Apologies to anyone impacted by yesterday's outage. I'm taking it as an opportunity to shore up the infrastructure — adding the kind of things you'd expect for a production deployment (synthetics, alerting, etc.).
It's a curious thing about software, and really any shared endeavor: as something becomes depended upon by more people, the nature of what it demands from you changes. It's silly for a hobby project nobody uses to have a lot of process. But at some point neglect stops being cool-kid pragmatism and becomes stubborn recklessness. What made you fast early on becomes a liability.
It's a lot like growing up. You do it reluctantly, and usually because you learned something the hard way.
https://t.co/pb1UMcV8mC now has a CLI
```
npx @nshipster/sosumi fetch /documentation/swift/array
```
This lets you fetch and render Developer content directly into LLM-readable Markdown, without connecting to a remote MCP server.
...which ironically would have been especially useful during the outage https://t.co/pb1UMcV8mC had yesterday in the process of shipping the CLI 🫠
https://t.co/Mm7KORlJV8
Today https://t.co/jFknDoasSy joins Hugging Face
Together we will continue to build ggml, make llama.cpp more accessible and empower the open-source community. Our joint mission is to make local AI easy and efficient to use by everyone on their own hardware.
We want to be good citizens of the ecosystem, so https://t.co/pb1UMcV8mC respects robots.txt on every external host. Site owners can opt out by disallowing the sosumi-ai user-agent.
Full crawler policy and contact info: https://t.co/NQ6YytjaxH
Other than that, this is getting pretty close to a feature-complete 1.0 release, I reckon.
Big thanks to everyone who's opened issues and contributed PRs. If you hit any bumps or have anything you'd like to share, please reach out!
AnyLanguageModel 0.7.0 is here, with guided generation for all models — Core ML, MLX, llama.cpp, Ollama, OpenAI, Anthropic, Gemini — plus, a new Open Responses model, and a ToolExecutionDelegate for observing and controlling tool calls.
Check it out!
https://t.co/yhfftTXZ6H
Next up: Improving guided generation with XGrammar.
This release uses a hand-rolled implementation of constrained generation, but we do much better.
I wrote a Swift wrapper around the C++ library and published it here: https://t.co/avlLjRpbah
PR here: https://t.co/adXCIXdnob
Just released AnyLanguageModel 0.6.0, with support for guided generation for MLX and Llama, custom JSON mode for Gemini, model prewarming for MLX, improved DeepSeek API compatibility, and the ability to create custom adapters in your own project.
https://t.co/RFlrPf1O1H
🎉 llama.cpp now has Ollama-style model management.
• Auto-discover GGUFs from cache
• Load on first request
• Each model runs in its own process
• Route by `model` (OpenAI-compatible API)
• LRU unload at `--models-max`
https://t.co/yfmfHL7zzj
In my collaboration with @huggingface, I'm on a mission to make it easier to build apps that leverage local, open-source AI models.
So much of this has been community-driven, and I really appreciate the feedback I've received so far.
What would you like to see next?
I built this package in response to feedback we received about the HubAPI implementation in swift-transformers.
Try swift-huggingface today, and stay tuned for its integration into swift-transformers in an upcoming release.
https://t.co/v5yaB36NFf
Just published an article on @huggingface's blog about AnyLanguageModel that goes into more detail about the motivation behind the library and what we're building toward:
https://t.co/xu79guTCFR
🆕 Be sure to check out our new chat-ui-swift example project (linked at the end)
@pj4533@bcherny Sosumi currently serves text by default for docs. It'd be great if Apple's website supported Markdown automatically, but until then, we'll need something like Sosumi.
```
❯ curl -I -H "Accept: text/markdown" https://t.co/LYpt6W8oRf
Content-Type: text/html; charset=utf-8
```
@adityadaniel Yes, that's correct. This is something we'd like to support, and is actively being worked on now.
Follow this issue to track our progress: https://t.co/XgX2tQ6shF
Thanks to everyone who's checked out AnyLanguageModel so far!
I just cut a new 0.3.0 release with a Gemini adapter that supports Google search grounding, code evaluation, and more: https://t.co/JjDXwGTOh9
Next up — Documenting best practices for managing API credentials in your app. The current README shows usage with environment variables, but that only works server-side. There are a few good options (and one very bad option), and it'd be helpful to write those up.
Introducing AnyLanguageModel: A Swift package that provides a drop-in replacement for Apple's Foundation Models framework with support for custom language model providers.
https://t.co/2sbM9ITVie
Just change your import statement:
AnyLanguageModel 0.4.0 is out with support for multi-modal inputs.
Vision language models are incredibly useful — extract text from receipts, analyze diagrams, describe images for accessibility, answer questions about photos, and much more.
You can now pass images directly to models running with MLX & Ollama, plus Anthropic, Gemini, and OpenAI providers. Core ML support coming soon.