Hi there! We’re seriously considering using Koog f...
# koog-agentic-framework
j
Hi there! We’re seriously considering using Koog for our production solutions, but the current 0.x versioning raises some concerns among our clients. From what I understand, Koog is already production-ready — could you please share how close we are to a 1.0.0 release?
a
сс @Vadim Briliantov
j
@Jose in my experience Jetbrains versioning strategy has nothing to do with reliability or readiness, but rather with how prone the interfaces and behaviour are to change. Some clear examples are KMP (used in Prod by major companies while still in Beta), Datetime library and Kotlin DL
👍 1
v
That’s exactly the point, thanks @JCollardBovy! Going back to the original question — Koog is production ready and it has nothing to do with the version number. We’re keeping some room for ourselves so that we can change some APIs in the future. Although, whenever we change some API even now (in 0.*) we always add a
@Deprecated
annotation with replacement (where relevant) and still keep the existing API supported and working — feel free to search for Deprecations, Breaking Changes and Removals here: https://github.com/JetBrains/koog/releases Before 1.0.0 we still would like to: • polish and refine PromptExecutor/LLMClient/LLMProvider API • provide Java API • polish the nodes API and node naming (ex: nodeLLMRequest -> llmRequestNode change, and similar. Of course with deprecation and also keeping the older version working) • provide integration with Spring AI • Add integration with Quarkus, Micronaut • Add integrations with major vector databases (other integrations are also planned, as well as additions to the API, including new strategy types — but those are not a pre-requisite for a 1.0.0 release) As for the date, we’re considering to have 1.0.0 around March/April approximately. This should provide us with enough time to finalize and polish everything. But again — it doesn’t mean that Koog is not production ready (it is). And it doesn’t mean that we will be breaking the API all the time without providing deprecations and replacements (we always do).
👍 2
❤️ 1
j
Thanks @JCollardBovy @Vadim Briliantov — that makes total sense. Really appreciate the detailed context and roadmap!
👍 1
kodee loving 1
r
I've been rewriting
langchain
implementation of
deep-research-agent
into
Koog
for couple of weeks, and this is just a mess of side-effects. Would agree to all Vadim's points.