fix(cua-bench-rl): documented PyPI installation fails
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 73/100
- Issue type
- Bug
- Clarity
- Clearly specified
- Activity status
- Active
- Tech stack
- python
- Domain
- build-system, documentation
Research direction
Read the install block in libs/cua-bench-rl/README.md and the compatibility hint in libs/cua-bench/cua_bench/compat/__init__.py; the issue also links the reinforcement-learning guide. Check the documented base install against the available source-install route and confirm the worker-server import succeeds. Done means the README and compatibility hint direct users to a working installation method.
Written by the indexing model from the issue text.
Description
Primary area
Documentation
Summary
The RL package README tells users to run pip install cua-bench-rl, but PyPI has no project by that name. A clean install fails before the documented worker command can run. The released Bench compatibility error also directs users to the same unavailable package.
Reproduction
- Use a clean Python 3.12 virtual environment with pip and the public PyPI index.
- Follow the base install command in the RL README:
python -m pip --isolated install --no-cache-dir --index-url https://pypi.org/simple cua-bench-rl
curl -i https://pypi.org/pypi/cua-bench-rl/json
- Optionally, in a separate environment with the published Bench package installed, run
python -c 'import cua_bench.workers'and follow its install hint. Both commands were executed; the optional import produced the error below.
Expected behavior
The documented install should obtain the package needed by python -m cua_bench_rl.workers.worker_server --port 8001, or identify an available source-install route. The compatibility error should point to an install method that works.
Actual behavior
The install exits 1:
ERROR: Could not find a version that satisfies the requirement cua-bench-rl (from versions: none)
ERROR: No matching distribution found for cua-bench-rl
The PyPI JSON endpoint returns HTTP 404 with {"message": "Not Found"}. Importing the moved workers module in the published Bench environment exits 1 with:
ModuleNotFoundError: cua_bench.workers moved to cua_bench_rl.workers: pip install cua-bench-rl
Environment
- Documentation and source checked at default-branch commit
5ee5974492839bb3b5a3db8919e3cf517e23ffe3. - PyPI
cua-bench0.4.9, resolvingcua-sandbox0.10.2 andcua0.25.2 in the separate compatibility-import environment. - Linux x86_64, CPython 3.12.15, pip 26.3; clean virtual environments and the public PyPI index. No VM, container, worker server, or model was started.
- There is no published PyPI release of
cua-bench-rlto test; its source metadata declares 0.1.0.
Evidence
- Expected source: RL README install block.
- Failure source: the executed pip output and HTTP response above; the released compatibility import emitted the same install hint as the current compatibility source.
- Evidence provenance: observed
- Local verification: reproduced
- Reproduction completeness: complete
The Spaces reinforcement-learning guide describes this as the package in the repository. The report concerns the README and emitted install hint; it does not assume that publishing a new distribution is the intended solution.
Restoration check
Starting with a clean supported Python environment, the documented base installation method should finish successfully and python -c 'import cua_bench_rl.workers.worker_server' should succeed. The compatibility error should direct users to that available method. A local source install with registry dependencies passed that import check; no worker was started and RL execution was not tested.
Workaround and additional context
From a repository checkout, python -m pip install ./libs/cua-bench-rl is a source-install route. The equivalent isolated installation using uv pip install --no-sources ./libs/cua-bench-rl succeeded with registry dependencies and passed the worker-server import smoke check. This does not verify the optional torch or Tinker extras.
Triage assessment
- Impact level: P2
- Assessment status: supported
- Impact basis: Following the documented base install fails for the RL workers package, and the released compatibility error points users to the same unavailable distribution. Ordinary benchmark execution is outside this defect.
- Workaround status: available
- Workaround basis: Installing the base package from a checkout with registry dependencies succeeded and its worker-server module imported successfully; optional training extras were not tested.
Submission checks
- I searched for duplicate issues and active pull requests.
- This issue describes a single problem.
- I removed credentials, private data, sensitive screenshots, and vulnerability details from this public report.
- Dominant language
- Rust
- Stars
- 32.3k
- Forks
- 2.4k
- Avg merge
- 4h 47m
- Merged PRs (30d)
- 711
Getting set up
- No Dockerfile or Docker Compose file
- Has a pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from trycua/cua
-
Difficulty 1/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 86/100
Maintainers usually reply within 1 day
Similar issues
-
app-request
Difficulty 2/5 1-3 hours Newbie friendliness 65/100
flatpark/flatpark#1423 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
helix-editor/termina#35 ·
-
Difficulty 1/5 Under an hour Newbie friendliness 75/100
-
Difficulty 2/5 1-3 hours Newbie friendliness 77/100
Maintainers usually reply within 5 days
-
Add coverage reportsOpentype:ci
Difficulty 2/5 1-3 hours Newbie friendliness 68/100