Skip to content

fix(tbench2_env): bring Docker mode up to the canonical scoring contract #1012

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
burtenshaw merged 3 commits into from
Jul 29, 2026
Merged

fix(tbench2_env): bring Docker mode up to the canonical scoring contract #1012

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
9c2fd89
fix(tbench2_env): bring Docker mode up to the canonical scoring contract
nblintao Jul 24, 2026
7bef0ba
docs: sync tbench2 env docs after README update
nblintao Jul 24, 2026
4180b7c
fix: make docker reset fail closed
burtenshaw Jul 29, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion docs/source/environments/tbench2.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,7 +96,7 @@ TB2_MODE=local python -m tbench2_env.server.app

### Docker Mode

Each task runs in its own Docker container, using the image specified in the task's `task.toml`:
Each task runs in its own Docker container, using the image specified in the task's `task.toml` (tasks that declare no `docker_image` are rejected at reset — there is no host-execution fallback). Agent commands run in the image's own `WORKDIR`, and `evaluate` scores with the task's canonical `tests/test.sh`, staged for exactly the verify window — the same scoring contract as local mode:

```bash
# Enable Docker mode
Expand Down
2 changes: 1 addition & 1 deletion envs/tbench2_env/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -110,7 +110,7 @@ TB2_MODE=local python -m tbench2_env.server.app

### Docker Mode

Each task runs in its own Docker container, using the image specified in the task's `task.toml`:
Each task runs in its own Docker container, using the image specified in the task's `task.toml` (tasks that declare no `docker_image` are rejected at reset — there is no host-execution fallback). Agent commands run in the image's own `WORKDIR`, and `evaluate` scores with the task's canonical `tests/test.sh`, staged for exactly the verify window — the same scoring contract as local mode:

```bash
# Enable Docker mode
Expand Down
14 changes: 8 additions & 6 deletions envs/tbench2_env/server/app.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,7 +28,7 @@
python -m server.app

Environment Variables:
TB2_MODE: Execution mode - "local" (default), "docker", or "auto"
TB2_MODE: Execution mode - "local" (default) or "docker"
MAX_CONCURRENT_ENVS: Maximum concurrent WebSocket sessions (default: 8)
"""

Expand Down Expand Up @@ -61,13 +61,15 @@
if _TB2_MODE == "docker":
_DEFAULT_ENVIRONMENT = Tbench2DockerEnvironment
_ENV_SUFFIX = " (Docker mode)"
elif _TB2_MODE == "auto":
# Auto-detect: try Docker, fall back to local
_DEFAULT_ENVIRONMENT = Tbench2Environment
_ENV_SUFFIX = " (auto-detect mode)"
else:
elif _TB2_MODE == "local":
_DEFAULT_ENVIRONMENT = Tbench2Environment
_ENV_SUFFIX = " (local mode)"
else:
# No silent aliases: the old "auto" value claimed to auto-detect Docker
# but always ran local mode, and a typo'd mode would quietly do the same.
# Which class serves decides the isolation and scoring contract, so
# refuse to guess.
raise ValueError(f"Unknown TB2_MODE {_TB2_MODE!r}: expected 'local' or 'docker'.")


# Create the app with web interface and README integration
Expand Down
Loading
Loading