When ingesting json data, do the json objects have to be flat? I can't find docs about parsing json object trees, but I see the github event example is a tree

It depends on what you are trying to do:

1. If you want to store data as JSON blob in Pinot and use JSON indexing: <https://docs.pinot.apache.org/basics/indexing/json-index>

2. If you want to extract data from nested input JSON and store in a flat schema: 

<https://docs.pinot.apache.org/developers/advanced/ingestion-level-transformations#column-transformation>

<https://docs.pinot.apache.org/users/user-guide-query/scalar-functions?q=groovy>

#2. The row data is several levels deep, and it's not an array but rather a list of id properties { node1: { node2: { node3: { id1: { col1: value1, col2: value2 }, id2: { col1: value1, col2: value2 } } } }

You can write a transform to extract that data and map to a column in Pinot schema

sorry, hit send before my example was finished. id1 is a row, id2 is a row, etc

You should still be able to write per column transform (would be a little inefficient to do it once per column)?

How does Pinot decide what the row level object is? ex: id1, id2, etc vs some other object

Your transform function would need to decide that