Skip to content

feat(studio): frame prefetch window with generation-keyed cache #67

Description

@LeadcodeDev

Preview playback still not fluid enough after the release-build fix: frames are rendered strictly on demand (one HTTP request per displayed frame, rendered under the model lock, no cache). Requested: pre-render N frames ahead of the playhead into a bounded cache; on any modification (generation bump) discard and re-prime the window. Design frozen: Arc-snapshot of scenario+tasks so rendering leaves the model lock entirely; cache keyed (generation, side, frame) bounded ~120 entries with distance+generation eviction; background prefetcher keeps [cur..cur+30] warm while playing, [cur-5..cur+30] while paused, nearest-first.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions