spotify/scio ? reverse-engineered prompt

Build me a Scala library for writing Apache Beam pipelines that feels simple and familiar, kind of like Spark or Scalding. I want it to support both batch and streaming jobs with one API, and make it easy to read from and write to common cloud and data tools like BigQuery, Pub/Sub, Cloud Storage, Avro, Parquet, JDBC, Redis, Cassandra, Elasticsearch, and TensorFlow records. Include a small example app like a word count job and make it easy to run locally and on Google Cloud Dataflow. If it helps, use the current docs online for Beam and Dataflow details. Also include a clean test setup and a few example pipelines so I can see how the pieces fit together.