madroid

[madroid @Madroidmaq]

🚀 mlx-omni-server v0.4.3 just dropped!
✨ Revolutionized reasoning model handling:
✅ Auto-parses responses into reasoning + content fields
✅ Toggle logic chain visibility with enable_thinking
🔥 Try with deepseek-ai/DeepSeek-R1-0528-Qwen3-8B
🔗 GitHub: https://t.co/eMCmeJubFC

[madroid @Madroidmaq]

Built a @raycast extension for Claude Code @claudeai

Manage your Claude plugins visually:
✨ Browse plugin marketplace
⚡️ One-click install/update
🎯 Filter by scope & status

Beats typing `claude plugin install` every time!

see https://t.co/6lRcr1KZRC

[madroid @Madroidmaq]

@Rh1ve @0ssamaak0 Maybe you can provide more detailed error information on GitHub to help solve the problem.

[madroid @Madroidmaq]

🎉 mlx-omni-server v0.4.0 is out!

✨ New /v1/embeddings service (via mlx-embeddings, thx @0ssamaak0) - generate embeddings easily!
🔊 More TTS models (kokoro, bark & more!) with mlx-audio (thx @ zboyles) !
🧠 mlx-lm upgraded for models like qwen3!

https://t.co/iGK9vfbRuK

[madroid @Madroidmaq]

@stanley520 @0ssamaak0 you can try pip install -U mlx-omni-server

[madroid @Madroidmaq]

@0ssamaak0 Also, a huge thanks to @Prince_Canuma and @awnihannun for their continuous iterations on the MLX framework, making all of the above much easier to achieve!

[madroid @Madroidmaq]

🚀 MLX Omni Server Update!

New version adds Prompt Cached for faster AI inference + cached_tokens to track cached counts.
Boost efficiency now!

👉 https://t.co/eMCmeJuJva

#mlx #openai

[madroid @Madroidmaq]

@ivanfioravanti I didn't notice this, theoretically there should be no breaking change, could you describe the problem you encountered in detail?

[madroid @Madroidmaq]

🔍 New in v0.3.0:
- OpenAI-compatible Structured Output
- Configurable worker parallelism
- Pydantic model integration

Check out how easy it is to use! ⬇️

#MLX #LocalAI

[madroid @Madroidmaq]

🎉 mlx-omni-server v0.3.0 is out!

Introducing Structured Output support for type-safe AI applications on Apple Silicon 💻

❤️ Huge thanks to @OutlinesOSS for the implementation!

🔗 https://t.co/eMCmeJuJva

#MLX #LocalAI #AppleSiliconc

[madroid @Madroidmaq]

@PHIDATA 🛠️ MLX Omni Server updates for developers:

- Enhanced Function Calling accuracy
- Prefill response support
- Stream options with token stats
- Custom stop tokens
- Improved code structure

One command to start:

pip install -U mlx-omni-server

https://t.co/5z4TKLDnEU

[madroid @Madroidmaq]

🚀 MLX Omni Server v0.2.0 is here!

Major improvements in Function Calling accuracy:
- Llama3.2 3B (4bit): 2.9% → 99.6%
- Qwen2.5 3B (4bit): 48.4% → 99.0%

Now you can run powerful AI tools locally on your Mac with minimal latency!

Try it now: https://t.co/eMCmeJuJva

#MLX

[madroid @Madroidmaq]

💡 Function Calling just got much better!

With @phidata, it's incredibly simple to add web search to your AI app - even with a tiny 4-bit quantized 3B model!

Check out how clean this implementation and the result

[Awni Hannun @awnihannun]

Very cool new project: MLX Omni Server by @Madroidmaq

Local APIs for text generation, speech generation and transcription, image generation, and more.

[madroid @Madroidmaq]

@peduarte Maybe add slash command as well as show md raw mode

[madroid @Madroidmaq]

@ollama When will Phi-3.5-vision be supported?

[madroid @Madroidmaq]

@jantxuuu @huggingface @zeddotdev @MistralAI That's awesome. How did you do that?

[madroid @Madroidmaq]

@karpathy It's great that if there are other display formats for content to appear, its distribution will be even wider. Regarding the presentation formats you mentioned, I found a tool that can do something similar, such as one of the courses below.https://t.co/ANwcOsHPrF

[madroid @Madroidmaq]

@botteaap @Google @huggingface This demo is awesome, I would like to know what App you used for the demo and how should I get access to it? Looking forward to your reply!