Fixed issue #1074

[alexlin2]

Fixed issue #1074

[greptile-apps]

Greptile Overview

Greptile Summary

Refactored MoondreamHostedVlModel.point() to query_points() to match the base VlModel interface and return properly typed ImageDetections2D[Detection2DPoint] instead of raw pixel coordinates.

Key changes:

• Renamed method from point() to query_points() to align with base class interface defined in dimos/models/vl/base.py:294-342
• Changed return type from list[tuple[float, float]] to ImageDetections2D[Detection2DPoint]
• Updated implementation to create Detection2DPoint objects with all required fields (x, y, name, ts, image, track_id)
• Added validation via is_valid() check before appending detections
• Added Detection2DPoint import
• Added test coverage for MoondreamHostedVlModel in both bbox and point detection test cases

Confidence Score: 5/5

• This PR is safe to merge with no risk
• The changes are a straightforward refactoring to fix an interface mismatch. The new implementation follows the exact pattern used in the local MoondreamVlModel.query_points() method (lines 181-220), properly converts normalized coordinates to pixels, creates typed detection objects with validation, and adds test coverage. No logic bugs, security issues, or breaking changes detected.
• No files require special attention

Important Files Changed

Filename	Overview
dimos/models/vl/moondream_hosted.py	Refactored point method to query_points to match base class interface and return ImageDetections2D[Detection2DPoint], added Detection2DPoint import

Sequence Diagram

sequenceDiagram
    participant Test as test_vlm.py
    participant Model as MoondreamHostedVlModel
    participant Client as Moondream API Client
    participant Image as Image/Detection Types
    
    Test->>Model: query_points(image, query)
    Model->>Model: _to_pil_image(image)
    Model->>Client: point(pil_image, query)
    Client-->>Model: {"points": [{"x": 0.5, "y": 0.3}, ...]}
    
    loop For each point
        Model->>Model: Convert normalized coords to pixels
        Model->>Image: Create Detection2DPoint(x, y, name, ts, image, track_id)
        Model->>Image: detection.is_valid()
        alt Valid detection
            Model->>Image: image_detections.detections.append(detection)
        end
    end
    
    Model-->>Test: ImageDetections2D[Detection2DPoint]

Loading

[greptile-apps]

_{No files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}