philler 0.1.0

Install the mcp extra and start the server:

pip install "philler[mcp]"
phil-mcp
# or ephemeral: uv tool run --from "philler[mcp]" phil-mcp

Example Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "phil": {
      "command": "uv tool run",
      "args": ["--from", "philler[mcp]", "phil-mcp"]
    }
  }
}

Key tools: ingest_dataset, characterize_dataset, profile_predictability, generate_research_template, recommend_grid, list_grids, create_config, validate_config, run_imputation_sweep, diagnose_sweep, list_collapse_strategies, inspect_ect_distribution, collapse_ensemble, export_imputed_data.

Agents can read phil://docs/imputation-matrix for grid comparison metadata, use profile_predictability + generate_research_template to separate empirically predictable columns from domain/literature priors (DomainKnowledge), and compare ECT signature collapse strategies (barycenter, trimmed_barycenter, frechet_medoid, euclidean_medoid, mode_medoid, cluster_barycenter, auto). Polars users write to Parquet and ingest the file path.

See docs/source/userGuides/mcp.rst for the full tool table and example dialog. Local end-to-end testing: demos/medical.

Imputation grids

Named grids via GridGallery:

Custom grid:

from phil import Phil, ImputationConfig
from sklearn.model_selection import ParameterGrid

config = ImputationConfig(
    methods=["KNNImputer"],
    modules=["sklearn.impute"],
    grids=[ParameterGrid({"n_neighbors": [3, 5, 7]})],
)
phil = Phil(param_grid=config)

ECT descriptor

from phil import Phil, ECTConfig

phil = Phil(config=ECTConfig(num_thetas=64, radius=1.0, resolution=100, scale=500, normalize=True, seed=42))

from sklearn.pipeline import Pipeline
from sklearn.ensemble import RandomForestClassifier
from phil import PhilTransformer

pipe = Pipeline([
    ("imputer", PhilTransformer(samples=20, random_state=0)),
    ("model", RandomForestClassifier()),
])
pipe.fit(X_train, y_train)

• FastMCP server — phil-mcp exposes the imputation sweep pipeline as MCP tools for agents.
• Grid recommender — recommend_grid, declarative GridMetadata, and the phil://docs/imputation-matrix resource.
• Medical demo — demos/medical with covariate sampling, masked iterative imputation, and MDS visualization of descriptor space.

See CHANGELOG.md for full release notes.

uv sync --all-extras
uv run pytest -v
uvx ruff format phil/ tests/
uvx ruff check phil/ tests/

Contributors: see AGENTS.md for package layout and design notes.

Sphinx docs live under docs/source. Build locally:

uv run sphinx-build -M html docs/source docs/build