sangkyoonnam commented on issue #1190: URL: https://github.com/apache/flink-agents/issues/1190#issuecomment-6083511961
An update since my note above: #1155 made an explicit `durable_id` bypass the argument digest, and `tool_call_action.py` now passes `"tool-call:" + execution.id`. The built-in chat action still calls `durable_execute` / `durable_execute_async` without one, so the instability reported here is still on main (d6e2ecc9). That leaves two scopes. The narrow one gives each chat model attempt an explicit id built from the model resource name and the retry attempt, which doesn't depend on the request UUID. It fixes the built-in path without touching `_compute_args_digest`. The broad one is the canonical digest I described above, which also covers user calls without an id. Either way, existing identities change, so persisted results inside unfinished actions can be invalidated and those calls run again after the upgrade; completed actions still replay. @wenjin272, which would you prefer? I can do the narrow one first and leave the digest change as a follow-up. -- 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]
