UI

[3a1b2c3]

No description provided.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Greptile Summary

The PR adds a LingBot browser UI and example-scene catalog while extending live prompt handling and WebRTC integration.

• Adds prompt-bar and model-specific browser controls.
• Introduces dynamic prompt propagation and conditioning-state tracking.
• Adds LingBot setup scripts, documentation, launch helpers, and example assets.
• Updates shared WebRTC serving and image-encoding integration points.

Confidence Score: 3/5

The PR is not yet safe to merge because prompt submissions remain excluded from canonicalization and clearing a prompt leaves stale model conditioning active.

The current source capability still cannot feed the changed text-event converter, while the clear path continues selecting the active prompt instead of restoring the base prompt; these previously reported failures remain reachable in the browser prompt flow.

Files Needing Attention: integrations/lingbot/lingbot/input_mapping.py, integrations/lingbot/lingbot/webrtc/session.py

Important Files Changed

Filename	Overview
integrations/lingbot/lingbot/input_mapping.py	Extends text-event payloads with dynamic prompts and adds persistent active-prompt state across inference steps.
integrations/lingbot/lingbot/webrtc/session.py	Updates the live LingBot session to apply prompt-conditioning changes during chunk generation.
integrations/lingbot/lingbot/webrtc/web/adapter.js	Adds the model-specific browser UI, prompt submission, and scene-control behavior.
flashdreams/flashdreams/serving/webrtc/manager.py	Adjusts shared WebRTC session and event handling used by the new LingBot interface.
flashdreams/flashdreams/infra/encoder/image/clip.py	Updates image-encoder integration used by the expanded interactive workflow.

Sequence Diagram

sequenceDiagram
    participant B as Browser UI
    participant W as WebRTC input source
    participant C as Canonicalizer
    participant M as LingBot input mapping
    participant S as Model session
    B->>W: Camera or text event
    W->>C: Timed user-input window
    C->>M: Canonical controls
    M->>S: Camera trajectory and prompt conditioning
    S-->>B: Generated video chunk

Loading

_{Reviews (15): Last reviewed commit: "run" | Re-trigger Greptile}

[greptile-apps]

Clear retains stale conditioning

When a user activates a catalog or dynamic prompt and then clears it, this branch prefers _active_prompt, which still contains the selected prompt, instead of restoring _base_prompt. Subsequent chunks therefore continue using the scene conditioning the user attempted to clear.

Knowledge Base Used: Interactive world-model integrations

[greptile-apps]

Want your agent to iterate on Greptile's feedback? Try greploops.