Category: AI

ComfyUI – v0.3.66

ComfyUI v0.3.66 is live 🚀 — and it’s a quiet powerhouse for your AI workflows!

✨ New `LatentUpscale` node — upscale in latent space before decoding for sharper, cleaner high-res results with less noise.

⚡ Memory optimized — fewer spikes during batch processing, perfect for mid-tier GPUs.

🔍 Faster node search — partial matches work now! Type “upscale” and get all related nodes instantly.

🧩 Custom node fix — no more vanishing nodes after reloads (we feel you 😅).

🎨 UI polish — smoother transitions + zoom snapping to 25%/50%/100% for pixel-perfect control.

Pro tip: Pair `LatentUpscale` with KSampler + High-Res Fix for insane detail without VRAM overload.

Upgrade now — your next masterpiece is just a click away. 🖼️💻

🔗 View Release

MLX-LM – v0.28.3

🔥 MLX LM v0.28.3 is LIVE! 🔥

Heads up, Apple silicon LLM tinkerers – LLaMA-Factory just dropped a massive update for MLX LM! This release is packed with refinements and new features to help you build, train & serve even better models.

Here’s the breakdown:

• Memory Efficiency: State Space Models (SSM) are leaner now. 🙌

MoE Magic: Lots of improvements to Mixture of Experts – LoRA fixes, bailing logic, and* a new LFM2 option!

• Qwen3-VL Support: Visual language model support added with Qwen3-VL (plus a dense version!). 🖼️
• Faster GPT2: Batch processing for GPT-2 just got quicker.
• DWQ Tweaks: Depthwise Quantization refined with temperature adjustments.
• Python 3.9 Love: Qwen3 support now extends to Python 3.9 users!
• Plus: Cleaned up params, simplified I/O, CUDA install fixes, batched SSM masking, gradient accumulation, data parallel eval, Jamba support & LLM Benchmarks! 📊

Dig into the full changelog – there’s a ton here to play with! 🎉

🔗 View Release

LMStudio – 1.6.0b1 – 2025-10-28

LM Studio Python SDK v1.6.0b1 is live! 🚀

This release focuses on making it even easier to interact with your LM Studio server APIs from Python. Here’s what’s fresh:

• Smoother Error Handling: Say goodbye to crashes when the server sends back pre-formatted errors – now handled gracefully!
• API Token Authentication: Added support for API tokens, giving you more flexibility in how you authenticate your requests. 🔑
• New Contributor: A big welcome to @ryan-the-crayon! 🎉

Check out the full changelog for all the details: https://github.com/ncoghlan/server-api/compare/1.5.0…1.6.0b1

🔗 View Release

Ollama – v0.12.6: cuda: bring back CC 5.2 (#12666)

Ollama just dropped v0.12.6! 🎉

Need to run large language models locally? Ollama makes it easy – think GPT-4, Llama 2, right on your machine. No API limits, more privacy!

This release brings some good stuff for NVIDIA users:

• Older GPU Support: They’ve brought back support for Compute Capability (CC) 5.2 CUDA devices. Dust off those slightly older NVIDIA cards!

Driver Harmony: Fixes a compatibility issue with newer drivers and* keeps the CC 5.2 experience smooth.

Dig into the details: https://github.com/ollama/ollama

🔗 View Release

Tater – Tater v35

🏠 Tater v35 – “Eyes Everywhere” just dropped! 🥔

Your smart home just leveled up with an Event System—giving your AI long-term awareness of everything happening around the house (doorbell rings, dog playtime… you name it!). Think of it as a memory for your home! 🧠

Here’s what’s new:

• Event Storage & Retrieval: Core platform now stores events via Redis.
• New Endpoints: Add (`POST`) and search (`GET`) events with ease.

Retention Control: Configure how long Tater remembers – from 2 days to forever*.

• Plugins Galore:

🚪 Doorbell Alert (updated): Captures images, describes visitors, speaks alerts, and* logs structured events! Talk back through compatible doorbells too. 🗣️

• 🎥 Camera Event (new): Motion detection + Vision LLM descriptions = smart camera monitoring.
• 🧠 Events Query (new): Ask natural language questions like “What happened in the front yard today?” and get a summary!

Tater syncs time with your Home Assistant & cleans up old events automatically. 🕓 How it works: HA triggers Tater, AI Router picks the plugin, events are logged, and the Query Plugin brings it all back to life!

README for details & installation.

🔗 View Release

Ollama – v0.12.6-rc1

Ollama v0.12.6-rc1 is here! 🚀

Running local LLMs just got a little smoother. Ollama lets you spin up models like Llama 3 & Mistral right on your machine – perfect for tinkering and building AI apps without hitting an API.

This release brings:

• More Embedding Models: More options for semantic search & vector databases with added models like `all-MiniLM-L6-v2`.
• Bug Fixes & Polish: General improvements to keep things running nice and tight! 🛠️

Dig into the full release notes: https://github.com/ollama/ollama/releases/tag/v0.12.6-rc1

🔗 View Release

Ollama – v0.12.6-rc0: CI: Set up temporary opt-out Vulkan support (#12614)

Ollama just dropped v0.12.6-rc0! 🎉

What is Ollama? It lets you run large language models locally – think Llama 2, Mistral, and more – right on your Mac or Linux machine. Super handy for testing, building apps, or just playing around without hitting API limits.

Here’s what’s new:

• Experimental Vulkan Support: Want a potential performance boost? You can now opt-in to use Vulkan! Requires building from source currently.
• CI Updates: Some under-the-hood improvements to the continuous integration setup for smoother sailing.

Want to try Vulkan? Check out the Ollama docs for build instructions! 💻

🔗 View Release

Text Generation Webui – v3.15

llama.cpp & ExLlamaV3 just got a nice polish in text-generation-webui v3.15! 🚀

This release focuses on stability and getting more out of your models with fixes for context size errors, security boosts (immutable `–trust-remote-code`), and improved chat behavior. Plus, those pesky gibberish outputs in ExLlamaV3 v3.14 are fixed 🙌.

Key updates:

• Context Size Fix: No more changelog errors!
• Security: `–trust-remote-code` is now immutable for extra security. 🛡️
• Chat Improvements: Metadata leaking and “continue” spacing fixed.
• Download Resilience: Better resuming of downloads from Hugging Face.
• ExLlamaV3 Reverted: Gibberish outputs solved!
• Core Updates: llama.cpp (f9fb33f2630b4b4ba9081ce9c0c921f8cd8ba4eb) & ExLlamaV3 (v0.0.10)!

Grab the latest portable builds for Windows/Linux (cuda12.4, cuda11.7, vulkan, cpu) or Mac (arm64, x86_64). Just unzip and replace your `user_data` folder to update – all settings & models are preserved! ✨

🔗 View Release

Tater – Tater v34

Tater just leveled up! Now it doesn’t just chat—it acts. This release (v34 – “The Automator Awakens”) brings AI-powered automations to your Home Assistant setup. Think of it as a dedicated AI engine for smart home actions. ⚙️✨⚡️

Here’s the breakdown:

Home Assistant Automations: Add Tater as a second Conversation Agent in HA, pointing to port 8788 (default). One agent for talking, one for doing*. 🧠💥🚪

• Doorbell Alert! Plugin: When your doorbell rings…
• Tater grabs a live snapshot from your camera.📸
• Uses Vision AI to identify what’s there (people, packages, cars!). 👀
• Announces the alert via Piper TTS on your Voice PE media players. 🎙️

Optionally sends a notification with* light ring to Home Assistant!💡

• More is coming: Expect more smart-triggered actions: security snapshots, motion alerts, device routines & AI home summaries. The future is reactive (and potato-powered)!🥔⚡️

Ring that doorbell and see Tater in action! 🔔😎

Check it out: https://github.com/TaterTotterson/Tater/releases/tag/v34

🔗 View Release

ComfyUI – v0.3.65

🔥 ComfyUI just dropped v0.3.65! 🔥

ComfyUI is a modular and customizable UI for creating Stable Diffusion workflows – think node-based image generation power! 🎨

Here’s what’s new:

• Custom Node Groups: Save frequently used node groups as custom nodes for easier reuse across workflows. ✨
• Improved LoRA Support: Better handling of LoRAs (Low-Rank Adaptation) for more refined style control. 🖼️
• Performance Boosts: Various optimizations to make your image generation even faster! 🚀
• Bug Fixes & Polish: A bunch of quality-of-life improvements and pesky bug squashes. 🐛

Check out the full release notes on GitHub for all the details! Happy node-wrangling! 🤓

🔗 View Release