Model Landscape

Databricks-native model observability. Two halves of one product:

• mlflow-lens — a small SDK that enriches MLflow runs with interactive Plotly panels (classification, regression, model selection), training-time drift, structured summaries, and workspace context.
• The warehouse app — a Databricks App that monitors any Unity Catalog inference table for drift, performance degradation, data-quality regressions, and incidents. Data never leaves the workspace.

Docs

The full documentation site is the canonical source — install guide, architecture, SDK reference, panel gallery, and the intro deck all live there:

→ https://databricks-solutions.github.io/model-landscape/

Run it locally:

uv pip install mkdocs-material 'mkdocstrings[python]>=0.26' \
               mkdocs-include-markdown-plugin
uv run python docs/_build_gallery.py
uv run mkdocs serve

Quick deploy

git clone https://github.com/databricks-solutions/model-landscape.git
cd model-landscape
uv build --wheel --out-dir dist

databricks bundle deploy -t warehouse_only \
  --var "sql_warehouse_id=<id>" \
  --var "refresh_node_type_id=<node-type>" \
  --var "control_plane_catalog=<catalog>" \
  --var "control_plane_schema=model_landscape_control_plane"

databricks apps start model-landscape
databricks apps deploy model-landscape \
  --source-code-path /Workspace/Users/<email>/.bundle/model-landscape/warehouse_only/files
databricks bundle run tutorial_mlops

Full first-time-setup guide: Get started.

Local development

uv sync --extra dev               # install workspace + dev deps
uv run pytest -q                  # run tests
uv build --wheel --out-dir dist   # build SDK + app wheel
uv run python -m model_landscape.app  # run app locally

License

See LICENSE.md. Third-party notices in NOTICE.txt.