rapiddweller/datamimic/setup-datamimic

Set up DATAMIMIC in the project

Prepare the selected local project for CE authoring, then continue the user's data task. A connected Platform project uses its existing remote service and does not need a local CE installation.

Select the environment

Read the project's instructions, Makefile, dependency declarations and lockfile. Use its configured Python interpreter and dependency manager. Resolve absolute project and interpreter paths; a global CLI on PATH is not evidence that it belongs to this project. Preserve the existing environment and dependencies.

If no environment is configured, find Python 3.11+ and create .venv in the project with that interpreter's -m venv .venv. Keep model resources outside root .venv/venv directories, which bounded snapshots exclude. Use the environment's binaries explicitly, including Scripts on Windows. If Python is unavailable, report that prerequisite; do not install a system interpreter.

Install only when needed

Try the selected environment's datamimic authoring version and datamimic authoring reference --topic assembly --name starter. A compatible CE 5 runtime must expose Build protocol 5.0: a starter with entry_xml and an acceptance carrying expectations. If that contract is already available, setup is complete: record the resolved paths/version and continue the data task. Do not invoke pip or another dependency manager, even to request an already satisfied package.

Only if CE is absent or incompatible, install the published CE 5 dependency through the project's dependency manager and record it in the existing dependency/lock files. For an unmanaged virtual environment, use its Python directly:

python -m pip install "datamimic-ce>=5.0.0,<6"

Here python means the absolute selected environment interpreter. Complete this setup as part of the requested task; do not send the user a command to run when you can execute it. Respect host permissions. An existing incompatible explicit pin needs a reported dependency conflict, not silently removing the pin or replacing the environment. CLI authoring needs no MCP extra.

If the published version cannot be resolved, installation fails, or the live contract remains incompatible, preserve the project and report the exact stage and error. Do not substitute an old API, source checkout or local wheel. An access/configuration failure is not an authoring result.

Verify and continue

Run both discovery commands with the installed project's absolute CLI. Check the actual returned contract; installation exit status alone is insufficient. Record the project root, interpreter, CLI path, CE version and any dependency changes. A second invocation should reuse the verified environment unchanged. Return to the original data task using CE authoring.