Skip to content

Refactor/jevduo env and wording - #16

Merged
xxxyxun merged 2 commits into
AI45Lab:mainfrom
two-tiger:refactor/jevduo-env-and-wording
Nov 6, 2026
Merged

xxxyxun merged 2 commits into
AI45Lab:mainfrom
two-tiger:refactor/jevduo-env-and-wording

Conversation

@two-tiger

Copy link
Copy Markdown
Collaborator

Summary

Follow-up to #15. Renames the TANDEM_* environment variables to JEVDUO_* (with a legacy fallback) and updates the remaining brand wording from Tandem to JevDuo. Two commits:

  1. refactor(scripts): rename TANDEM_* env vars to JEVDUO_* with a legacy fallback
  2. docs: update brand wording from Tandem to JevDuo

Changes

Env vars (the only behavior change)

  • scripts/run_experiment.sh reads JEVDUO_EXPERIMENT_ROOT / JEVDUO_CONFIG_FILE.
  • The old TANDEM_EXPERIMENT_ROOT / TANDEM_CONFIG_FILE still work but print a deprecation warning on stderr, so a stale variable never silently falls back to the default directory. The new name wins if both are set.
  • No Python code reads these variables; the script and its tests are the only users.

Wording (display text only)

  • README / AGENTS.md / docs/CLUSTER_RUNTIME.md: title and intro.
  • demo: page title, header, labels, console tag, build_site.py description, two HTTP User-Agent strings.
  • tools/render_summary.py: page title.
  • DEVIATIONS.md: a naming note at the top; historical entries kept verbatim.

Not in this PR

  • harness.name, the episode.json harness string, the version number, and the schema $id / $ref / title (next PR).
  • Shared-volume paths owned elsewhere (tandem_api_pool, tandem_jev_images, tandem_osworld21_*), historical data under demo/data/, and the real-machine fixture in tests/eval/rjob_fakes.py.

Verification

  • ruff check src tests demo: pass. jevduo check: 17 samples pass.
  • pytest: 659 passed, 1 skipped (the existing optional-Chromium test). Two new tests cover legacy-only and both-set env vars.
  • node --check on the demo JS files: pass.
  • Manual run_experiment.sh --dry-run: new var only → no warning; legacy var only → warning and the legacy path is used; both set → the new path is used and the legacy path is not created.

Smoke: not run on the cluster. A real run_experiment.sh run always submits --benchmark osworld21, so a mock rjob smoke would not exercise the changed code. The changed path was verified with --dry-run. The second commit is display text only.

tandem-demo added 2 commits November 6, 2026 10:19
… fallback

- scripts/run_experiment.sh reads JEVDUO_EXPERIMENT_ROOT / JEVDUO_CONFIG_FILE.
  The pre-rename TANDEM_* names are still honored but print a deprecation warning
  on stderr, so a stale variable never silently falls back to the default directory.
  The new name wins when both are set.
- tests: switch the existing cases to JEVDUO_*, add legacy-only and both-set cases.
- README upgrade note and docs/CLUSTER_RUNTIME.md document the new names and the transition.
- README / AGENTS.md / docs/CLUSTER_RUNTIME.md: title and intro wording
- demo: page title, header, eyebrow, player and result-table labels, console tag,
  build_site.py description, and the two HTTP User-Agent strings
- tools/render_summary.py: page title
- DEVIATIONS.md: add a naming note at the top; historical entries are kept verbatim

Display text only, no behavior change. Still unchanged (follow-up PR): harness.name,
the episode.json harness string, the version number and the schema $id/$ref/title.
@xxxyxun
xxxyxun self-requested a review November 6, 2026 02:58
@xxxyxun
xxxyxun merged commit 6801092 into AI45Lab:main Nov 6, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants