Writing Online Resolvers

When to Use

Writing a new @online Python resolver
Choosing between a Python resolver and a feature expression (F.http_post, _.foo, SQL file)
Picking the right input/output shape: scalar, Features[...], DataFrame[...], has-many
Pulling fields from related namespaces via dotted feature paths
Adding Now for "as-of" / point-in-time arithmetic (optional; mostly for time-dependent features like age calculations)
Parsing protobuf / JSON / vendor payloads into many features at once
Aggregating a has-many DataFrame down to a scalar (risk score, totals, ratios)
Tagging resolvers, scoping to an environment, or routing to a specific machine type
Writing unit + integration tests for resolvers