feat(evals): add local VQA dataset editor

[ruthwikdasyam]

Contribution path

• Linked PR stack: feat(evals): support point-cloud VQA families #3492

Problem

VQA datasets need a local workflow for frame-level generation, review, and correction.

Solution

Add a browser-based dataset editor with atomic submission, generation feedback, and editor-startup model preloading.

How to Test

uv run dimos evals vqa edit /path/to/recording.db /path/to/vqa-dataset

AI assistance

OpenCode with GPT-5.6-sol provided substantial implementation and validation assistance.

Checklist

• I have read and approved the CLA.

[codecov]

Codecov Report

❌ Patch coverage is 87.21400% with 95 lines in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
dimos/evals/vqa/editor.py	82.96%	34 Missing and 28 partials ⚠️
dimos/evals/vqa/visualizer.py	72.94%	18 Missing and 5 partials ⚠️
dimos/evals/vqa/preprocessing.py	28.57%	5 Missing ⚠️
dimos/evals/vqa/test_visualizer.py	93.65%	3 Missing and 1 partial ⚠️
dimos/evals/vqa/test_editor.py	99.44%	1 Missing ⚠️

@@                       Coverage Diff                       @@
##           ruthwik/feat/vqa-pointcloud    #3605      +/-   ##
===============================================================
+ Coverage                        76.61%   76.67%   +0.05%     
===============================================================
  Files                             1260     1265       +5     
  Lines                           122285   123061     +776     
  Branches                         10874    10949      +75     
===============================================================
+ Hits                             93691    94355     +664     
- Misses                           25444    25521      +77     
- Partials                          3150     3185      +35     

Flag	Coverage Δ
OS-ubuntu-24.04-arm	71.30% <87.21%> (+0.09%)	⬆️
OS-ubuntu-latest	73.01% <87.21%> (+0.08%)	⬆️
Py-3.10	73.01% <87.21%> (+0.07%)	⬆️
Py-3.11	73.01% <87.21%> (+0.07%)	⬆️
Py-3.12	73.01% <87.21%> (+0.07%)	⬆️
Py-3.13	73.01% <87.21%> (+0.08%)	⬆️
Py-3.14	73.01% <87.21%> (+0.07%)	⬆️
Py-3.14t	73.01% <87.21%> (+0.07%)	⬆️
SelfHosted-Large	29.76% <21.93%> (-0.05%)	⬇️
SelfHosted-Linux	35.59% <21.93%> (-0.10%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
dimos/cli/test_vqa.py	100.00% <100.00%> (ø)
dimos/cli/vqa.py	96.66% <100.00%> (+0.66%)	⬆️
dimos/evals/vqa/author.py	91.48% <100.00%> (+3.98%)	⬆️
dimos/evals/vqa/generate.py	81.31% <100.00%> (ø)
dimos/evals/vqa/test_contracts.py	98.92% <100.00%> (+<0.01%)	⬆️
dimos/evals/vqa/test_preprocessing.py	100.00% <100.00%> (ø)
dimos/evals/vqa/test_editor.py	99.44% <99.44%> (ø)
dimos/evals/vqa/test_visualizer.py	93.65% <93.65%> (ø)
dimos/evals/vqa/preprocessing.py	60.33% <28.57%> (+1.81%)	⬆️
dimos/evals/vqa/visualizer.py	72.94% <72.94%> (ø)
... and 1 more

... and 5 files with indirect coverage changes

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.