fix: include view data in pipeline responses and skip zero-delta commits

[staging-devin-ai-integration]

Summary

Fixes a regression in the compositor node UI (Monitor view) where stale layout data appeared on the client and was sent back to the server on the next interaction (e.g. click-to-select).

Root cause — four interacting bugs:

A) Missing initial view data: The compositor server emits resolved layout on startup and the engine stores it in node_view_data, but pipeline API responses (both REST and WebSocket) never included this snapshot. New clients fell back to config-parsed positions which could disagree with the server's aspect-fit adjusted layout.

B) Zero-delta click commits stale data: handlePointerUp fired throttledConfigChange unconditionally — even for click-to-select with zero pointer movement. This sent stale config-parsed positions for ALL layers back to the server, overwriting the server's correct resolved layout.

C) "Sync from props" effect overwrites server geometry: When the server echoes back updated params, the "sync from props" effect in useCompositorLayers re-parses all layer positions from params via parseLayers(). For layers with rect: null (auto-layout), this produces {x:0, y:0, width:canvasWidth, height:canvasHeight} — overwriting the server's aspect-fit adjusted positions that useServerLayoutSync had applied. This caused a visible revert ~2 seconds after any drag.

D) preserveGeometry only preserved x/y/w/h — not all server-resolved fields: The initial fix (C) only kept x, y, width, height from existing state, but rebuilt the overlay by spreading { ...parsed, ... }. This lost runtime-only fields (measuredTextWidth, measuredTextHeight) and allowed stale param-derived values for opacity, rotationDegrees, zIndex, and mirror flags to overwrite the server's resolved state — causing text overlays to revert size/position on focus change.

Fix:

Server-side:

• Add Session::get_node_view_data() to expose the engine's stored view data
• Add Pipeline.view_data field (runtime-only, follows Node.state pattern)
• Populate view_data in both REST and WebSocket get_pipeline handlers

Client-side:

• Extract pipeline.view_data into sessionStore.nodeViewData on receipt (both setPipeline and batchSetPipelines)
• Guard handlePointerUp: skip config commit when pointer delta is zero
• Rewrite mergeOverlayState Monitor-view merge: when preserveGeometry is true, start from the full existing overlay (preserving ALL server-resolved OverlayBase fields + runtime measurements like measuredTextWidth), then overlay only type-specific config fields (text, fontSize, fontName, color, dataBase64) from parsed params. Uses pickConfigFields helper + OVERLAY_BASE_KEYS set to cleanly separate server-owned vs config-only fields.
• Regenerated TypeScript types for the new Pipeline.view_data field

Tests:

• Unit tests for mergeOverlayState preserveGeometry behavior (9 tests): OverlayBase field preservation, config field updates, runtime field retention, referential equality, image overlay support, new/removed overlay handling
• Unit tests for setPipeline view_data extraction (initial load, merge, null, batch)
• Unit tests for zero-delta pointer up guard (click does NOT commit, drag DOES commit)
• Integration tests for Monitor view data flow (7 tests): server layout survives params echo-back, text measurements survive echo-back, config changes picked up while preserving server geometry, opacity/rotation/zIndex survive echo-back, focus changes don't reset video or text overlay positions, Design view control test

Review & Testing Checklist for Human

• Verify with a live webcam PiP pipeline in Monitor view: Start a pipeline with a compositor node (e.g. webcam PiP), open the Monitor view in a separate tab/window. Layers should render at the correct server-resolved positions immediately, not jump on first load.
• Click-to-select should not cause layout jumps: In the Monitor view, click on a layer to select it (without dragging). The layer positions should remain stable — no snapping to different coordinates.
• Drag then wait — no revert: Drag a layer to a new position and wait 5+ seconds. The layer should stay in its new position — no revert to the original position.
• Text and image overlays: Position text overlays and image overlays centrally, then wait. They should remain in position (not revert).
• Focus change stability: Select different layers in sequence (text, video, image). No overlay should revert its size or position when losing/gaining focus.
• Multiple clients: Open the Monitor view in two browser tabs. Change the layout in one and verify the other receives the correct positions (not stale data).

Notes

• The pre-existing clippy lint failures in crates/nodes/src/video/compositor/kernel.rs and tests.rs are NOT introduced by this PR — they fail on the video branch without any changes.
• The Pipeline.view_data field uses skip_serializing_if = "Option::is_none" and serde(default) so it's backward-compatible with existing pipeline definitions (YAML configs never include it).
• The preserveGeometry flag only applies in Monitor view (when sessionId is set). Design view (no sessionId) continues to use config-parsed positions as the source of truth, which is the correct behavior for editing.

Link to Devin session: https://staging.itsdev.in/sessions/a5c78299b1204298b83f286a29918a25
Requested by: @streamer45

---

[staging-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring