Feat/rerun latency panels

[Nabla7]

Rerun Visualization Improvements

Addresses feedback from issue #922

Visual Fixes:

• Costmap now uses Foxglove-style colors (blue-purple free, black occupied)
• Fixed z-fighting white dots by increasing z_offset to 0.05
• Removed broken 2D image panel logging
• Made path visualization thicker (5cm) and elevated above floor

New Features:

• TFRerunModule: Auto-visualizes all TF transforms by subscribing to /tf LCM topic
• log_timing_to_rerun() decorator for easy performance metrics

Fixes:

• Command center routing now uses global_config instead of os.environ
• A* planner logs paths manually without polluting stream.py core code

Result: Clean Rerun integration without adding complexity to core stream infrastructure. (cleaned up all vibeslop from previous experiments my bad)

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[greptile-apps]

Greptile Summary

This PR cleans up Rerun visualization by removing experimental autolog_to_rerun() infrastructure from stream.py and implementing cleaner, module-specific logging patterns. The changes address issue #922 feedback comprehensively.

Major improvements:

• Visual fixes: Costmap now uses Foxglove-style colors (blue-purple #484981 for free space, black for occupied), z-offset increased from 0.02m to 0.05m to eliminate z-fighting, paths made thicker (5cm) and elevated (0.2m) for better visibility
• New TFRerunModule: Auto-visualizes all TF transforms by subscribing to /tf LCM topic, properly conditional on viewer_backend setting
• Performance metrics: New @log_timing_to_rerun() decorator enables easy function timing without queues
• Configuration cleanup: Command center and modules now use global_config.viewer_backend instead of os.environ
• Camera fix: Unitree camera correctly marked as RGB format (was implicitly BGR before)

Architecture changes:

• Removed generic autolog_to_rerun() from Out class (was coupling visualization to core streaming infrastructure)
• A* planner now manually subscribes for Rerun logging (cleaner than autolog pattern)
• Removed broken 2D costmap image logging (only keeping working 3D mesh)

The PR successfully achieves its goal of "clean Rerun integration without adding complexity to core stream infrastructure."

Confidence Score: 5/5

• Safe to merge - well-tested refactor with clear improvements and no breaking changes
• All changes are architectural improvements with proper cleanup. The removal of autolog_to_rerun() is compensated by explicit patterns. Visual fixes directly address reported issues. New modules properly check configuration before initializing. No logic errors or security concerns identified.
• No files require special attention

Important Files Changed

Filename	Overview
dimos/core/stream.py	Removed autolog_to_rerun() infrastructure - clean removal of experimental feature without side effects
dimos/dashboard/tf_rerun_module.py	New TF visualization module - properly checks viewer_backend and cleans up LCM subscription on stop
dimos/utils/metrics.py	Added log_timing_to_rerun() decorator for performance metrics - simple and well-documented
dimos/mapping/costmapper.py	Removed broken 2D image logging, updated z_offset to 0.05 to fix z-fighting
dimos/msgs/nav_msgs/OccupancyGrid.py	Implemented Foxglove-style colormap (blue-purple for free, black for occupied) with smooth gradient
dimos/navigation/replanning_a_star/module.py	Replaced autolog with manual subscribe pattern, cleaned up Disposable usage

Sequence Diagram

sequenceDiagram
    participant BP as Blueprint
    participant TFM as TFRerunModule
    participant LCM as LCM /tf Topic
    participant TF as TFMessage
    participant RR as Rerun

    BP->>TFM: start()
    TFM->>TFM: Check viewer_backend
    alt viewer_backend starts with "rerun"
        TFM->>RR: connect_rerun()
        TFM->>LCM: subscribe("/tf", callback)
        loop On each TF message
            LCM->>TFM: _on_tf_message(TFMessage)
            TFM->>TF: msg.to_rerun()
            TF-->>TFM: [(entity_path, Transform3D), ...]
            loop For each transform
                TFM->>RR: rr.log(entity_path, Transform3D)
            end
        end
    end

    Note over BP,RR: Manual Rerun Logging (A* Planner)
    participant AP as ReplanningAStarPlanner
    participant GP as GlobalPlanner
    
    BP->>AP: start()
    AP->>AP: Check viewer_backend
    alt viewer_backend starts with "rerun"
        AP->>RR: connect_rerun()
        AP->>GP: path.subscribe(_log_path_to_rerun)
        loop On path updates
            GP->>AP: path published
            AP->>RR: rr.log("world/nav/path", path.to_rerun())
        end
    end

Loading

[greptile-apps]

Greptile's behavior is changing!

From now on, if a review finishes with no comments, we will not post an additional "statistics" comment to confirm that our review found nothing to comment on. However, you can confirm that we reviewed your changes in the status check section.

_{This feature can be toggled off in your Code Review Settings by deselecting "Create a status check for each PR".}