weiqingy commented on issue #829: URL: https://github.com/apache/flink-agents/issues/829#issuecomment-4700857162
For what it's worth, here's where I personally land after the audit — sharing it as one opinion to react to, not a direction. The priority call is the community's. I'd lean toward eventually retiring the *user-facing* local path (the no-arg `get_execution_environment()` plus the `from_list` / `to_list` quickstart and `ReActAgent`'s default), rather than keeping it indefinitely or removing it outright today. Two things push me there: the dual-API divergence is a recurring tax — it's what made #722 awkward — and the local path is already a partial simulation, since long-term memory, metrics, and durable execution are all no-ops there. So the "run without Flink" experience doesn't actually mirror production, which undercuts its value as the thing we point new users at. I wouldn't commit to removal yet, though, because the main argument *for* keeping it — fast JVM-free unit tests — is currently unmeasured, and the migration isn't mechanical (`from_list` / `to_list` have no MiniCluster equivalent, so affected tests/examples would need rewriting onto sources/sinks). So the sequence I'd suggest: 1. Fix the dead Java no-arg `getExecutionEnvironment()` first — it resolves a `LocalExecutionEnvironment` class that doesn't exist and throws at runtime, independent of this decision. I split it out as #847. 2. Measure the JVM-free unit lane against an embedded-MiniCluster run on a few representative agent-logic tests, so the keep-vs-remove call rests on a real number rather than an assumption. 3. Then choose between deprecate-and-migrate, or shrinking to a thin in-process harness for unit tests while dropping just the user-facing quickstart slice. Either way `LocalMemoryObject` can stay as a test fixture — it's separable from the runner. Happy to take any of these forward if there's appetite — I'll leave the direction and priority to the community. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
