Split off from #2285's gap 3.
create_topics and delete_topics used to be the only two binding-mcp-kafka tools exposing a caller-settable timeout argument, overriding zilla.binding.mcp.kafka.request.timeout (default PT30S) for that one call. That was an inconsistency: every other tool making a blocking Kafka call (alter_configs, produce, consume, describe_configs, list_acls/create_acls/delete_acls, describe_topic, list_consumer_groups, describe_consumer_group(_lag), reset_offsets) has no such override, relying entirely on the configured default.
Resolved for now by removing the argument from create_topics/delete_topics (see the PR linked from #2285) rather than adding it everywhere, since generalizing it would mean touching ~12 more tool schemas, their transform Source parsers, and k3po fixtures -- too large to decide unilaterally.
Open question for a maintainer: is a general-purpose timeout argument (uniformly available across every blocking tool, or scoped to a specific subset like admin operations that can legitimately hang on controller/coordinator state) worth adding later? If so, this issue tracks that follow-up design + implementation.
Split off from #2285's gap 3.
create_topicsanddelete_topicsused to be the only twobinding-mcp-kafkatools exposing a caller-settabletimeoutargument, overridingzilla.binding.mcp.kafka.request.timeout(defaultPT30S) for that one call. That was an inconsistency: every other tool making a blocking Kafka call (alter_configs,produce,consume,describe_configs,list_acls/create_acls/delete_acls,describe_topic,list_consumer_groups,describe_consumer_group(_lag),reset_offsets) has no such override, relying entirely on the configured default.Resolved for now by removing the argument from
create_topics/delete_topics(see the PR linked from #2285) rather than adding it everywhere, since generalizing it would mean touching ~12 more tool schemas, their transformSourceparsers, and k3po fixtures -- too large to decide unilaterally.Open question for a maintainer: is a general-purpose
timeoutargument (uniformly available across every blocking tool, or scoped to a specific subset like admin operations that can legitimately hang on controller/coordinator state) worth adding later? If so, this issue tracks that follow-up design + implementation.