Create a new workflow to run self_hosted_large tests

[Dreamsorcerer]

No description provided.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!

[greptile-apps]

Greptile Summary

This PR replaces the dimsim pytest marker with self_hosted_large and introduces a dedicated self-hosted-large-tests CI job that runs on a GPU-capable [self-hosted, Linux, large] runner, finally wiring up the three dimsim e2e tests that were previously marked but never executed in CI.

• The dimsim_process.py render-mode logic is refactored so that an explicit DIMSIM_RENDER env var takes priority over the CI auto-detection, allowing the new job to opt into GPU rendering while keeping other CI jobs on CPU.
• Explicit timeouts (120 s / 180 s) are added to wait_until_odom_position calls that previously relied on the 60-second default — appropriate given the heavier scenes.
• The self-hosted-tests matrix job filter does not exclude self_hosted_large, which is fine today but could silently run large tests on the base runner if both markers are ever combined on one test.

Confidence Score: 4/5

Safe to merge; the new CI job is well-guarded against fork abuse and the marker migration is complete across all files.

The self-hosted-tests matrix job does not exclude self_hosted_large from its marker filter, which is harmless today but leaves a small gap if both markers are ever combined on a test — it would run on the base runner without the required GPU and likely fail silently on that runner. Everything else — the render-mode refactor, timeout additions, and codecov wiring — is straightforward and correct.

.github/workflows/ci.yml — the self-hosted-tests run step could benefit from adding self_hosted_large to its exclusion filter for future safety.

Important Files Changed

Filename	Overview
.github/workflows/ci.yml	New self-hosted-large-tests job added for GPU-capable runner; dimsim marker exclusion replaced with self_hosted_large in regular test job; actions/checkout bumped to v6. The self-hosted-tests job filter does not explicitly exclude self_hosted_large, which could cause duplicate runs if both markers appear on one test.
dimos/simulation/dimsim/dimsim_process.py	Render mode selection refactored to respect DIMSIM_RENDER env var when explicitly set, falling back to CPU in CI and GPU otherwise — correctly enables the new large runner to opt into GPU rendering.
dimos/conftest.py	dimsim marker replaced with self_hosted_large; description updated to reflect high-memory runner requirement.
dimos/e2e_tests/test_dimsim_walk_forward.py	Marker updated to self_hosted_large; explicit timeout=120 added to wait_until_odom_position to prevent silent hangs under the default 60-second timeout.
dimos/e2e_tests/test_dimsim_spatial_memory.py	Marker updated to self_hosted_large; explicit timeout=180 added to wait_until_odom_position for the longer spatial-memory scenario.
pyproject.toml	Default addopts exclusion filter updated from dimsim to self_hosted_large, keeping local dev runs free of GPU-heavy simulation tests.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[CI Triggered] --> B{Event type}
    B -->|push / same-repo PR| C[self-hosted-large-tests job starts]
    B -->|fork PR| D[self-hosted-large-tests SKIPPED]
    C --> E[Checkout + Fix permissions]
    E --> F[Install uv + apt deps]
    F --> G[uv sync --group tests-self-hosted]
    G --> H[Build C++ extensions]
    H --> I{DIMSIM_RENDER set?}
    I -->|Yes - gpu from job env| J[render = gpu]
    I -->|No| K{CI env set?}
    K -->|Yes| L[render = cpu]
    K -->|No| M[render = gpu]
    J --> N[Run pytest -m self_hosted_large with DISPLAY=:0]
    N -->|pass| O[Upload coverage to Codecov flag: SelfHosted-Large]
    N -->|fail| P[Re-run --lf with -vvvvv]
    P --> O
    O --> Q[ci-complete watcher allows skip of self-hosted-large-tests]

Loading

Comments Outside Diff (1)

1. .github/workflows/ci.yml, line 298-305 (link)

   The self-hosted-tests filter doesn't exclude self_hosted_large, so any future test decorated with both self_hosted and self_hosted_large would run on the base runner (which lacks the required GPU/memory) and on the large runner — duplicating the run and likely failing on the smaller machine. Adding an explicit exclusion here is a small, cheap guard against that.

_{Reviews (1): Last reviewed commit: "Merge branch 'main' into sam/self-hosted..." | Re-trigger Greptile}