google / google/artemis

start.sh: MCP installation fails with older uv due to directory shadowing

Open Beginner friendly
#106 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
5.9k
Forks
516
Avg merge
22m
Merged PRs (30d)
5

Description

### Description

When running `./start.sh` from the repository root and choosing to install MCP configuration, installation fails with:

```text
ImportError: cannot import name 'RunTree' from 'langsmith'
(/.venv/lib/python3.14/site-packages/langsmith/__init__.py)
```

The Showcase UI build completes successfully before this failure.

### Environment

- macOS
- Python 3.14.5
- uv 0.5.9
- Repository commit: 371aa6d

### Reproduction

From the repository root:

```bash
uv run artemis mcp --help
```

This reproduces the same import error without modifying IDE configuration.

Verbose uv output shows:

```text
DEBUG Running `python artemis mcp --help`
```

### Cause

In this environment, uv interprets `artemis` as the local directory instead of invoking the installed console script. This places the package directory on the module search path, allowing `artemis/platform` to shadow Python's standard-library `platform` module.

### Expected behavior

The MCP command should run successfully from the repository root.

### Verified workaround

```bash
uv run python -m artemis mcp --help
uv run python -m artemis mcp --install all
```

Both commands succeed with the same Python environment and dependencies.

### Proposed fix

Use `uv run python -m artemis` for MCP installation and the retry instructions in `start.sh`, consistent with the script's existing UI launch command.

Contributor guide

Open the contributing guide

Research direction

Open start.sh and compare the existing UI launch command with the MCP installation and retry commands. Reproduce the failure from the repository root with uv run artemis mcp --help, then verify the MCP help and installation commands succeed with the Python module entry point under the reported older uv environment.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
cli
Issue type
Bug
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Active
Clarity
Clearly specified
Newbie friendliness
88/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.