Turn any REST API into a standalone Lakeflow Declarative Pipeline, just generated code.

Welcome to Polymo

Setting up API ingestion in Databricks Lakeflow Declarative Pipelines is tricky, because it involves classes and requires knowledge of inner working of declarative pipelines. Polymo is a tool that helps to define the pipeline from a UI. It allows you to preview the ingestion dataframes locally, and export a standalone Lakeflow Declarative Pipelines pipeline. polymo is never a dependency of the scripts it generates. The output is plain Python that only needs requests, the standard library, and pyspark. Nothing is imported from polymo, and there is no config file to load at runtime. The generated code is yours: edit it, version it, review it like any other code.

What Polymo generates

Pointing Polymo at https://jsonplaceholder.typicode.com/posts produces a single file. Shown here, the real script also contains the retry/backoff and response-normalization helpers.

"""Lakeflow Declarative Pipelines connector for posts."""

import time
from typing import Any, Iterator

import requests

BASE_URL: str = "https://jsonplaceholder.typicode.com"
PATH: str = "/posts"
PARAMS: dict[str, Any] = {"_limit": 20}
HEADERS: dict[str, str] = {}
TIMEOUT: float = 30.0


def fetch_records() -> Iterator[dict[str, Any]]:
    ...


from pyspark import pipelines as dp
from pyspark.sql import SparkSession
from pyspark.sql.datasource import DataSource, DataSourceReader


class RestSource(DataSource):
    ...


spark = SparkSession.getActiveSession()
spark.dataSource.register(RestSource)


@dp.table(name="posts")
def posts():
    return spark.read.format("posts_source").load()

Every field you fill in: pagination, incremental sync, partitioning, filter pushdown, error handling, headers, query parameters, an explicit schema, XML responses, streaming tables is baked in as a constant or a small block of specialized code. Options you don't use produce no code at all.

How does it work?

Open Polymo, describe your API (base URL and path are the only required fields), and press Preview to see the DataFrame, the raw records, and the raw API responses side by side. When you're happy, switch to the Generated Code tab and download the script or bootstrap a Databricks project.

Ready to run it on Databricks? The Deploy tab walks through it: Select Profile → Target → Bootstrap → Deploy → Run:

1. Pick a Databricks CLI profile, then a catalog and schema (or type a schema name).
2. Bootstrap writes a complete Databricks Asset Bundle project: the connector as an installable package under src/, the pipeline under pipelines/, and a databricks.yml that builds the wheel and attaches it to a serverless pipeline.
3. Deploy and Run drive databricks bundle deploy and databricks bundle run without leaving Polymo, with the CLI output docked below.

Secrets never land in generated code. Each auth field (bearer token, API key, OAuth2 client secret) and any {{ options.<name> }} placeholder can reference a Databricks secret scope or a Unity Catalog service credential backed by Azure Key Vault; the pipeline resolves the value on the driver and passes it to the reader, and Polymo redacts it from every preview. Requires the Databricks CLI and a ~/.databrickscfg profile, see Deploy to Databricks for the full walkthrough.

See the Connector options reference for what every field generates, and the Polymo UI walkthrough for a guided tour.

How to start?

pip install polymo

Launch the UI

polymo

Or without installing it first:

uvx polymo

(Optional) Run Polymo in Docker

docker compose up --build ui

The service listens on port 8000; open http://localhost:8000 once Uvicorn reports it is running.

Migrating from polymo 0.x?

The YAML runtime (spark.read.format("polymo"), PolymoConfig, polymo smoke) is gone since 1.0. See docs/migration-1.0.md for what changed, and pin polymo<1.0 (0.11.0 is the last release with the old runtime) if you're not ready to move yet.

Where to Next

Read the docs here.

Other material:

• Step by step example: medium blogpost (written for the 0.x YAML runtime, see the migration guide for what changed)

Contributing

Is there something missing? Raise an issue or contribute! Contributions and early feedback welcome.

pytest runs the unit suite against a local mock API and a local Spark session. POLYMO_LIVE=1 pytest tests/live additionally reads every connector option against public APIs (jsonplaceholder, PokeAPI, GitHub, GitLab, Wikipedia, httpbin, arXiv, ...) through Spark's Python Data Source API; the live workflow runs that weekly and on pull requests that touch code generation.

---

If Polymo helped, a ⭐ makes my day