Harden replay repository operations

[xiaoyaoqilan]

Summary

• require TLS 1.2+ when TLS is enabled
• add per-object and per-repository quotas
• expose Prometheus-compatible operational metrics
• support CDN download base URLs
• audit upload operations
• recover interrupted .part uploads safely on service start
• document production deployment and operational settings

Motivation

This is PR 6 of the #3158 implementation stack. It adds the bounded resource use, observability, recovery, and CDN integration required to move the validated MVP toward a reusable service.

Stack dependency

• depends on Select healthy China and US replay nodes #3187, Resolve remote memory2 replays for Go2 #3186, Add indexed memory2 replay uploads #3185, Add S3-compatible replay object storage #3184, and Add WebRTC cross-region and replay repository MVP #3177
• review the commit after f251779 for the isolated change

Validation

• quota, TLS configuration, metrics, CDN URL, audit, and interrupted-upload recovery paths are covered by tests
• full six-PR targeted suite: 56 passed
• ruff check and ruff format --check passed
• targeted mypy checks passed

[codecov]

Codecov Report

❌ Patch coverage is 79.85005% with 430 lines in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
dimos/protocol/pubsub/impl/webrtc/mvp_cli.py	16.94%	195 Missing and 1 partial ⚠️
...s/protocol/pubsub/impl/webrtc/replay_repository.py	80.74%	67 Missing and 42 partials ⚠️
...rotocol/pubsub/impl/webrtc/replay_repository_s3.py	74.64%	23 Missing and 13 partials ⚠️
dimos/protocol/pubsub/impl/webrtc/replay_nodes.py	76.19%	19 Missing and 16 partials ⚠️
dimos/memory2/remote_replay.py	79.46%	15 Missing and 8 partials ⚠️
dimos/memory2/remote_dataset.py	87.36%	6 Missing and 6 partials ⚠️
...otocol/pubsub/impl/webrtc/replay_repository_cli.py	86.36%	6 Missing and 3 partials ⚠️
dimos/protocol/pubsub/impl/webrtc/mvp_metrics.py	88.46%	3 Missing and 3 partials ⚠️
...os/protocol/pubsub/impl/webrtc/providers/broker.py	80.00%	1 Missing and 1 partial ⚠️
.../pubsub/impl/webrtc/test_memory2_remote_dataset.py	98.97%	1 Missing ⚠️
... and 1 more

@@            Coverage Diff             @@
##             main    #3188      +/-   ##
==========================================
- Coverage   73.77%   73.71%   -0.07%     
==========================================
  Files        1109     1125      +16     
  Lines      104383   108053    +3670     
  Branches     9567    10147     +580     
==========================================
+ Hits        77013    79648    +2635     
- Misses      24695    25591     +896     
- Partials     2675     2814     +139     

Flag	Coverage Δ
OS-ubuntu-24.04-arm	68.23% <79.85%> (+0.52%)	⬆️
OS-ubuntu-latest	70.24% <79.85%> (+0.47%)	⬆️
Py-3.10	70.24% <79.85%> (+0.47%)	⬆️
Py-3.11	70.24% <79.85%> (+0.47%)	⬆️
Py-3.12	70.24% <79.85%> (+0.47%)	⬆️
Py-3.13	70.24% <79.85%> (+0.47%)	⬆️
Py-3.14	70.25% <79.85%> (+0.47%)	⬆️
Py-3.14t	?

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

Files with missing lines	Coverage Δ
dimos/core/global_config.py	88.63% <100.00%> (+5.68%)	⬆️
dimos/core/test_global_config.py	100.00% <100.00%> (ø)
dimos/memory2/cli/app.py	84.21% <100.00%> (+14.21%)	⬆️
...otocol/pubsub/impl/webrtc/providers/video_track.py	24.39% <ø> (-2.28%)	⬇️
...os/protocol/pubsub/impl/webrtc/test_mvp_metrics.py	100.00% <100.00%> (ø)
.../protocol/pubsub/impl/webrtc/test_remote_replay.py	100.00% <100.00%> (ø)
...s/protocol/pubsub/impl/webrtc/test_replay_nodes.py	100.00% <100.00%> (ø)
...tocol/pubsub/impl/webrtc/test_replay_repository.py	100.00% <100.00%> (ø)
...l/pubsub/impl/webrtc/test_replay_repository_cli.py	100.00% <100.00%> (ø)
...ol/pubsub/impl/webrtc/test_replay_repository_s3.py	100.00% <100.00%> (ø)
... and 12 more

... and 31 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.

[xiaoyaoqilan]

Public end-to-end validation (2026-07-26)

The six-PR stack is now deployed on an Alibaba Cloud node in Beijing and has been validated from an external Windows client, not only through server-local requests.

• Public health check: GET http://39.106.170.53:8765/healthz returned HTTP 200 with node cn-beijing, region china.
• H.264 object: 640x360, 24 FPS, 2 seconds, 16,763 bytes. External download SHA-256 matched object id 22d3353f196c7ebca06979e03dc68e952d0b92f5162fbe105dc5fdab8655c8c7.
• Official Go2 memory2 dataset: 229,855,232 bytes. External download completed in about 60 seconds and SHA-256 matched object id 32c2971034e449136a9196a2ec50dadd046c04a2783cf56513e2b613f92af0bd.
• SQLite PRAGMA integrity_check: ok.
• Streams discovered from _streams: color_image (855 rows), color_image_embedded (108), lidar (461), and odom (1122).
• Public object pages:
  • http://39.106.170.53:8765/r/xiaoyaoqilan/go2-nav-demo
  • http://39.106.170.53:8765/r/xiaoyaoqilan/go2-short
• The memory2 upload path created a WAL-consistent snapshot plus JSON index; the remote resolver downloaded, content-hash verified, SQLite-validated, and cached the dataset successfully.

CI follow-up was also pushed: project logging now uses setup_logger(), and Typer/Rich CLI error assertions normalize ANSI styling. The targeted replay suite passes (56 passed), Ruff passes, and scoped mypy passes.

The repository/data path is therefore validated end to end. Running the complete dimos --replay run unitree-go2 navigation graph remains a robot/developer-host test because the repository VM is intentionally a small storage service rather than a full Open3D/navigation runtime.

[xiaoyaoqilan]

Superseded by #3201. The implementation is consolidated into the dimos/hosted_data package and unified dimos data CLI.