Update `mlx-swift-lm` dependency to 3.x to support newer model architectures (Gemma 4, etc.)

[james-333i]

Summary

AnyLanguageModel 0.8.0 pins mlx-swift-lm to from: "2.25.5", which resolves to the 2.x line (<3.0). New model architectures are now landing only in mlx-swift-lm 3.x, which is unreachable from the current dependency range, so AnyLanguageModel consumers can't run them.

Evidence (mlx-swift-lm releases)

• 2.30.6 (18 Feb) — added Qwen3-VL (dense)
• 2.31.3 (1 Apr) — added Qwen3.5 (+ MoE), text & vision
• 3.31.3 (15 Apr) — first 3.x release (breaking API); adds Gemma 4 (text, vision, MoE / E2B & E4B edge variants)

So Qwen3.5 and Qwen3-VL are reachable today (we pin mlx-swift-lm 2.31.3, within AnyLanguageModel 0.8.0's range and it works great). But Gemma 4 — and any future architecture added only on the 3.x line — is blocked, because 3.x has breaking changes that 0.8.0 wasn't built against.

Request

Update AnyLanguageModel's mlx-swift-lm dependency to 3.x (adapting to the 3.x API changes) and cut a release, so newer model families (Gemma 4 first) become available to consumers. More broadly: keeping the mlx-swift-lm dependency current would let new architectures work without each one waiting on an AnyLanguageModel bump.

Happy to test a branch/pre-release against a real iOS app (on-device, multi-turn) if that helps.

[pcuenca]

This has been addressed in #161. We need to publish a new release, which will be 0.9.0. I'm waiting, in turn, for a new release to land in mlx-swift-lm, because we discovered a regression when building with Xcode 16 (SDKs earlier than 26) that has been fixed, but not released as a version yet. When it does, rerunning CI on this PR should be green.

Happy to test a branch/pre-release against a real iOS app (on-device, multi-turn) if that helps

This would be awesome!

Also, please let me know if a new release would be helpful even if builds are broken for Xcode 16.

[james-333i]

@pcuenca Thank you. While I'm not using Xcode 16, some may be. I'll wait to test when you roll out as 0.9.0 given the Xcode issue.

[SpiraMira]

@pcuenca - I humbly vote for a new release that supports Xcode 26+ as soon as possible. Am also waiting on support for newer models. Thx in advance.

[ also available to test against my (unreleased) multiplatform app that leverages this package for coreml, coremlkv(stateful) and mlx. minimum sdk = 15.0]

[pcuenca]

Version 0.9.0 has just been released 🚀

I left out #168 for now, as it touches the new beta APIs. I'll test it locally before submitting a review, then we can have it land on main when ready. Thanks @james-333i @SpiraMira!

[james-333i]

Thanks @pcuenca . I tested Gemma 4 and it seems to have issues with tool calling.

It has this error:
LLM Streaming failed: decodingFailure(AnyLanguageModel.LanguageModelSession.GenerationError.Context(debugDescription: "Detected repeated MLX tool-call signature and aborted to avoid an infinite tool loop."))

[james-333i]

@pcuenca Also, Vision is not working on Gemma 4 even though its listed as having support.

I specifically tested:
mlx-community/gemma-4-E2B-it-qat-4bit

I attach images but it acts like they are not attached.

[james-333i]

@pcuenca Happy to help test any changes as well if you find a solution. For reference, here is a Test Flight of our app Infinity Vaultz: Private AI where you can test Gemma 4 (along side many other models). Currently, I have tooling and vision turned off because of these issues.

https://testflight.apple.com/join/Zm3TX3Fa

Btw, it would be cool to start compiling a list of apps using the library as a showcase on the main ReadMe so use cases can be seen by the community.

[SpiraMira]

Version 0.9.0 has just been released 🚀

I left out #168 for now, as it touches the new beta APIs. I'll test it locally before submitting a review, then we can have it land on main when ready. Thanks @james-333i @SpiraMira!

@pcuenca - fyi smoked 0.9.0 + the #168 feature locally and all green so far. thx.