News
Latest news and updates on local artificial intelligence
10 news items in Use cases
Open WebUI 0.12.0: realtime voice calls
October 10, 2026The new Open WebUI release adds a Realtime mode for voice calls: a voice model handles the conversation and hands real questions to the selected chat model, which answers using its history and tools. It can be enabled by administrators in the audio settings.
RunningTab: direct workspace interaction with file tabs for AI agents
October 8, 2026A new Hugging Face paper proposes that LLM agents work directly on workspace files through dedicated tabs, with no indexing: the agent searches and reads documents from a terminal and produces deliverables from multiple files in one flow.
AI recreates Photoshop and Premiere in Rust, open source
October 8, 2026The ArtCraft project released Photocraft, Filmcraft and Lightcraft: Rust reimplementations of Photoshop, Premiere and Lightroom written with AI agents, available as native binaries for Linux, macOS and Windows. An interesting demonstration of what agents can do, with open legal questions.
Moose: the GNOME local-AI app runs on Ollama
October 3, 2026Moose is a new native GTK4/libadwaita app for GNOME, written in Rust under GPLv3, that lets you chat with local models via Ollama: it handles Ollama setup, model management and loading local documents and images. It's on Flathub, the code is on GitHub, and it was announced on This Week in GNOME.
funcroute: a router that gives a text-only LLM model eyes and ears
October 2, 2026A small open-source router (v0.4.0) that adds multimodal capabilities to a text-only LLM model: images and audio are handled at the routing layer, so the local model stays the one you already run, without swapping it out.
A wall that listens: the local-LLM pipeline behind an AI party in Vilnius
September 27, 2026A creative experiment in Vilnius: a wall with microphones and speakers powered by a fully local LLM pipeline. Participants talk to the wall and get responses generated on site — no cloud, no external servers.
ComfyUI v0.37.4: stability patch for local generation workflows
September 25, 2026ComfyUI reaches v0.37.4, a patch in the 0.37.x series following the native Qwen-Image 2.1 support introduced in v0.37.0. If you use ComfyUI for local image generation, this is the build to run right now.
Open WebUI v0.11.4: per-file metadata and faster hybrid search on large knowledge bases
September 24, 2026Open WebUI release v0.11.4, published on September 21, brings two direct RAG improvements: metadata attached to uploaded files now travels with each chunk and arrives with the retrieved sources, and hybrid search now runs natively in the database (pgvector) instead of loading the whole collection into memory, making large knowledge bases dramatically faster. The slim image also drops to around 175 MB.
Mistral and Mozilla bring open, private and multilingual AI to the browser with Firefox Smart Window
September 22, 2026Mozilla and Mistral announced a partnership integrating Mistral's AI — open, multilingual models with a privacy focus — into Firefox Smart Window, the browser's AI panel. A signal in the direction of today's article: bringing AI into everyday work tools, with data staying under the user's control as much as possible.
Open WebUI v0.11.1: tool approvals, ask_user, and faster streams
September 18, 2026Open WebUI v0.11.1 adds tool-call approvals (a pause button between the model and its tools), a built-in ask_user function that lets the model pause and ask you a question mid-answer, a terminal file browser with in-place document previews, and Delta streaming that cuts reply traffic by roughly 99%. The full release includes more than ninety additions and over two hundred fixes.