* fix(studio): prevent preview hang on burst external file rewrites
Two interacting bugs caused Studio to freeze when multiple processes
(generator, check, snapshot) burst-wrote index.html within seconds:
1. SSE listener leak: the /api/events handler added a watcher listener
per client connection but never removed it on disconnect. Reconnects
accumulated dead listeners, each triggering readFileSync on every
file change and writing to closed streams.
2. Generation starvation: processChange incremented generationRef and
awaited drainPendingChanges. A second event arriving mid-drain bumped
the generation, causing the first drain to bail at the generation
check. With rapid writes, no drain ever completed and Studio stayed
frozen on stale content.
Fix 1: use stream.onAbort() to remove the watcher listener when the
SSE connection closes.
Fix 2: gate processChange with a draining ref. While a drain is in
progress, stash the latest event. On completion, process the stashed
event — the last write in a burst always completes its reload.
Closes#3646
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(studio): align coordinator tests with drain serialization
Update existing test to expect the new behavior: when two events
fire in quick succession, the first drain completes and triggers a
reload (previously it was silently discarded). The stashed event
then starts a second drain.
Also fix the burst-write test to use the drains array pattern and
explicit act() flushes for stashed event processing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(studio): stash events with allowDuplicate and harden listener cleanup
Address Rames's review findings:
- Stash with allowDuplicate: true so re-dispatched events are not
swallowed by the duplicate guard (the identity was already written
on the way in, so the stashed event matched itself on re-entry).
- Wrap SSE keepalive loop in try/finally so the listener is removed on
both abort and throw, not just abort.
- Restore stale-completion guard test coverage lost in the rename.
- Use await act(async () => {...}) for burst dispatches so assertions
depend on the stash guard rather than scheduling.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix(studio): simplify drain serialization and restore SSE cleanup
Restructure processChange into intake + drain loop:
- processChange is now synchronous — validates, dedupes, checks own
echoes, enqueues the accepted payload, and starts the drain loop
- startDrainLoop runs while the pending slot is non-null, draining
one event per iteration via drainOnePending
- No recursive void processChange(...) from finally, so no
allowDuplicate escape hatch needed — stashed events never re-enter
intake guards
SSE listener: restore stream.onAbort alongside try/finally. Hono's
sleep() never throws, so finally alone doesn't fire on disconnect.
Both paths call removeListener (Set.delete is idempotent).
Tests: remove stale-drain test that contaminated subsequent tests by
emptying the shared roots array mid-test. Use sync act() for burst
dispatches — the stash decision is synchronous.
All 10 coordinator tests pass locally (NODE_ENV=test).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(studio): let an agent drive Studio's selection and playhead
Adds `studio_select` and `studio_seek`, so an agent and the human are looking
at the same element and the same instant. Selecting reveals the inspector,
exactly as a click does, which is what makes the agent's move visible.
Selection is shared state, not a per-call argument, and that is forced rather
than chosen. Most of Studio's edit handlers read the ambient React selection,
and `applyDomSelection` only schedules a state update, so selecting and
committing inside ONE call would write to whatever was selected before. Two
tool calls are separated by a render, so the contract is select first, then
act. That is also how a human works: click, then type.
`studio_seek` uses `requestSeek`, not `setCurrentTime`. The latter only moves
the timeline's displayed number and leaves the composition where it was.
Two things the tools refuse to fake:
Seek does not clamp. `seek()` already clamps against the adapter's duration,
which can differ from the store's, and clamping again would give that
invariant two owners that can disagree. The tool reports where the playhead
actually landed instead, read back afterwards.
`requestSeek` is fire-and-forget, so it cannot report that no adapter was
mounted to receive it. The tool compares the playhead before and after and
fails rather than claiming a seek that never happened.
Select separates three failures that a single message would have merged: the
preview is not mounted yet (wait), no element matches the handle (re-read),
and the element cannot be selected (try a neighbour). The agent's next move
differs for each, so collapsing them would cost it a round trip or a retry
loop.
* feat(studio): give an agent eyes with studio_frame
Renders the composition to a PNG at a given time and returns the URL. This is
what turns the tool set from a remote control into a loop: author a change,
capture the instant it affects, look, adjust. No agent can judge motion from
source, because "what does this look like at 2.4 seconds" is not a question a
file answers.
Reuses Studio's existing capture endpoint via `buildFrameCaptureUrl` rather
than inventing a second one.
Two things this does not fake:
It reports the time the playhead LANDED on, not the time requested. The player
clamps, so those differ at the ends, and attaching the wrong time to a frame is
how an agent draws a confident wrong conclusion about motion.
It waits before capturing, by default 150ms. The frame is rendered from the
file on disk, and the render cache is cleared by a file watcher with a 40ms
write-stability threshold, so a capture that beats the watcher renders the
PRE-edit composition. That exact staleness was a real bug here once. An agent
reading a stale frame as "my edit failed" would thrash, so the wait is on by
default, `settleMs` makes it tunable, and the tool description names the
failure rather than leaving it to be rediscovered.
It probes with HEAD before returning, so a URL that 404s comes back as a
failure with a hint instead of as a link the agent cannot render.
* feat(studio): add studio_inspect, so an agent reads before it writes
Everything about one element in one call: resolved styles, text fields, box,
data attributes, GSAP animations, and what the element will and will not
accept.
The point is to prevent a failed write rather than to satisfy curiosity.
`can.reasonIfDisabled` is passed through verbatim from Studio's own
capabilities, so an agent that reads first should never attempt an edit the
element would refuse.
Three things it refuses to get wrong:
Animations are reported ONLY for the current selection, because that is the
only element Studio parses them for. Attributing them to any other element
would be reporting the wrong element's motion, which is worse than reporting
none. When a handle names something else the field is empty and
`animationEditingBlocked` says why.
`animationEditingBlocked` also carries the two states where animation editing
is off entirely, multiple timelines and an unsupported timeline pattern. Both
live on the selection context. Learning them from a read costs one call;
learning them from a failed write costs a retry loop.
Inspecting a handle does NOT change what is selected. It is a read, and
stealing the human's selection would be a side effect they did not ask for.
There is a test asserting `applySelection` is never called.
Nothing selected and no handle given is a failure, not an empty result. An
empty result would assert "this element has nothing", which is a different and
false claim.
* feat(studio): let an agent edit text and styles, guarded
The first tools that change the composition. Both act on the current
selection and take no handle, which is forced rather than chosen: the
handlers read the ambient React selection, and `applyDomSelection` only
schedules a state update, so selecting and committing inside one call would
write to whatever was selected before. Select first, then edit.
Also plumbs the write-blocked state, which was the blocker for shipping any
write at all. `domEditSaveQueuePaused` and the external-file conflict both
lived on App and were unreachable from the tool surface, so `canWrite` was
optimistic and a comment said so. They now derive into a single
`writeBlockedReason` on the shell context: one field, one owner, conflict
taking precedence because resolving it is what unblocks the queue.
That guard matters more than it looks. Both states are BANNERS in Studio with
no lock behind them, so nothing else was stopping a programmatic write from
landing on top of a conflict the user had been asked to adjudicate.
Three things the tools refuse to fake:
They check the outcome, not the absence of a throw. Studio has several paths
where a failed commit resolves anyway, so awaiting the handler proves nothing.
The tagged outcome added earlier is what proves the write landed.
A partial style result is reported as partial. `handleDomStyleCommit` is one
property per call, so N properties are N commits; the result carries `applied`
and `rejected` maps rather than a single boolean that would have to pick a
side.
Style commits run sequentially, never concurrently. Two commits racing through
Studio's client-side read-modify-write can record undo entries that both claim
the same starting content. There is a test that measures concurrency rather
than trusting the loop.
Every decline reason maps to a hint naming what to do instead, so a refusal
routes the agent rather than just stopping it.
* feat(studio): move, resize and rotate, verified by reading back
`studio_transform` does what a drag does, and then checks. The box in the
result is READ BACK after the write, never echoed from the request, and
`applied` lists what actually took effect.
That is not belt-and-braces. The plan for this unit said to re-derive the
geometry handlers' behaviour rather than trust any description of them, and
doing that turned up three different behaviours behind one interface.
The handlers on `DomEditActionsValue` are the GSAP-AWARE wrappers, aliased in
`useDomEditSession.ts:534-538`, not the CSS ones in `useDomGeometryCommits.ts`
that an earlier note in this workstream described.
`handleGsapAwarePathOffsetCommit` and `handleGsapAwareRotationCommit` are
`if (gsapCommitMutation) { ...intercept... }` with no else branch. Their own
comments say the absence is deliberate: position and rotation are written as
GSAP code and there is no CSS fallback to write to. So they can return having
done nothing.
`handleGsapAwareBoxSizeCommit` is not like the other two. It runs through
`runGestureTransaction` with separate scale and width/height routes, so resize
works more generally.
Reading back is what turns that middle case from a silent lie into a reported
one. A move that did nothing comes back in `unchanged` with a reason.
Three smaller decisions:
Operations re-read between each other, so a move is judged against the box
AFTER a resize in the same call. Comparing against the original would credit
the resize's change to the move.
Rotation is reported as dispatched, not verified. `rotate` is an individual
transform property and does not appear in the computed transform, so there is
no honest box-derived signal, and claiming one would be worse than saying so.
x pairs with y and width pairs with height. Accepting one alone would mean
inventing the other from the current value, which moves the element somewhere
the caller did not ask for. The pairing rule and its minimum live in one
`parsePair` helper rather than as four separate branches.
* feat(studio): let an agent author motion
Four tools: add an animation, change its duration/ease/position, add a
keyframe, delete it. This is the capability that makes the tool set worth
having, because motion is the one thing an agent cannot judge or author from
source.
These are deliberately less confident than the rest of the set, and the
reason is the handlers underneath them:
`handleGsapAddAnimation(method)` takes only a method. Its insert position
comes from the live playhead, not the caller, and the call is `void ...catch()`
so it returns nothing.
`handleGsapAddKeyframeBatch` returns a promise but catches its own failure, so
awaiting proves the call finished, not that it landed.
`handleGsapDeleteAnimation` discards its promise entirely.
`handleGsapUpdateMeta` is the one honest signal. It returns a boolean.
U8 handled the same problem by reading the result back. That does not work
here: the animation list comes from React state that only refreshes on a
render, and no render happens inside one tool call. Rather than fake a
verification with a frame-timer, these report what was DISPATCHED and the
descriptions tell the agent to call studio_inspect to see the result. Saying
"I asked for this" is honest; saying "this happened" would not be.
Three consequences worth stating:
`studio_add_animation` takes no position. The handler reads the playhead, so
accepting one would report a number that had no effect. It reports where the
playhead actually was and tells the agent to seek first.
`studio_update_animation` rules out the no-selection case BEFORE dispatch. The
handler answers `false` for both "nothing selected" and "the write failed", so
eliminating one is what makes the other legible.
Keyframe percent and properties are validated in the tool, because nothing in
the platform checks input against the declared schema.
* feat(studio): add studio_inspect, so an agent reads before it writes (#3517)
Everything about one element in one call: resolved styles, text fields, box,
data attributes, GSAP animations, and what the element will and will not
accept.
The point is to prevent a failed write rather than to satisfy curiosity.
`can.reasonIfDisabled` is passed through verbatim from Studio's own
capabilities, so an agent that reads first should never attempt an edit the
element would refuse.
Three things it refuses to get wrong:
Animations are reported ONLY for the current selection, because that is the
only element Studio parses them for. Attributing them to any other element
would be reporting the wrong element's motion, which is worse than reporting
none. When a handle names something else the field is empty and
`animationEditingBlocked` says why.
`animationEditingBlocked` also carries the two states where animation editing
is off entirely, multiple timelines and an unsupported timeline pattern. Both
live on the selection context. Learning them from a read costs one call;
learning them from a failed write costs a retry loop.
Inspecting a handle does NOT change what is selected. It is a read, and
stealing the human's selection would be a side effect they did not ask for.
There is a test asserting `applySelection` is never called.
Nothing selected and no handle given is a failure, not an empty result. An
empty result would assert "this element has nothing", which is a different and
false claim.
* feat(studio): move, resize and rotate, verified by reading back (#3519)
`studio_transform` does what a drag does, and then checks. The box in the
result is READ BACK after the write, never echoed from the request, and
`applied` lists what actually took effect.
That is not belt-and-braces. The plan for this unit said to re-derive the
geometry handlers' behaviour rather than trust any description of them, and
doing that turned up three different behaviours behind one interface.
The handlers on `DomEditActionsValue` are the GSAP-AWARE wrappers, aliased in
`useDomEditSession.ts:534-538`, not the CSS ones in `useDomGeometryCommits.ts`
that an earlier note in this workstream described.
`handleGsapAwarePathOffsetCommit` and `handleGsapAwareRotationCommit` are
`if (gsapCommitMutation) { ...intercept... }` with no else branch. Their own
comments say the absence is deliberate: position and rotation are written as
GSAP code and there is no CSS fallback to write to. So they can return having
done nothing.
`handleGsapAwareBoxSizeCommit` is not like the other two. It runs through
`runGestureTransaction` with separate scale and width/height routes, so resize
works more generally.
Reading back is what turns that middle case from a silent lie into a reported
one. A move that did nothing comes back in `unchanged` with a reason.
Three smaller decisions:
Operations re-read between each other, so a move is judged against the box
AFTER a resize in the same call. Comparing against the original would credit
the resize's change to the move.
Rotation is reported as dispatched, not verified. `rotate` is an individual
transform property and does not appear in the computed transform, so there is
no honest box-derived signal, and claiming one would be worse than saying so.
x pairs with y and width pairs with height. Accepting one alone would mean
inventing the other from the current value, which moves the element somewhere
the caller did not ask for. The pairing rule and its minimum live in one
`parsePair` helper rather than as four separate branches.
* docs: document Studio's WebMCP agent tools, proven end-to-end in a browser (#3521)
* docs: document Studio's WebMCP agent tools
Adds `guides/webmcp`, under Developers > Agent setup.
Its first job is to defuse a name collision. `guides/mcp` already exists and
covers HeyGen's HOSTED MCP connector, which builds a video from a chat. This
page is about an agent working inside Studio on a composition already open in
front of you. Different feature, confusingly similar name, so the page says
what it is not before it says what it is.
Written to DOCS_GUIDELINES: one-sentence intro, outcome before implementation,
real values rather than placeholders, and three callouts.
The three things a reader most needs are the ones easiest to get wrong:
The API is `document.modelContext`, not `navigator.modelContext`. Most
published examples use the second, which is a polyfill compatibility shim
rather than a spec member, so feature-detecting it misleads.
Select first, then edit. Most editing tools act on the current selection, and
an agent that skips it gets an error rather than a wrong-element write.
Leave Studio visible. Some of Studio's write paths report failure through a
toast rather than a return value, so the human is the one who sees it. That is
a real property of the co-pilot design, not a nicety, so the page says it
plainly.
Verified with `npx mint validate` and `npx mint broken-links --check-redirects`,
both passing.
* fix(studio): target the text field that exists, not one named self
Found by running the tools end to end in a browser, which is the only way it
could have been found: the unit tests mock `setText`, so they never crossed the
boundary where this breaks.
An element's text usually lives in a CHILD field, keyed like `self:0:h1` or
`child:0:h1`. `studio_set_text` passed no field key, so
`buildNextDomTextFields` planned zero operations, the request went out with an
empty patch, and the server answered:
POST /api/projects/<id>/file-mutations/patch-element
-> 400 {"error":"target and operations required"}
Which surfaced as `persist-failed`. The tool was telling the truth, so the
reporting work in the earlier PRs did its job, but the failure looked like a
server problem and was not.
The tool now resolves the field: the one the caller named, or the element's
single field when it has exactly one. An element with several fields is asked
to name one; an element with none is reported blocked. Naming a field the
element does not have is rejected with the list of the ones it does have,
rather than silently writing nowhere.
Four regression tests, including the exact `child:0:h1` shape that failed. One
existing assertion changed: it expected the field to be `undefined`, which is
precisely the bug, so it now expects the resolved key.
Also documents two things the browser run surfaced, both real and neither a
defect: registration is asynchronous, so a caller reading `getTools()` too
early sees a partial list; and the tools that act on the current selection need
a render between the select and the edit, which a real agent gets for free
because its calls arrive as separate messages.
* docs: give the agent-tools kill switch instructions that work
The page told readers to set agentToolsEnabled in Studio's preferences.
Nothing writes that flag: it is read in useStudioAgentTools and parsed in
studioUiPreferences, but there is no settings UI and no toggle, so the
instruction could not be followed. Replace it with the localStorage write
that actually flips it, and spell out the merge, since overwriting the key
drops every other stored preference.
* docs: do not promise a per-call permission prompt we have not verified
The page said the browser asks before any agent calls a tool. Prompt
granularity is browser-specific and unsettled during the origin trial, and
we have not observed it on the native path. Say what holds, that access is
gated, and name the part that is still moving.
* fix(studio): re-apply WebMCP test polyfill fix (#3532 regression)
The squash merge of #3518 re-introduced the old assertion that
document.modelContext is absent. The polyfill from #3514 installs it
as a fallback — that is expected behavior.
Same fix as #3532: remove the assertion, keep the boot-cleanly contract.
---------
Co-authored-by: miga-heygen <miguel.sierra_miga@heygen.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The eye on an expanded sub-composition child always rendered as "Hide",
whatever the source said. One click hid the element and every click after
that rewrote the same attribute, so the row could never be shown again,
not even after a reload, since data-hidden is in the file.
buildChildElements synthesizes a child row from a manifest clip with no
element to read, and compensated by inheriting hidden/timelineLocked/
timelineRole/fxChain/automation from the child's flat store twin. That
twin does not exist for a real sub-composition: processTimelineMessage
drops any clip whose parent composition is itself in the manifest before
building the flat store, so the lookup always missed and the inheritance
was dead code for the one case it was written for. It worked only for a
phantom-wrapper parent, where the child does keep a store entry.
Read the state off the live element instead. collectSubCompositionHostState
walks each sub-composition host in the preview document and records the
data-* state of every id'd descendant, keyed by dom id. The existing
sibling walk cannot serve this: it defines which rows exist and writes
parentMap, and it stops at the first id'd descendant, so scene footage
sitting one level below an id'd region wrapper is never reached. The new
walk descends the whole subtree and touches neither rows nor parentage.
Reproduced on a 9-scene storyboard project where every scene is a
sub-composition. Before: a scene video and title carrying data-hidden both
announced "Hide track N", and clicking left the file byte-identical. After:
both announce "Show track N", and hide/show round-trips the attribute.
A top-level clip with the same attribute always announced "Show", which is
what made the gap specific to expanded child rows.
The existing regression test passed throughout because its fixture hands
the child a flat twin with hidden: true and gives the host no
compositionSrc, so the child key falls back to the index.html scope and a
twin can exist. The added test models a real sub-composition instead.
* feat(studio): let an agent drive Studio's selection and playhead
Adds `studio_select` and `studio_seek`, so an agent and the human are looking
at the same element and the same instant. Selecting reveals the inspector,
exactly as a click does, which is what makes the agent's move visible.
Selection is shared state, not a per-call argument, and that is forced rather
than chosen. Most of Studio's edit handlers read the ambient React selection,
and `applyDomSelection` only schedules a state update, so selecting and
committing inside ONE call would write to whatever was selected before. Two
tool calls are separated by a render, so the contract is select first, then
act. That is also how a human works: click, then type.
`studio_seek` uses `requestSeek`, not `setCurrentTime`. The latter only moves
the timeline's displayed number and leaves the composition where it was.
Two things the tools refuse to fake:
Seek does not clamp. `seek()` already clamps against the adapter's duration,
which can differ from the store's, and clamping again would give that
invariant two owners that can disagree. The tool reports where the playhead
actually landed instead, read back afterwards.
`requestSeek` is fire-and-forget, so it cannot report that no adapter was
mounted to receive it. The tool compares the playhead before and after and
fails rather than claiming a seek that never happened.
Select separates three failures that a single message would have merged: the
preview is not mounted yet (wait), no element matches the handle (re-read),
and the element cannot be selected (try a neighbour). The agent's next move
differs for each, so collapsing them would cost it a round trip or a retry
loop.
* feat(studio): give an agent eyes with studio_frame
Renders the composition to a PNG at a given time and returns the URL. This is
what turns the tool set from a remote control into a loop: author a change,
capture the instant it affects, look, adjust. No agent can judge motion from
source, because "what does this look like at 2.4 seconds" is not a question a
file answers.
Reuses Studio's existing capture endpoint via `buildFrameCaptureUrl` rather
than inventing a second one.
Two things this does not fake:
It reports the time the playhead LANDED on, not the time requested. The player
clamps, so those differ at the ends, and attaching the wrong time to a frame is
how an agent draws a confident wrong conclusion about motion.
It waits before capturing, by default 150ms. The frame is rendered from the
file on disk, and the render cache is cleared by a file watcher with a 40ms
write-stability threshold, so a capture that beats the watcher renders the
PRE-edit composition. That exact staleness was a real bug here once. An agent
reading a stale frame as "my edit failed" would thrash, so the wait is on by
default, `settleMs` makes it tunable, and the tool description names the
failure rather than leaving it to be rediscovered.
It probes with HEAD before returning, so a URL that 404s comes back as a
failure with a hint instead of as a link the agent cannot render.
* feat(studio): add studio_inspect, so an agent reads before it writes
Everything about one element in one call: resolved styles, text fields, box,
data attributes, GSAP animations, and what the element will and will not
accept.
The point is to prevent a failed write rather than to satisfy curiosity.
`can.reasonIfDisabled` is passed through verbatim from Studio's own
capabilities, so an agent that reads first should never attempt an edit the
element would refuse.
Three things it refuses to get wrong:
Animations are reported ONLY for the current selection, because that is the
only element Studio parses them for. Attributing them to any other element
would be reporting the wrong element's motion, which is worse than reporting
none. When a handle names something else the field is empty and
`animationEditingBlocked` says why.
`animationEditingBlocked` also carries the two states where animation editing
is off entirely, multiple timelines and an unsupported timeline pattern. Both
live on the selection context. Learning them from a read costs one call;
learning them from a failed write costs a retry loop.
Inspecting a handle does NOT change what is selected. It is a read, and
stealing the human's selection would be a side effect they did not ask for.
There is a test asserting `applySelection` is never called.
Nothing selected and no handle given is a failure, not an empty result. An
empty result would assert "this element has nothing", which is a different and
false claim.
* feat(studio): let an agent edit text and styles, guarded
The first tools that change the composition. Both act on the current
selection and take no handle, which is forced rather than chosen: the
handlers read the ambient React selection, and `applyDomSelection` only
schedules a state update, so selecting and committing inside one call would
write to whatever was selected before. Select first, then edit.
Also plumbs the write-blocked state, which was the blocker for shipping any
write at all. `domEditSaveQueuePaused` and the external-file conflict both
lived on App and were unreachable from the tool surface, so `canWrite` was
optimistic and a comment said so. They now derive into a single
`writeBlockedReason` on the shell context: one field, one owner, conflict
taking precedence because resolving it is what unblocks the queue.
That guard matters more than it looks. Both states are BANNERS in Studio with
no lock behind them, so nothing else was stopping a programmatic write from
landing on top of a conflict the user had been asked to adjudicate.
Three things the tools refuse to fake:
They check the outcome, not the absence of a throw. Studio has several paths
where a failed commit resolves anyway, so awaiting the handler proves nothing.
The tagged outcome added earlier is what proves the write landed.
A partial style result is reported as partial. `handleDomStyleCommit` is one
property per call, so N properties are N commits; the result carries `applied`
and `rejected` maps rather than a single boolean that would have to pick a
side.
Style commits run sequentially, never concurrently. Two commits racing through
Studio's client-side read-modify-write can record undo entries that both claim
the same starting content. There is a test that measures concurrency rather
than trusting the loop.
Every decline reason maps to a hint naming what to do instead, so a refusal
routes the agent rather than just stopping it.
* feat(studio): add studio_inspect, so an agent reads before it writes (#3517)
Everything about one element in one call: resolved styles, text fields, box,
data attributes, GSAP animations, and what the element will and will not
accept.
The point is to prevent a failed write rather than to satisfy curiosity.
`can.reasonIfDisabled` is passed through verbatim from Studio's own
capabilities, so an agent that reads first should never attempt an edit the
element would refuse.
Three things it refuses to get wrong:
Animations are reported ONLY for the current selection, because that is the
only element Studio parses them for. Attributing them to any other element
would be reporting the wrong element's motion, which is worse than reporting
none. When a handle names something else the field is empty and
`animationEditingBlocked` says why.
`animationEditingBlocked` also carries the two states where animation editing
is off entirely, multiple timelines and an unsupported timeline pattern. Both
live on the selection context. Learning them from a read costs one call;
learning them from a failed write costs a retry loop.
Inspecting a handle does NOT change what is selected. It is a read, and
stealing the human's selection would be a side effect they did not ask for.
There is a test asserting `applySelection` is never called.
Nothing selected and no handle given is a failure, not an empty result. An
empty result would assert "this element has nothing", which is a different and
false claim.
* feat(studio): move, resize and rotate, verified by reading back (#3519)
`studio_transform` does what a drag does, and then checks. The box in the
result is READ BACK after the write, never echoed from the request, and
`applied` lists what actually took effect.
That is not belt-and-braces. The plan for this unit said to re-derive the
geometry handlers' behaviour rather than trust any description of them, and
doing that turned up three different behaviours behind one interface.
The handlers on `DomEditActionsValue` are the GSAP-AWARE wrappers, aliased in
`useDomEditSession.ts:534-538`, not the CSS ones in `useDomGeometryCommits.ts`
that an earlier note in this workstream described.
`handleGsapAwarePathOffsetCommit` and `handleGsapAwareRotationCommit` are
`if (gsapCommitMutation) { ...intercept... }` with no else branch. Their own
comments say the absence is deliberate: position and rotation are written as
GSAP code and there is no CSS fallback to write to. So they can return having
done nothing.
`handleGsapAwareBoxSizeCommit` is not like the other two. It runs through
`runGestureTransaction` with separate scale and width/height routes, so resize
works more generally.
Reading back is what turns that middle case from a silent lie into a reported
one. A move that did nothing comes back in `unchanged` with a reason.
Three smaller decisions:
Operations re-read between each other, so a move is judged against the box
AFTER a resize in the same call. Comparing against the original would credit
the resize's change to the move.
Rotation is reported as dispatched, not verified. `rotate` is an individual
transform property and does not appear in the computed transform, so there is
no honest box-derived signal, and claiming one would be worse than saying so.
x pairs with y and width pairs with height. Accepting one alone would mean
inventing the other from the current value, which moves the element somewhere
the caller did not ask for. The pairing rule and its minimum live in one
`parsePair` helper rather than as four separate branches.
---------
Co-authored-by: miga-heygen <miguel.sierra_miga@heygen.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(studio): let an agent drive Studio's selection and playhead
Adds `studio_select` and `studio_seek`, so an agent and the human are looking
at the same element and the same instant. Selecting reveals the inspector,
exactly as a click does, which is what makes the agent's move visible.
Selection is shared state, not a per-call argument, and that is forced rather
than chosen. Most of Studio's edit handlers read the ambient React selection,
and `applyDomSelection` only schedules a state update, so selecting and
committing inside ONE call would write to whatever was selected before. Two
tool calls are separated by a render, so the contract is select first, then
act. That is also how a human works: click, then type.
`studio_seek` uses `requestSeek`, not `setCurrentTime`. The latter only moves
the timeline's displayed number and leaves the composition where it was.
Two things the tools refuse to fake:
Seek does not clamp. `seek()` already clamps against the adapter's duration,
which can differ from the store's, and clamping again would give that
invariant two owners that can disagree. The tool reports where the playhead
actually landed instead, read back afterwards.
`requestSeek` is fire-and-forget, so it cannot report that no adapter was
mounted to receive it. The tool compares the playhead before and after and
fails rather than claiming a seek that never happened.
Select separates three failures that a single message would have merged: the
preview is not mounted yet (wait), no element matches the handle (re-read),
and the element cannot be selected (try a neighbour). The agent's next move
differs for each, so collapsing them would cost it a round trip or a retry
loop.
* feat(studio): give an agent eyes with studio_frame
Renders the composition to a PNG at a given time and returns the URL. This is
what turns the tool set from a remote control into a loop: author a change,
capture the instant it affects, look, adjust. No agent can judge motion from
source, because "what does this look like at 2.4 seconds" is not a question a
file answers.
Reuses Studio's existing capture endpoint via `buildFrameCaptureUrl` rather
than inventing a second one.
Two things this does not fake:
It reports the time the playhead LANDED on, not the time requested. The player
clamps, so those differ at the ends, and attaching the wrong time to a frame is
how an agent draws a confident wrong conclusion about motion.
It waits before capturing, by default 150ms. The frame is rendered from the
file on disk, and the render cache is cleared by a file watcher with a 40ms
write-stability threshold, so a capture that beats the watcher renders the
PRE-edit composition. That exact staleness was a real bug here once. An agent
reading a stale frame as "my edit failed" would thrash, so the wait is on by
default, `settleMs` makes it tunable, and the tool description names the
failure rather than leaving it to be rediscovered.
It probes with HEAD before returning, so a URL that 404s comes back as a
failure with a hint instead of as a link the agent cannot render.
* feat(studio): add studio_inspect, so an agent reads before it writes (#3517)
Everything about one element in one call: resolved styles, text fields, box,
data attributes, GSAP animations, and what the element will and will not
accept.
The point is to prevent a failed write rather than to satisfy curiosity.
`can.reasonIfDisabled` is passed through verbatim from Studio's own
capabilities, so an agent that reads first should never attempt an edit the
element would refuse.
Three things it refuses to get wrong:
Animations are reported ONLY for the current selection, because that is the
only element Studio parses them for. Attributing them to any other element
would be reporting the wrong element's motion, which is worse than reporting
none. When a handle names something else the field is empty and
`animationEditingBlocked` says why.
`animationEditingBlocked` also carries the two states where animation editing
is off entirely, multiple timelines and an unsupported timeline pattern. Both
live on the selection context. Learning them from a read costs one call;
learning them from a failed write costs a retry loop.
Inspecting a handle does NOT change what is selected. It is a read, and
stealing the human's selection would be a side effect they did not ask for.
There is a test asserting `applySelection` is never called.
Nothing selected and no handle given is a failure, not an empty result. An
empty result would assert "this element has nothing", which is a different and
false claim.
---------
Co-authored-by: miga-heygen <miguel.sierra_miga@heygen.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adds `studio_select` and `studio_seek`, so an agent and the human are looking
at the same element and the same instant. Selecting reveals the inspector,
exactly as a click does, which is what makes the agent's move visible.
Selection is shared state, not a per-call argument, and that is forced rather
than chosen. Most of Studio's edit handlers read the ambient React selection,
and `applyDomSelection` only schedules a state update, so selecting and
committing inside ONE call would write to whatever was selected before. Two
tool calls are separated by a render, so the contract is select first, then
act. That is also how a human works: click, then type.
`studio_seek` uses `requestSeek`, not `setCurrentTime`. The latter only moves
the timeline's displayed number and leaves the composition where it was.
Two things the tools refuse to fake:
Seek does not clamp. `seek()` already clamps against the adapter's duration,
which can differ from the store's, and clamping again would give that
invariant two owners that can disagree. The tool reports where the playhead
actually landed instead, read back afterwards.
`requestSeek` is fire-and-forget, so it cannot report that no adapter was
mounted to receive it. The tool compares the playhead before and after and
fails rather than claiming a seek that never happened.
Select separates three failures that a single message would have merged: the
preview is not mounted yet (wait), no element matches the handle (re-read),
and the element cannot be selected (try a neighbour). The agent's next move
differs for each, so collapsing them would cost it a round trip or a retry
loop.
The header's Export button started renders with no options at all, so the
request carried no `composition` and the server fell back to index.html.
Selecting a sub-composition in the Comps panel showed its canvas and timeline
but exported the root file instead.
Studio starts renders from three controls, and the render target was owned by
each of them separately: the Renders panel resolved it, the header omitted it,
the sidebar's per-composition button named one explicitly. Give it one owner
in `startRender`, which all three route through, defaulting to the active
composition and leaving an explicit argument to win.
Fixes#3549
The Comps panel sets activeCompositionPath to the selected file, but
useCompositionStack's effect only pushed a stack level when that path
started with compositions/. A project laying its comps out anywhere
else, for example a generated multi-part build with parts/part-1.html
next to the root index.html, matched no branch at all: the row
highlighted and the URL hash updated while the stack silently kept the
master mounted, so the canvas and timeline stayed on index.html and any
edit landed in the root file instead of the part.
Replaced the prefix test with a plain truthiness check, so the root
stays on the master level and every other path pushes its own level.
Label derivation is unchanged, matching CompositionsTab's own
convention.
The "registers nothing when the browser has no WebMCP" test asserted
that document.modelContext was absent after mount. Since #3514 added
the @mcp-b/global polyfill fallback, the hook now installs
document.modelContext even when the browser has no native support —
that is the polyfill's job. The real assertion is that mounting does
not throw, which still holds.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat(studio): fall back to a WebMCP polyfill where the browser has none
WebMCP is an Origin Trial. Chrome 149 and Edge 150 have it behind a flag,
ChatGPT Desktop ships it, and everything else does not. Without a fallback the
tools registered in the previous change are invisible on stable Chrome, which
is exactly where a bridge extension would connect from.
Adds `@mcp-b/global` (MIT) as a DYNAMIC import, so a browser with native
support never fetches it. Verified in the build output rather than asserted:
the bundle keeps a bare `import("@mcp-b/global")` instead of inlining it.
Chosen over the smaller `@mcp-b/webmcp-polyfill` because that one only defines
`document.modelContext`. `@mcp-b/global` also stands up the in-page MCP server
a bridge extension attaches to, and serving that case is the only reason the
fallback exists at all.
The load is guarded by a module-level promise so two mounts racing share one
load, and an import failure is caught and logged rather than thrown: a missing
agent surface must never stop Studio booting. The registration path re-checks
the abort signal after the await, so unmounting mid-import registers nothing.
Two things the type checker forced, both worth keeping:
Installing the package brings its own global `Document.modelContext`
declaration, which collided with the local one. Studio now reads the property
through a type guard instead of augmenting `Document`, so there is only one
declaration of that global and it is the package's.
Studio keeps its own narrow tool types rather than importing the package's.
Theirs overload `registerTool` to infer argument types from a literal
`inputSchema`, which helps when registering one tool inline and fights a
uniform registration loop. The comment in `types.ts` says so, and names the
drift risk that choice accepts.
The polyfill test asserts promise identity rather than counting imports. The
ESM registry dedupes the import either way, so a call count would pass whether
or not the guard existed.
* fix(studio): observe and retry WebMCP fallback
* feat(studio): expose Studio's live state to an agentic browser
Registers a `studio_look` tool on `document.modelContext`, so an agent in a
browser that supports it can read what Studio knows: the open project and
composition, the playhead, the human's current selection with its
capabilities, and the timeline's elements with a handle for each.
The API is `document.modelContext`, not `navigator.modelContext`. The latter
is a polyfill compatibility shim rather than a spec member, so feature
detecting it is wrong even where a published sample appears to work.
Three decisions worth knowing:
Registration happens ONCE per mount, with the dependencies held in a ref that
every render refreshes. Depending on the handlers instead re-runs on nearly
every interaction, because the DomEdit actions object changes identity with
the selection and the element list. Each re-run aborts the registration signal
and unregisters everything, and the spec warns that a quick unregister-then-
reregister can apply an old call's arguments against the new schema. The test
for this is the important one in the unit; breaking the empty dependency array
fails it and nothing else.
Tools resolve with a tagged result, they never reject. That is forced by the
spec: a rejected `execute` has its reason discarded and the caller sees a bare
UnknownError, so rejecting would guarantee the agent cannot learn why an edit
failed.
Elements are addressed by a minted handle, not by `TimelineElement.id`. That
id is a synthesised identity, so `getElementById` misses most elements; the
handle carries `data-hf-id`, else the DOM id, else a selector plus occurrence.
Mounted from `EditorShell` rather than `App`, because the DomEdit contexts are
only readable below `DomEditProvider` and `App.tsx` is three lines under the
600-line cap.
The undo signal is reported as the shell actually exposes it, `canUndo` and a
label, rather than as a revision counter. The depth lives in component-local
state and is not reachable without plumbing it through the shell context, so
the field says what it is instead of implying precision it does not have.
Writes are not in this change. `canWrite` is optimistic and the comment says
so; the write tools need a real guard against the paused-save and external-
conflict states, which are not on any context this component can reach yet.
* fix(studio): bound WebMCP look filters
* fix(studio): remove premature WebMCP write state
* docs(studio): name WebMCP singleton assumption
* fix(studio): surface WebMCP registration failures
* fix(studio): let a failed text or style commit report itself
`runDomEditCommit` catches a persist failure, reverts, fires `onError` and
then resolves. That contract is deliberate and its docstring says so: the
human path learns the write failed from the toast `onError` puts on screen,
so a rejection would be redundant. It also means a caller awaiting
`handleDomTextCommit` or `handleDomStyleCommit` cannot tell a landed write
from a reverted one, because both resolve with `undefined`.
The runner already offers `onSettled` as the way out. Text and style were
the two commits that never got it wired.
Add `runReportedDomEditCommit`, which owns `onSettled` (forwarding to a
caller-supplied one rather than dropping it) and returns whether the write
landed. Both handlers now return a tagged outcome, so the three preconditions
that previously returned early and silently are each distinguishable:
no selection, a manual-geometry property the style path refuses, and a
selection that cannot edit styles. Same for text: no selection versus not
text-editable.
Human-facing behaviour is unchanged and the tests assert that: the toast
still fires and the optimistic DOM change is still reverted.
The callback props that carry these handlers ignore the result, so their
declared type widens from `Promise<void>` to `Promise<unknown>`. That type is
hand-copied in fourteen places; consolidating it is worth its own change.
`useDomEditTextCommits.ts` is now 593 lines against the 600-line cap. The
next change to it needs a split.
* fix(studio): stop a paused save queue reporting a position edit as saved
Two more commits that could not tell a caller they had failed.
`useDomEditPositionPatchCommit` swallowed `DomEditSaveQueueOpenError` and
resolved. The intent was right, a paused save queue already puts a banner on
screen and one toast per blocked edit is noise, but swallowing it also
skipped the caller's revert: `useDomGeometryCommits` only restores the
optimistic offset, size or rotation from its `.catch`. So once the breaker
opened, a drag left the element where the user dropped it while nothing
reached the file, and the next reload snapped it back.
It now rejects without toasting. The banner still does the telling; the
caller gets to revert.
`handleDomEditElementsDelete` caught everything and only toasted, so an
unpatchable target and a completed delete were indistinguishable to a caller.
It now returns an outcome, with `no-project` and `no-selection` separated from
a failed write rather than all three sharing an early `return`.
Adds the first test for `useDomEditPositionPatchCommit`, covering the paused
queue, an ordinary failure, and success.
* fix(studio): honor DOM edit failure outcomes
* fix(studio): classify stale delete previews
* fix(studio): enforce DOM edit outcome types
* feat(studio): drag automation segments
* fix(studio): clear clip selection for group effects
* fix(studio): replace clip selection with audio bus
* fix(studio): make audio bus selection authoritative
* fix(core): harden audio FX and group identity
* fix(core): address audio group review feedback
* fix(core): align preview transport with grouped audio
* test(core): pin audio group gain ceiling
* fix(core): preserve solo bridge through stack
* fix(engine): harden grouped audio rendering
* docs(engine): explain grouped mix fallback invariant
* test(engine): allow grouped mixes to finish on Windows
* feat(lint): validate audio group membership and timing
* test(lint): pin audio group membership guards
* fix(studio): unify audio IDs and group state
* fix(studio): make audio-group edits transactional
* fix(studio): keep preview state synchronized
* fix(core): harden audio FX and group identity
* fix(core): address audio group review feedback
* fix(core): align preview transport with grouped audio
* test(core): pin audio group gain ceiling
* fix(core): preserve solo bridge through stack
* fix(engine): harden grouped audio rendering
* docs(engine): explain grouped mix fallback invariant
* test(engine): allow grouped mixes to finish on Windows
* feat(lint): validate audio group membership and timing
* test(lint): pin audio group membership guards
* fix(studio): unify audio IDs and group state
* fix(studio): make audio-group edits transactional
Miguel Ángel