desktop change raises that to
0.9.67. The launcher's comment now says its floor stays at or above the oldest release
Desktop accepts, without naming a Desktop version, so it stays true when Desktop moves.
* fix(studio-server): proxy piece list names only the pieces, not their folder
The chunked proxy join wrote each piece as an absolute path into a concat
list in the system temp folder. A line break in a folder name above the
project ended an entry early, so the rest of the path could be read as
further list lines naming other local files.
The list now sits in the cache folder beside the pieces and names each
piece by its own generated file name, and ffmpeg runs the join from that
folder. No project or folder name is written to the list, so folder names
with line breaks, quotes, '#' or '?' all join only the copy's own pieces.
The CLI publish bake uses the same join and is covered too.
* fix(studio-server): a proxy copy lock dated ahead of the clock no longer holds waiters
A copy waits for another process's lock while that owner's lock, temp and
piece files keep changing. The newest modified time was taken as is, so a
lock or temp file dated in the future looked like fresh progress forever,
and the waiter held its transcode slot without end. Preview and the
publish bake could then wait with nothing running.
Each file's modified time now counts as no later than the moment the
waiter first saw that time. A file dated ahead of the clock ages from
then, so the waiter gives up within the stall window and frees its slot.
An owner that keeps writing still sets a fresh time on each change and is
waited for exactly as before, even beside a future-dated file.
* fix(studio-server): keep the absolute piece list and default folder on Windows
On Windows the working folder is part of the DLL search, and a long cache
folder path may fail to start a process from it. Windows paths cannot hold
a line break, so the piece join there lists absolute piece paths again and
starts ffmpeg from the default folder. Elsewhere it still runs from the
cache folder with bare piece names.
The line-break test now injects an unquoted list line, so the old join
really pulls in another file, and it checks that the copy holds none of that
file's frames. A spawn test checks the join's folder and bare list name.
* fix(studio): the property panel shows a sealed preview's animated values
* fix(studio): a sealed preview's panel shows rotation and scale from tweens that only turn or grow
* feat(engine): renders send page traffic through the egress proxy in HYPERFRAMES_CAPTURE_PROXY
* feat(engine): render media downloads go through the egress proxy in HYPERFRAMES_CAPTURE_PROXY
* fix(engine): downloads through the egress proxy keep undici's fetch, so gzip text and headers work
* fix(cli): batch render, cloud render and publish print project text as visible escapes
A batch render printed each failed row's error straight to the terminal. The
error can quote file names and attribute values from the project, so control
characters in them could move the cursor, rewrite earlier lines, set the
window title or write the clipboard. That line, and the manifest path in the
batch header, now go through terminalLines() and terminalText(), which keep
real line breaks and show every other control character as an escape.
cloud render does the same for the media lines it offloads, the "also stays in
the zip" line, the largest-files list, the aspect-ratio detection line and the
names of fonts that are not on this computer. publish does the same for the
composition path in the blank-entry hint and for the publish failure message.
The manifest file and --json output are unchanged.
* fix(cli): sign-in callback ignores requests without the login's state
The browser sign-in listens on a local port for the redirect back from the
sign-in page. It acted on an error in that redirect before checking the
state value, so any page or local process that reached the port could end a
pending sign-in and put its own text in the terminal. A request with a
missing or wrong state also ended the sign-in.
Now only a request carrying the login's state can end it. Any other request
gets the same 400 page as before and the listener keeps waiting for the real
redirect or the timeout. A sign-in that is cancelled or fails on the sign-in
page still carries the state, so it reports the failure as it does today.
The error text from the redirect, and the "Sign-in failed" line, now show
control characters as visible escapes.
* fix(cli): variable issues and oversize file list print control characters as escapes
Variable issue lines from render and render --batch printed the variable ids
and allowed options from the project's data-composition-variables as they
were. Each line now goes through terminalText(), so control characters,
line feeds included, show as visible escapes. The issue objects and their
messages are unchanged for other callers.
When a cloud render project is over the upload limit, the list of its largest
files now escapes each path, line feeds included, so a file name cannot add a
line of its own.
Font names in the cloud render notice are looked up as written and escaped
only when printed, so a name with a trailing control character keeps its
look-alike suggestion.
* fix(cli): a callback from another sign-in says the terminal is still waiting
The 400 page for a callback without the login's state now tells the person to finish signing in
from the newest tab, since the terminal keeps waiting instead of ending the login.
* fix(studio): copy, duplicate and audio lane updates work in a sealed preview
* fix(studio): a sealed copy keeps its starting selection, and a stale lane read is dropped
* fix(studio): a saved lane value outlasts an older sealed read, and cut deletes in its own project
* fix(studio): a sealed preview pauses on press and reopens a link's selection
* fix(studio): a linked selection retries on a timer, whatever the playhead
* refactor(studio): one lookup-state hook keeps the URL-state hook's size where main had it
* fix(studio): a linked group member keeps the lookup waiting until the preview can answer for it
* fix(cli): reject bad billing --timeout and --needed values as usage errors
Closes#7909
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* refactor(cli): share one positive-integer flag parser across billing, compare and share
Refs #7909
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix(producer): resume checks the segment file the join reads, not the manifest's path
A resumed segmented render now validates each reusable segment at its own
computed path in the segment folder. It must be a plain file, not a link, and
match a slice of this render's plan. The path a manifest records is never
opened or probed.
* fix(engine,producer): concat joins open only the list and encoder containers
Both concat-demuxer joins (the engine's chunk and segment concat and the
distributed assemble join) now pass a format whitelist of concat, mov and
matroska. ffmpeg applies it to every file the list opens, so a listed file
that is a playlist or manifest cannot pull in other files, while mp4, mov and
webm joins work as before.
* fix(producer): resume reuses only segments this install recorded
Segment manifests are now signed with an HMAC keyed by a per-install secret
kept in the HyperFrames config folder, outside every project. Resume trusts a
manifest only when its signature checks out, so a segment folder copied from
another machine or shipped inside a project is captured again, while a
render's own interrupted run resumes as before.
* docs(producer): restore the segment path ordering comment and rewrap resume docs
* fix(producer): replace an empty or short resume key file
A key file too short to use is now replaced by a fresh key instead of
disabling resume for good. A missing key is still linked into place, so
renders that create it at the same time share one key; a test runs eight
processes against one path to check that.
* fix(producer): warn under --resume when no resume key is available
When the resume key can be neither read nor created, for example on a
read-only home folder, a file system without hard links or a serverless
runtime, --resume now logs one warning that every segment will be captured
again instead of only reporting zero complete segments.
* fix(engine): fail a concat join at a list entry ffmpeg refuses
The concat demuxer ends the stream at a listed file it cannot open and ffmpeg
still exits 0, which cut the output short without an error. -xerror on the
concat input makes the join fail instead. mp4, mov and webm joins still exit
0.
* fix(producer): resume checks each reused segment's SHA-256
Each finished segment's SHA-256 is now recorded in the signed manifest. The
hash is taken once when the segment closes, and on resume only the files
about to be reused are hashed again, so a file swapped for another of the
same size is captured again.
* docs(cli): note that resume re-checks segment hashes and recaptures older segments once
* fix(studio-server): proxy joins only its own pieces, through the shared concat input
Studio's chunked proxy copy now names its pieces with a tag keyed by the
user's preview token, so pieces that arrive inside a project folder sit at
names it never looks up and every piece is encoded fresh. A stopped copy of
the user's own still continues from its finished pieces.
The piece join uses concatListInputArgs, now in @hyperframes/parsers so the
engine, producer and Studio share it. It allows only local files in the
concat, mov or matroska formats and fails at a refused entry. The cache sweep
recognizes both tagged and older piece names.
* test(parsers): prove the concat protocol whitelist with a data: entry
ffmpeg already refuses http:// in a concat list opened from a file, so the
test passed without the whitelist. A data: entry opens by default; the test
now asserts it is refused as not on the 'file' whitelist.
* fix(studio-server): only the copy's own tagged pieces count as lock progress
The copy lock counted any .part-<key>- file as the owner's progress, so a
piece shipped with a future modified time kept a stalled lock looking active.
It now counts only names under the copy's own tagged piece prefix. The cache
sweep removes untagged pieces past the usual age even while that copy runs,
and measures age in both directions so a future-dated piece is swept too.
* fix(studio-server): say once when proxy pieces cannot be keyed to the user
Without a preview token, piece names use a per-process secret, so pieces are
reused only within that Studio process. Studio now logs that once, so a copy
restarting from its first piece after a restart has a visible cause.
* test(studio-server): name the proxy join test for what it checks
* fix(producer): derive the segment manifest key from the preview token
Sign segment manifests with an HMAC key derived from the per-user preview
token instead of a separate key file. The token is created with an exclusive
open, so concurrent renders never end up with different keys, and it is only
used when the file is private to this user.
Resume now checks a segment's SHA-256 before probing it, and a real-ffmpeg
test documents how the concat join treats an entry that is itself an ffconcat
list.
* docs(producer): describe the manifest key per user and the no-token case
The generic "Try --docker for containerized rendering" hint was printed
under every failed render. When the failure is a video source FFmpeg could
not read, a missing file or a 404 URL, a container reads the same source
and fails the same way. Those three now get a hint about the file.
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
"1,080" and "3.5%" were split at the separator and counted as two words,
so a two-word line such as "1,080억 달러" was timed as three and its
shortest allowed hold went from 0.4 s to 1.05 s.
Co-authored-by: Claude Fable 5.2 <noreply@anthropic.com>
* fix(studio): a value or keyframe edit shows in place in a sealed preview
* fix(studio): a sealed group edit reloads once at its end, and a batch renders its last tween
* refactor(studio): the sealed tween patch path lives in its own module
Keeps useGsapScriptCommits under Studio's 600-line limit, and records the authoredOpacity helpers' move into core as an intended deletion.
* refactor(studio): the sealed tween patch code passes the quality audit without suppressions
Splits the frame-side change parser per kind, the tween lookup and keyframe rebuild into small checks, and the reachable-preview patch out of applyPreviewSync; drops an unused type re-export.
* fix(cli): play and present answer unknown paths with an inert refusal (SECOPS-2642 part 3)
An OAuth return that lands on the play or present port was a 404 document on
the same origin as the composition frame, so project script could read its
?code=&state= through a frame, a popup or the authorize redirect. Unknown
paths, missing composition files and any request carrying code or state now
get a fixed text/plain 404 on an opaque origin (CSP sandbox) that cannot be
framed and drops its opener (COOP same-origin).
play's app setup moves into an exported buildPlayApp so tests can serve it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(cli): trim the inert refusal's comments to the package share
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(producer): a render that fails on a missing file says what was gone again
Restores the diagnostic the previous commit on main took out, as it was; the commits after it confine what it looks at.
* test(producer): render failure diagnostics never probe a path from error text or outside the render
* fix(producer): probe a missing path only from a real error, inside the render's folders
describeMissingPath now reads only the path Node set on an ENOENT error, never
one quoted in a message the page can write, and looks on disk only under the
render's temp root, extract cache or work folder. Any other path is named from
its text, so a network share is never contacted and the facts cannot tell an
existing path from a missing one.
* test(producer): a missing-path probe never passes an outside link or climbs above its folder
* fix(producer): walk down from the render's own folder and check each link before following it
* build(producer): give the packed size budget its 5% headroom with the missing-file diagnostic back
* test(studio): a sealed scene's children and clip parents fill as an open preview's do
* fix(studio): a sealed preview's scene children and clip parents come from the editing port
* test(studio): a sealed preview's first-frame event carries its clips and length
* fix(studio): a sealed preview's first-frame event counts its clips from the document copy
* refactor(studio): drop two comments the names already carry
* test(core): reading an svg id scope's size never walks its ids
* perf(core): count an svg id scope's renames once, not on every stylesheet it rewrites
* test(cli): time where the very-many-files case spends its time on each runner
* fix(cli): the received-render size tests cap the walk instead of writing 20,000 entries
* fix(studio-server): an inserted scene names its file from the project root, as the renderer reads it
* test(studio-server): keep a test comment under the line limit
* fix(studio): refused group, freeze frame and animation saves keep the server's status
* refactor(studio): withStatus says what it does without a comment
* test(producer): remote scripts and stylesheets inlined at compile time stay within the size limit
* fix(producer): fetch each remote script and stylesheet once, a few at a time, within the size limit
Every inlined copy is charged to the render's expansion budget as it arrives;
going over fails the compile with COMPOSITION_TOO_LARGE and stops the
downloads still queued. Remote CSS is spliced in literally, so `$` sequences
in it are no longer expanded.
* fix(producer): drop the budget label comment the comment gate counts
* fix(producer): count a remote script when it arrives and stop the queue on an integrity mismatch
* test(producer): font normalization tests answer Google Fonts from a stub, not the network
* style(producer): wrap the stubbed font response
#8173 set the budget to 82,000 against 81,378 bytes measured, about 0.8%
over. The budget file's own rule is 5% above measured, so the next small
change would trip it again: ceil(81,378 × 1.05) = 85,447.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The bench's openStudio now waits in the page for Studio's own word that the
composition loaded (studio_look reports a duration) under the one 90 s boot
budget, instead of a separate 30 s poll that failed a case Studio was still
loading after a reload under load. A boot that never loads names Studio's last
report, and seekBound names the condition it was still waiting on.
hiddenTarget no longer throws on a preview frame detached mid-scan: puppeteer
throws that synchronously, so the per-frame .catch never ran.
* revert(producer): a render that fails on a missing file stops naming what was gone, for now
This reverts commit cfea8cfff5 (#7955). The detail comes back in a follow-up that only looks inside the folders the render owns.
* revert(producer): name the files the #7955 revert removes in the deletion guard
* fix(aws-lambda): match the shell to its own pin, ship only hashed files, refuse the shell ZIP in CDK and SAM
Review follow-ups to the pinned shell route:
- build-zip --source=chrome-headless-shell accepts the folder only when its
digest is pinned for this version, not for any pinned build.
- It leaves out the files the digest skips (debug.log and chrome_debug.log,
which Chrome writes as it runs), so nothing unhashed ships.
- The CDK construct throws on chromeSource "chrome-headless-shell" and the
SAM template allows only sparticuz, as lambda deploy already refused it:
a shell ZIP cannot fit a function. The CHROME_PATH wiring for that ZIP is
gone with it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore(aws-lambda): raise the packed size budget to 82,000 bytes
The shipped build-zip script now checks the shell against its own
version's pin and skips the files the digest leaves out. That takes the
tarball to 81,378 bytes in CI, over the 80,500 budget's 1% tolerance;
main was already within about 200 bytes of it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>