I've created an agentic framework to make working ...

# feed

[Jeff Hudson]

I've created an agentic framework to make working with LLMs a breeze!

Our company is rewriting our RAG chatbot & AI Agent tasks to use a network of smaller, more focused agents, which seems to be what the broader industry is moving towards.

https://github.com/hudson155/osiris-ai

Here's a trivial example, but there's more examples in the README!

val modelFactory: ModelFactory =
  modelFactory {
    openAiApiKey = ProtectedString("...")
  }

val response = llm(
  model = modelFactory.openAi("gpt-4.1-nano"),
  messages = listOf(
    UserMessage("What's 2+2?"),
  ),
).get()

response.convert<String>()
// 2 + 2 equals 4.

[Fudge]

I assume this only supports remote LLMs?

[Jeff Hudson]

It's built around Langchain4j, and should support anything and everything that Langchain4j supports!

I've only used it with remote LLMs, but it looks like local Ollama & a few more are supported
https://docs.langchain4j.dev/integrations/language-models/

[Jeff Hudson]

The ModelFactory part in the example won't work without some adjustments, but in

val response = llm(
  model = modelFactory.openAi("gpt-4.1-nano"),
  messages = listOf(
    UserMessage("What's 2+2?"),
  ),
).get()

model = actually takes an instance of dev.langchain4j.model.chat.ChatModel, so you can instantiate that however you wish

[hho]

Have you seen Jetbrains Koog, introduced on this years KotlinConf? It seems to be pretty much the same thing?

[Jeff Hudson]

Oh wow, this looks very promising! I hadn't found this yet. Definitely going to take a good look at it

[hho]

I just looked through your repo and found that you started before they published it. Maybe you inspired them? 🙂

[Jeff Hudson]

I'll take credit for that /s

[hho]

Seriously, I think it's cool how similar your example above looks to their "getting started" one

[Jeff Hudson]

yeah, and the agentic example with tools looks even more similar!
(screenshot from my repo)

[Jeff Hudson]

Although I think it's more likely that we were both inspired by
https://github.com/openai/openai-agents-python