PydanTable

Strongly-typed DataFrame layer for FastAPI and Pydantic services, with a Rust execution core (Polars-backed inside the native extension). This site is the full manual; the repository README is the short entrypoint for install one-liners.

See also

Execution builds on Polars (see their user guide) and ships as a native Rust extension. PydanTable adds Pydantic-first schemas, SQLModel-style tabular models, and service / data I/O patterns on top.

Note

Current release: see CHANGELOG — stable 1.x under VERSIONING. Roadmap: ROADMAP; SQLModel milestones: SQLMODEL_SQL_ROADMAP.

Minimal example

from pydantable import DataFrame, Schema


class Row(Schema):
    id: int
    score: float


df = DataFrame[Row]({"id": [1, 2], "score": [10.0, 20.5]})

Choose your path

• Services (FastAPI)

  ---

  Shortest runnable path, HTTP patterns, register_exception_handlers (503 / 400 / 422), and cookbooks. Tests: pydantable.testing.fastapi. Troubleshooting: TROUBLESHOOTING.

  Golden path →

• Data & I/O

  ---

  DataFrameModel, lazy vs eager I/O, IO_OVERVIEW, and format guides (Parquet, CSV, SQL, …). Pandas-like helpers: PANDAS_UI.

  I/O decision tree →

• Library & contracts

  ---

  Public API guarantees, VERSIONING, plans/plugins (PLAN_AND_PLUGINS), and custom engine packages (CUSTOM_ENGINE_PACKAGE).

  Interface contract →

• Optional engines

  ---

  SQL (SQL_ENGINE, pydantable[sql]), Mongo (MONGO_ENGINE, pydantable[mongo]), Spark (SPARK_ENGINE, pydantable[spark]). Beanie: BEANIE. PySpark façade: PYSPARK_UI.

  Spark engine →

Where pydantable fits

• New here? Start with the Five-minute tour, then read the Mental model.
• Deciding between tools? Start with Why pydantable? and the Comparisons (Polars + pandas).
• Looking for recipes by intent? See Workflows (cookbook).

Guide map

• Services: GOLDEN_PATH_FASTAPI → FASTAPI → FASTAPI_ADVANCED → FASTAPI_ENHANCEMENTS → DATAFRAMEMODEL → EXECUTION. Cookbooks: fastapi_columnar_bodies, fastapi_observability, fastapi_background_tasks, async_lazy_pipeline. Layout: docs/examples/fastapi/service_layout/.
• Data workflows: DATAFRAMEMODEL → IO_DECISION_TREE → IO_OVERVIEW.
• Typing: TYPING — mypy plugin vs Pyright / Pylance / Astral ty, and as_model(...) on transform chains.
• Third-party execution: CUSTOM_ENGINE_PACKAGE. Mongo helpers and Beanie reads/writes: IO_DECISION_TREE and BEANIE.
• Not sure where something lives? DOCS_MAP.