ai

kapil 2025-10-29T08:12:23.447969Z

Announcing release of litelllm-clj ! https://github.com/unravel-team/litellm-clj This is a port of python library - https://docs.litellm.ai/. It’s an adapter layer that connects to a lot of providers and LLMs. I wanted Clojure to have a foundation piece so more AI work can happen. There is still a lot to be done in the library. But, I am happy about the progress so far. Happy to hear thoughts about API. Some things that are pending, 1. Tool calling API, It’s not quite right yet. I would like to have another go at it eventually 2. Observability, I want to include out of the box observability integration 3. Examples, Another repository with examples on how to integrate it with different Clojure libraries. 4. Reasoning tokens API - https://github.com/unravel-team/litellm-clj/issues/14 I had released instructor-clj a while back. It now uses litellm-clj https://github.com/kapilreddy/instructor-clj It’s a big surface area to cover so happy to hear feedback!

🚀 11
2025-10-29T08:35:07.524909Z

For reach maybe cross post in #announcements and #ai-assisted-coding?

kapil 2025-10-29T08:39:19.464559Z

Hey thanks. Doing it...

kapil 2025-10-29T08:43:11.257769Z

@trost.mario Since this is a library #ai-assisted-coding might not be suited for this.

2025-10-29T08:44:28.623799Z

Yeah that’s true, there’s just so much more activity over there that sometimes people post ai libraries there with the moniker like “Not strictly about ai assisted coding, but a library that might interest people here” or some such 🙂

2025-10-29T08:44:43.830619Z

But #announcements definitely has the bigger reach!

1
2025-10-29T08:44:51.070979Z

And is on-topic!

ericdallo 2025-10-29T12:20:03.318429Z

That's awesome! I really searched for this when making #eca, and most of eca's llm-api probably can be reused there, especially the tool calling, maybe in the future we can integrate with it! But I'd say eca LLM API is pretty dense now, we would need to add more capabilities to litellm-clj

kapil 2025-10-29T12:21:07.933219Z

@ericdallo thanks! I can try getting to have feature parity to what ECA needs.

💙 1
kapil 2025-11-03T14:52:56.578309Z

A brief intro to the API https://youtu.be/R7QLHxGf6GU

👍 2
ericdallo 2025-11-03T14:54:29.279099Z

question: when I see litellm/completion is not clear if for :openai it will use Responses API or Chat API (legacy) - you may want to support both

ericdallo 2025-11-03T14:55:23.604539Z

FYI @zikajk

kapil 2025-11-03T14:56:37.760939Z

Currently it does legacy API call. Yes I would need to support both.

kapil 2025-11-03T14:56:57.060679Z

I think default should be Responses API.

ericdallo 2025-11-03T14:57:43.788129Z

Makes sense! Also, stream/not stream would be nice to support

ericdallo 2025-11-03T14:58:02.956499Z

Recently I added support for non-stream in ECA for openai-chat, all other only support streaming, you may want to take a look and use as inspiration

kapil 2025-11-03T15:00:34.748219Z

Yes. There is a dedicated API for streaming. Everything else is non-stream.

oλv 2025-11-03T15:03:19.631609Z

Great to see this library! Definitely fills a vacuum :^) Could it support embedding models too?

🙌 1
ericdallo 2025-11-03T15:03:27.406539Z

awesome, API wise looks pretty solid!

🙌 1
kapil 2025-11-03T15:40:26.172709Z

@post972 I am not sure. I suspect it might need some dependency to support embeddings.

oλv 2025-11-03T15:58:37.067249Z

@kapil The embeddin APIs I've used are exposed via HTTP, e.g the OpenAI embedding API. I don't think you need any extra dependencies :^).

1
kapil 2025-11-03T15:59:13.588169Z

Then it should be straightforward to implement.

kapil 2025-11-08T14:28:24.200639Z

litellm-clj 0.3.0-alpha.1 released • Add support for Reasoning tokens • Add support for Embeddings API • Add support for GPT-5* models

🔥 6