Run open models for chat, images, speech and video on your own machines. The API speaks the OpenAI and Anthropic formats. Your team gets chats, files and documents on top, and the assistant can write and run code. 5.0 adds a desktop app, Projects, a Files page, Office documents, a Python sandbox, team sharing and an add-on library, all built on chats that now run on the server.
Highlights:
- 🖥️ Desktop app: a new app for macOS (Apple silicon) and Linux (x86-64 and ARM64) installs, updates, starts and stops LocalAI, or connects to a LocalAI server somewhere else.
- 🗂️ Projects: group chats around one folder, with notes the chats read and that you confirm before they are saved.
- 💬 Chats on the server: a chat keeps working when you close the tab. Steer it while it works, hand work to sub-agents, set a
/goal, or talk to it hands-free with Live. - 📁 Files: upload, preview, search, extract archives, Trash, and a History tab that undoes any change, yours or the model's.
- 📄 Office documents from a chat: the Simple Office add-on reads, edits and creates Word, Excel and PowerPoint files, and reads and creates PDF.
- 🐍 Sandbox: run Python and shell commands on your files in an isolated runtime with no network by default. Linux only in this release.
- 👥 Team and sharing: accounts and roles, plus sharing of folders, saved agents, Knowledge collections and connections with one person or with everyone.
- 🧩 Add-ons: a library of separate programs that add tools to a chat, from Slack and Discord bots to Mail, Google Drive, GitHub and Home Assistant.
- 🔌 Remotes: connect another LocalAI instance, or an OpenAI or Anthropic compatible endpoint, and use its models like your own.
- 🎨 New look and a better first run: a new theme, logo and mascot, and Home recommends a model for your hardware.
📊 This release in numbers
| Pull requests merged | 130 |
| Commits | 1,432 (non-merge), most of them direct commits |
| Files changed | 7,310 |
| Feature PRs | 41 |
| User-facing fix PRs | 35 |
| Model library | 2,212 to 2,325 entries (+134 added, 21 removed) |
| Backends | more than 80 |
Counts compare v4.12.1 with the release-preparation snapshot. Most of 5.0 landed as direct commits, so the sections below cite commits and docs pages as well as PRs.
📌 TL;DR
| Area | Summary |
|---|---|
| 🖥️ Desktop app | Electron app for macOS (Apple silicon) and Linux AppImages. It replaces the older launcher and reads the same ~/.localai folder.
|
| 🗂️ Projects | Chats grouped around one folder, with a memory/ folder of notes, defaults for new chats, and agents made from a project.
|
| 💬 Chats | Chats run on the server, with steering, a message queue, sub-agents, compaction, goals, slash commands and Live voice. /v1/chat/completions is unchanged.
|
| 📁 Files | Files page with folders, upload, preview, search inside documents, archive extraction, Trash and undoable history. |
| 📄 Office | Simple Office add-on: 11 tools for Word, Excel, PowerPoint and PDF, and PowerPoint decks built on your own template. |
| 🐍 Sandbox | bash tool in a KVM or gVisor sandbox with your files mounted. Each command is one undoable change. Linux only.
|
| 👥 Team | Roles, per-user features, and one Share dialog for folders, agents, collections and connections. |
| 📑 Pages | The assistant writes small HTML pages that keep their data in your files and open beside the chat. |
| 🧩 Add-ons | Add-on library with install by tag, upgrade review, one Updates page, OAuth login for add-ons, and Slack and Discord chat apps. |
| 🔎 Models | Remotes, model search in Explore, the model library, an experimental Strata backend, LFM3-Audio speech-to-speech and a forecast API. |
| 🎨 UI | New theme, logo and mascot, a first-run model recommendation, a redesigned Nodes page. |
⚠️ Upgrade notes
5.0 replaces the chat page and the agent runtime, and moves Studio and Talk to add-ons. Read this before you upgrade a server that other people use.
- Chats replace the old chat page, agents, agent jobs and skills. On the first start the server converts what the old features left on disk and in the database into saved agents, triggers, skills and chats. It does not change or delete the old data. Converted triggers, MCP connectors and connector bindings are switched off until you turn them on. Conversations kept only in a browser were never on the server: upload them with
POST /api/chats/import/browser.GET /api/chats/conversionreturns the report, and each dropped setting is named in it. See Chats: moving from the old chat, agents and skills. #12918 - Removed APIs.
/api/agents(agents, actions, SSE),/api/agents/skills,/api/agents/git-reposand/api/conversationsanswer404. Use/api/chats/agents,/api/chatsand/api/chats/skills. The collections API moved to/api/collections; the old path still works with aDeprecation: trueheader. The commandlocal-ai agentis gone. - Renamed and ignored options. The
LOCALAI_AGENT_POOL_*collection settings are nowLOCALAI_COLLECTIONS_*, andLOCALAI_DISABLE_AGENTSis split intoLOCALAI_DISABLE_CHATSandLOCALAI_DISABLE_COLLECTIONS. The old names still work and log a deprecation warning. Some agent pool options are now ignored; the migration page lists them. - Studio and Talk are add-ons. Nothing is preinstalled. An administrator installs them from Add-ons, and users need the
addonsfeature, which the upgrade does not grant. See Upgrading: Studio and Talk are add-ons now. - Permissions. On a server with accounts, saved agents, triggers, connectors, pages and add-ons are off by default for new users until an administrator grants them. The
skillsandchat_historyfeatures are gone from the user permissions. liquid-audiois removed. Configs withbackend: liquid-audiostop working. Use anlfm3-audio-2bentry onllama-cppwithpipeline.any_to_any: true. LFM2-Audio and LFM2.5-Audio have no replacement.llama-cppmodels withrealtime_audiomust now setpipeline.any_to_any: true. #12958- Simple Office creation tools take Markdown or CSV. The old
blocks(DOCX, PDF) andsheets(XLSX) arguments are rejected with migration guidance. #13029 - The desktop app replaces the older launcher. The older launcher does not update itself to the new app: download the new file once. Your models, backends and settings in
~/.localaistay. See Move from the older launcher. Commits 0a72543, 822a7bc. - Models and backends come from the libraries. Release builds read the model library (
official-models) and the backend library by default.gallery/index.yamlis now a generated copy, kept for older releases. See Model library. - Add-ons follow the release. A release build reads the add-on index attached to its own release. A source build reads the development channel.
📦 Downloads
| Platform | File |
|---|---|
| macOS, Apple silicon (desktop app) | LocalAI.dmg
|
| Linux x86-64 (desktop app) | LocalAI-Launcher-x86_64.AppImage
|
| Linux ARM64 (desktop app) | LocalAI-Launcher-arm64.AppImage
|
| Linux x86-64 (server binary) | local-ai-v5.0.0-linux-amd64
|
| Linux ARM64 (server binary) | local-ai-v5.0.0-linux-arm64
|
| macOS, Apple silicon (server binary) | local-ai-v5.0.0-darwin-arm64
|
Each file has a .sha256 file next to it. There is no macOS build for Intel: use a container there. The AppImage needs FUSE 2; on Ubuntu 24.04 and later install libfuse2t64. See the desktop app guide.
To install a server instead:
curl -sSL https://localai.io/install.sh | sh
or with a container:
docker run -p 8080:8080 --name local-ai -ti localai/localai:latest
🚀 New Features & Major Enhancements
🖥️ The desktop app
LocalAI has a desktop app for macOS and Linux, built on Electron. It installs, updates, starts and stops the LocalAI server on your computer, and shows the web interface in its own window with a status bar for the server. On the first start you choose to run LocalAI on this computer or to connect to a server that runs somewhere else, and you can keep several servers. It has pages for logs, settings, and a Connect page that shows how other devices and apps reach the server.
The DMG and the app are signed with an Apple Developer ID and notarized. On Ubuntu 23.10 and later, the AppImage starts with the Chromium sandbox off unless user namespaces are allowed; the status bar says so, and the app can open the web interface in your browser instead. The app replaces the older Go launcher.
On first start, the desktop app asks where LocalAI should run: on this computer or on a LocalAI server.
The Models page lists the models that fit this computer, with the best fit first.
🔗 Commits: Electron launcher, remove the Go launcher, DMG build. Docs: Desktop app.
🗂️ Projects
A project groups chats around one folder in your own files. The folder holds a files/ folder for the material and a memory/ folder of notes that the project's chats read at each turn. The model can propose a note, and you confirm it before it is saved. The Tools tab sets defaults for new chats: skills, add-ons, MCP servers, connectors and collections. A chat has one home project and can read up to eight others. You can make a saved agent from a project and share the project folder.
A project organises chats. It is not an access boundary: a chat outside the project can still read the project's files from your home folder.
A project page: recent chats, the project memory and the tabs for files, agents and tool defaults.
🔗 Commit: add Projects. Docs: Projects.
💬 Chats that run on the server
Chats are now kept and run by the server, so a chat keeps working when you close the tab. A message you send while the model works waits in a queue, can be edited or discarded, and reaches the model at its next step. The model can start sub-agents in the foreground or the background (up to four in the background), and long chats are compacted. The OpenAI-compatible /v1/chat/completions endpoint is unchanged.
- Goals.
/goal <text>keeps a chat working toward an objective over many turns. It pauses on Stop and at its limits: 20 turns and 400 steps by default (chats_goal_max_turns,chats_goal_max_steps). - Slash commands.
/compact,/clear,/loop,/skill, and your own commands. - Live. Talk to a chat hands-free from the chat or from Home. Live needs a microphone, a speech-to-text model and a voice model; an administrator installs small ones with one click from the setup sheet. Your audio stays on your LocalAI server.
- Memory and change proposals. Every chat can read the index of your notes with
memory_read. A model asks to change a note or an agent profile withpropose_change, and nothing is written until you confirm. A confirmed change can be undone. - MCP connector tools behind the tool search. With
chats_tool_searchon, MCP connector tools are found on demand instead of being sent on every turn. - Send a message from a chat. With
send_message, a chat and its loops can post to an approved channel of one of your messaging connections, for example to report a result. One chat sends at most 5 messages in 10 minutes. - Context size for the chat agent. For a chat model with no
context_size, the loader aims at 32,768 tokens and steps down to what fits, never below 16,384 while that fits. A banner says when a model's window is too small.
Create a saved agent in four steps: purpose, abilities, where it works, and a review.
🔗 PRs: #12918, #13064, #13066, #13069, #13163, #13153, #13199. Commits: Live, Live from Home, user slash commands, send_message. Docs: Goals, Live.
📁 Files
The new Files page holds your files. Upload files and folders, drag rows between folders, preview a document and download it. Search covers names and the text inside PDF, Word, Excel and PowerPoint files. Extract here unpacks ZIP and TAR archives; 7z and RAR stay off until an administrator turns them on. Deleting moves a file to the Trash, and the History tab undoes a change.
A chat uses the same files. Attach a file to a message, or let the model list, read, search, write and edit your files. What the model changes goes into the same history, so you can undo it. A saved agent can be limited to chosen folders, each as read or write. Distributed deployments without S3 can keep the files on a shared filesystem.
Every change is listed in History with an Undo, including the commands a chat ran in the Sandbox.
🔗 PRs: #13009, #13012. Commits: extract an archive, managed folders. Docs: Files and shared folders.
📄 Office documents from a chat
The Simple Office add-on has 11 tools: read, edit and create for Word, Excel and PowerPoint, and read and create for PDF. Creation takes Markdown (Word, PDF) or CSV (Excel). A PowerPoint deck can be built on your own .pptx or .potx template and keeps its masters, layouts, theme colours and fonts. When a file name is taken, the new file is saved next to the old one. Limits: no PDF editing, the PDF writer takes Latin-1 text only, and layouts are basic. For charts or exact layout, use the Sandbox.
Preview a spreadsheet in Files, next to the Word, PDF and PowerPoint files of the same plan.
🐍 Sandbox: run Python on your files
The Sandbox add-on gives a chat a bash tool that runs in an isolated runtime with your files mounted. Ask for a chart from a CSV file, and the model runs a Python script and saves the image in your files. What a command changed is saved as one change that you can undo. The add-on tries KVM (Cloud Hypervisor) first, then gVisor. The base profile has Python with pandas, matplotlib, openpyxl, python-docx, python-pptx, pypdf and Pillow; the media profile adds ffmpeg, sox and ImageMagick.
A sandbox has no network device. HTTP and HTTPS go through a proxy, and the default is Off, so pip install fails until an administrator allows package registries or public sites. A chat can ask for a wider setting, but never for more than the administrator allows. The Sandbox add-on runs on Linux (amd64 and arm64) in this release.
The chart comes back as a file in the chat, created by the assistant.
👥 Team and sharing
With accounts on, there are administrator and user roles, and each person has their own files, chats and saved agents. One Share dialog shares a folder of your home, a saved agent, a Knowledge collection or a connection with one person or with everyone. Revoking access stops the work that the new role no longer allows. A shared saved agent runs as the person who uses it, with that person's folders, memory and connections. An administrator can also make shared folders and choose their members.
Only a person can share, from the browser. No model, agent or add-on can share anything or take access away.
📑 Pages
Ask the assistant for a small HTML page, such as a task board, and it writes one that keeps its data in your files. Pages open in a dock beside the chat, and a trigger can keep a page's data current. For users on a server with accounts, pages are off until an administrator switches them on.
A task board page made by the assistant, open beside the chat.
🔗 PR: #13127. Commits: pages beside the chat. Docs: Pages and Home.
🧩 Add-ons and the add-on library
An add-on is a separate program that gives LocalAI new tools. An administrator installs it from the library, and its tools appear in the chat's Tools menu. The library includes Studio and Talk, Voice, Face, Fine-tuning, Quantization, Tables, Simple Office, Sandbox, Web, Calendar, Contacts, Mail (IMAP and SMTP), Gmail, Google Drive, GitHub, SSH, Home Assistant, Jellyfin and PiKVM.
- Chat apps. Slack and Discord add-ons let people talk to a saved agent from those apps. A binding answers nobody until you add allowed senders, and it can limit the kinds of file it reads.
- OAuth for add-ons. An add-on can ask a person to log in to an external service with OAuth 2.0; the host runs the login after an administrator approves it.
- Write modes. Mail, Gmail, Google Drive and PiKVM offer Read only, Asks you first and Allow all.
- Updates in one place. An Updates page covers backends and add-ons, with a review step before anything runs and inbox notifications.
- Install and run. Development add-ons install by tag and are verified against the resolved digest. The add-on owner can run as its own workload in a cluster (
--addons-role).
An add-on runs as the operating system user that runs LocalAI, so install only add-ons you trust. Only the Sandbox isolates what it runs.
The Tables add-on: a Library database with a Books table that you can filter, export and share.
Find add-ons in the library and install them from the card.
🔗 PRs: #12858, #12867, #12875, #12883, #12898, #12948, #13034, #13077, #13097, #13104, #13128, #13154, #13167, #13173. Docs: Add-ons.
🔌 Remotes
Connect another LocalAI instance, or an OpenAI or Anthropic compatible endpoint, from the first-run guide or from the Models page. Each model the remote offers becomes an ordinary model config, so it works in chats, the API and failover. The credential is sealed at rest and write-only in the API. The admin API is under /api/remotes.
🔎 Models, backends and engines
- Model search. The Explore tab searches Hugging Face, with the size of each file and whether it fits your machine. Gated Hugging Face repositories are not offered.
- Strata (experimental). A backend for the Strata engine runs Qwen3.8-Flash-Next and DeepSeek V4.1-Flash on supported GPUs, with gallery entries that build the engine's pack on first load. #13125, #13131, #13132
- LFM3-Audio speech-to-speech. LFM3-Audio-2B runs on
llama-cppin the Realtime API (WebSocket and WebRTC). #12958 - Forecasting. A
timestral-cppbackend and aPOST /v1/forecastendpoint return quantile forecasts for time series, on CPU. #12885 - Vision alias. An alias can name a
vision_aliasmodel that serves its requests with images. #13171 - llama.cpp. New
moe_cache_cpu_shareoption, and fixes forreasoning_budgetandmlock. #13155, #13157 - Backend operations.
--max-concurrent-backend-operationslets installs, upgrades and deletes run in parallel (default 1). #12888 - Realtime. Transcription-only sessions announce the speaker of each turn. #12871
🔗 Commits: model search API, search from Explore.
🎨 New look and first run
The web UI has a new theme, logo and mascot, in light and dark. On a fresh install, Home recommends a model for the hardware it detects, and the first-run guide asks whether your models run on this machine or on another server. The Nodes page is redesigned with a fleet card, grouped rows and an inspector. iOS shows the logo when you add LocalAI to the home screen.
Home in 4.12 on the left, and in 5.0 on the right.
🐛 Bug Fixes (recap)
fix(distributed): keep one idle replica when scaling down, and keep healthy replicas when one fails. #13201, #13233fix(chats): keep the system prompt stable across chats so the prompt cache hits, and start every shell in the chat's work folder. #13153, #13212fix(llama-cpp): fixreasoning_budgetandmlock, add missing options, harden raw flags; keep an existing image data URL. #13157, #12921fix(downloader): parse the Hugging Face@revisionsuffix without panicking. #13179fix(oci): fall back to the OS temp folder when the staging folder is not writable. #13186fix(vram): do not cache failed probes. #13203fix(realtime): never offer deprecated IPv6 addresses as WebRTC ICE candidates. #13172fix(cloud-proxy): collapse nullable type unions in OpenAI tool schemas. #13165fix(proxy): propagate model env and accept direct keys. #12940fix(trace): stop trace writes into the data path on shutdown. #13161fix(core): correct AMD discovery and MCP and P2P handling. #13044fix(strata): preserve image markers in tokenizer messages. #13254fix(mlx-vlm): bump the mlx-vlm pin to v0.6.12. #12818fix(ui): show GPU memory use on the Settings page, no "No models loaded" for non-admins, keep the Models page tabs usable. #13162, #13166, #12899fix(addons),fix(connectors),fix(studio): owner settings, upgrades, discovery and connector delivery fixes. #12850, #12878, #12943, #13026, #13030, #13033, #13035, #13062, #13078, #13127, #13129fix(kokoros): implement Forecast as unsupported. #13134fix(gallery): installing a recommended model with no variants from Home no longer fails with "pinned model variant not found". #13277fix(timestral-cpp): build on arm64 (Linux and macOS). #13276
🧠 Models
The model library grows from 2,212 to 2,325 entries: 134 added, 21 removed (the liquid-audio entries among them). Additions include the Strata GGUF sets (Qwen3.8-Flash-Next and DeepSeek V4.1-Flash), Phonon-2.5 and Fluid Speech Pro on parakeet-cpp, Flint, MiniCPM-OCR 1.0, Nanbeige4.3, Gemma 4 and Qwen3.5 Artemis variants, Ternary Bonsai 2, and BGE English embedding variants. Check each model card for its licence and hardware needs.
🔗 PRs: #12854, #12896, #12994, #13131, #13132, #13253, #13267.
👒 Dependencies
The range has 104 engine update commits, most made directly by the update bot, and 19 dependency PRs.
| Source | Updates |
|---|---|
ikawrakow/ik_llama.cpp
| 15 |
0xShug0/audio.cpp
| 15 |
ggml-org/llama.cpp
| 14 |
leejet/stable-diffusion.cpp
| 12 |
mudler/vllm.cpp
| 11 |
CrispStrobe/CrispASR
| 9 |
NVIDIA/NeMo-Speech.cpp
| 5 |
ggml-org/whisper.cpp
| 5 |
ServeurpersoCom/omnivoice.cpp
| 3 |
PrismML-Eng/llama.cpp
| 3 |
Niko1221/Strata
| 3 |
antirez/ds4
| 3 |
| Four other sources | 6 |
📖 Documentation
The docs are regrouped into nine sections with a new brand, website and README (#13178). New pages cover the desktop app, Projects, Files, chats and goals, Live, the migration from the old chat and agents, Remotes, the add-on catalogue, and connecting coding agents to the OpenAI Chat Completions, OpenAI Responses and Anthropic Messages APIs. The product lede names coding (#13275). The 5.0 blog post and the home page download buttons are in #13229 and #13271.
🐞 Reporting problems
5.0 carries a lot of new work. If something breaks after the upgrade, open an issue or ask in the Discord.
🙌 New Contributors
Thanks to the contributors making their first merged PRs:
- @genx7up in #12818.
- @ilsawa in #12921.
- @Lil-Clanky in #13162.
- @Rutvij97 in #13186.
- @zjuncal in #13203.
Thanks also to @walcz-de for the chat, connector and Simple Office work in this release.
Full Changelog: v4.12.1...v5.0.0
What's Changed
Breaking Changes 🛠
- feat(realtime): LFM3-Audio speech-to-speech on llama.cpp, remove liquid-audio by @mudler-agent in #12958
- feat(chats): server-side chats with steering, sub-agents and compaction by @mudler-agent in #12918
Bug fixes 🐛
- fix(addons): one page title, one primary action and clearer empty states by @mudler-agent in #12850
- fix(mlx-vlm): bump mlx-vlm pin to v0.6.12 by @genx7up in #12818
- fix(ui): keep backends off the add-on browse list by @mudler-agent in #12878
- fix(ui): keep the Models page fixed and give its tabs and filters room by @mudler-agent in #12899
- fix(proxy): propagate model env and accept direct keys by @mudler-agent in #12940
- fix(addons): verify the platform child of an add-on index by @mudler-agent in #12943
- [llama-cpp] keep an existing image data URL by @ilsawa in #12921
- fix(agents): allow aliases in the model picker by @mudler-agent in #13024
- fix(connectors): route calls through the owner by @mudler-agent in #13026
- fix(connectors): preserve disabled add-on discovery by @mudler-agent in #13030
- fix(addons): make owner settings authoritative by @mudler-agent in #13033
- fix(addons): run accepted upgrades on the owner by @mudler-agent in #13035
- fix(studio): tell the model of a connector chat that files are attached by @mudler-agent in #13062
- fix(core): correct AMD discovery and MCP/P2P handling by @mudler-agent in #13044
- fix(studio): always read the thread when the bot is mentioned in it by @mudler-agent in #13078
- fix(studio): keep trigger files reachable, let update_trigger change delivery and acts by @mudler-agent in #13127
- fix(connectors): link back to the agent, report what a save changes, keep non-ASCII file-change cards by @mudler-agent in #13129
- fix(kokoros): implement Forecast as unsupported by @mudler-agent in #13134
- fix(chats): keep the system prompt stable across chats so the prompt cache hits by @walcz-de in #13153
- fix(llama-cpp): fix reasoning_budget and mmlock, add missing options, harden raw flags by @mudler-agent in #13157
- fix(cloud-proxy): collapse nullable type unions in OpenAI tool schemas by @walcz-de in #13165
- [ui] fix(ui): show GPU memory use on the Settings page by @Lil-Clanky in #13162
- fix(trace): stop trace persistence writes into DataPath on Shutdown by @walcz-de in #13161
- fix(ui): no "No models loaded" for non-admins, Manage setups with an agent by @walcz-de in #13166
- fix(realtime): never offer deprecated IPv6 addresses as WebRTC ICE candidates by @walcz-de in #13172
- fix(downloader): parse the HuggingFace @revision suffix without panicking by @simpleqt in #13179
- fix(oci): fall back to OS temp when the staging dir is not writable by @Rutvij97 in #13186
- fix(distributed): keep one idle replica when scaling down by @mudler-agent in #13201
- fix(chats): start every shell in the chat's work folder and quote paths by @mudler-agent in #13212
- [vram] fix: avoid caching failed probes by @zjuncal in #13203
- fix(ui): word the first-run add-on step without add-on names by @mudler-agent in #13231
- fix(distributed): keep healthy replicas when one replica fails by @mudler-agent in #13233
- [strata] fix: preserve image markers in tokenizer messages by @leilei3167 in #13254
- fix(timestral-cpp): pin the arm64 build fix by @mudler-agent in #13276
- fix(gallery): accept the entry's own name as the pin on a single-build entry by @mudler-agent in #13277
Exciting New Features 🎉
- feat(addons): add-ons UI direction B: one name, Turn on and Turn off, preview pane by @mudler-agent in #12858
- realtime: announce speakers in transcription-only sessions by @mudler-agent in #12871
- feat(addons): log the reason behind a generic denial by @mudler-agent in #12875
- feat(addons): run the add-on owner as its own workload by @mudler-agent in #12867
- feat(addons): install development add-ons by tag, as backends do by @mudler-agent in #12883
- feat(backends): timestral.cpp forecast backend and /v1/forecast API by @mudler-agent in #12885
- feat(galleryop): limit concurrent backend operations by @mudler-agent in #12888
- feat(addons): show the tool count of each add-on before install by @mudler-agent in #12898
- feat(ui): serve an apple-touch-icon so iOS shows the logo on the home screen by @mauro-agent in #12917
- feat(ui): review add-on and backend upgrades before they run by @mudler-agent in #12948
- feat(agents): limit saved agents to one folder by @mudler-agent in #13009
- feat(workspace): use filesystem storage without S3 by @mudler-agent in #13012
- feat(addons): simplify Office and discover Tables by @mudler-agent in #13029
- feat(sandbox): own document authoring skills by @mudler-agent in #13038
- feat(addons): declare typed host capabilities by @mudler-agent in #13034
- feat(studio): load context files and add memory_read to every chat by @mudler-agent in #13064
- feat(studio): propose_change and revert_change with a proposal store by @mudler-agent in #13066
- feat(ui): proposal card, Proposals page and Memory settings for change proposals by @mudler-agent in #13069
- feat(addons): a network capability for the network-policy control by @mudler-agent in #13073
- feat(connectors): declare file limits and build the sentences from them by @mudler-agent in #13077
- feat(ui): redesign the Nodes page with a fleet card, grouped rows and an inspector by @mudler-agent in #13121
- feat(ui): list the connections of an agent on its page by @mudler-agent in #13128
- feat(strata): add an experimental backend for the Strata engine by @mudler-agent in #13125
- feat(addons): add the Mail add-on (IMAP and SMTP) by @mudler-agent in #13104
- feat(strata): add the GGUF models to the gallery and build the pack on first load by @mudler-agent in #13131
- feat(strata): run DeepSeek V4.1-Flash and add its gallery entries by @mudler-agent in #13132
- feat(sharing): share agents, connections, collections and folders between accounts by @mudler-agent in #13130
- feat(addons): generic OAuth 2.0 login for add-ons, with Google Drive and Gmail add-ons by @mudler-agent in #13097
- feat(llama-cpp): bump llama.cpp and add the moe_cache_cpu_share option by @mudler-agent in #13155
- feat(ui): one place for backend and add-on updates, with inbox notifications by @mudler-agent in #13154
- feat(chats): defer MCP connector tools behind the tool search by @walcz-de in #13163
- feat(addons): add an Allow all write mode to mail, Gmail, Drive and PiKVM by @mudler-agent in #13167
- feat(chats): let a connector binding limit the kinds of file it reads by @walcz-de in #13173
- feat(config): serve image requests to an alias on a vision_alias model by @walcz-de in #13171
- feat(ui): new LocalAI theme, logo and empty-state mascot for the web UI by @mudler-agent in #13187
- feat: recommend a model on first run and size the context for the chat agent by @mudler-agent in #13199
- feat(sandbox): add a macOS VM backend built on Virtualization.framework by @mudler-agent in #13230
- [office] feat: build a PowerPoint deck on a template by @walcz-de in #13227
- feat(remotes): connect a LocalAI instance or an OpenAI or Anthropic endpoint by @mudler-agent in #13235
🧠 Models
- feat(gallery): add flint and bump llama.cpp to a37cc461 by @mudler-agent in #12854
- feat(gallery): add Phonon-2.5 parakeet-cpp models by @mudler-agent in #12896
- fix(model-gallery): point pinned LFM3-Audio entries at main by @mudler-agent in #12994
- chore(model-gallery): ⬆️ update checksum by @localai-update-bot[bot] in #13253
- chore(model-gallery): ⬆️ update checksum by @localai-update-bot[bot] in #13267
📖 Documentation and examples
- docs: new brand, docs sections, website and README by @mudler-agent in #13178
- blog: LocalAI 5.0 launch post and film (draft, publish with the release) by @mudler-agent in #13229
- site: Download for Linux and Mac buttons on the home page by @mudler-agent in #13271
- docs: name coding in the product lede by @mudler-agent in #13275
👒 Dependencies
- chore(deps): bump actions/setup-node from 6 to 7 by @dependabot[bot] in #12837
- chore(deps): bump docker/login-action from 3 to 4 by @dependabot[bot] in #12836
- chore(deps): bump grpcio from 1.84.0 to 1.85.0 in /backend/python/rerankers by @dependabot[bot] in #12963
- chore(deps): bump docs/themes/hugo-theme-relearn from
657eb0cto53bcfeeby @dependabot[bot] in #12962 - chore(deps): bump grpcio from 1.84.0 to 1.85.0 in /backend/python/common/template by @dependabot[bot] in #12961
- chore(deps): bump torch from 2.14.0+xpu to 2.15.0+xpu in /backend/python/common/template by @dependabot[bot] in #12960
- chore: ⬆️ Update ggml-org/whisper.cpp to
5922450fbce1ed9d1648babd0d5ac4520055b436by @localai-update-bot[bot] in #13250 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
fcf7b364786ec621d0c53eae39223efc5492b8c5by @localai-update-bot[bot] in #13249 - chore: ⬆️ Update CrispStrobe/CrispASR to
9ddae4590d427f27e0ca04694864eed66bcb530eby @localai-update-bot[bot] in #13247 - chore: ⬆️ Update 0xShug0/audio.cpp to
89bccbea607d3bb35cc6aff17476ad97e472aa15by @localai-update-bot[bot] in #13246 - chore: ⬆️ Update ServeurpersoCom/omnivoice.cpp to
a5bd24f6b675335d1c9c0e945793d0de9fabe9a4by @localai-update-bot[bot] in #13251 - chore: ⬆️ Update ggml-org/llama.cpp to
fc504f91d8cfdfcd9b0fedcf14600fd2b0e63420by @localai-update-bot[bot] in #13245 - chore: ⬆️ Update ggml-org/whisper.cpp to
aba02defd6bb6ff5de63721c21e3456645c153e2by @localai-update-bot[bot] in #13266 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
a6383da414f74c944999e7b9312a5a5d76191f6bby @localai-update-bot[bot] in #13262 - chore: ⬆️ Update 0xShug0/audio.cpp to
2872e36e992c8dd98fcc47de094bd703e1ff6dabby @localai-update-bot[bot] in #13264 - chore: ⬆️ Update ggml-org/llama.cpp to
d998c676ff463b1b8ad94ad7c0b905e281422fadby @localai-update-bot[bot] in #13261 - chore: ⬆️ Update Niko1221/Strata to
v0.2.15 (c77572806996349b49995e4f98d41b5294d36a21)by @localai-update-bot[bot] in #13265 - chore: ⬆️ Update ServeurpersoCom/omnivoice.cpp to
a5bd24f6b675335d1c9c0e945793d0de9fabe9a4by @localai-update-bot[bot] in #13263 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
532d2c15c430a785a712b2d94281f04e733008a4by @localai-update-bot[bot] in #13248
Other Changes
- chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12817
- chore: ⬆️ Update CrispStrobe/CrispASR to
95e454b493bf6d1ea7171cb154eaace5c44c9ae3by @localai-org-maint-bot in #12815 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
a77d3623180b2bc1886f3dd04c0aa04156a987efby @localai-org-maint-bot in #12814 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
15e7ecacddd87455f3fbc4805bc0508df01fa8ccby @localai-org-maint-bot in #12813 - feat(i18n): add French (fr) locale by @gangna2s-arch in #12826
- chore: ⬆️ Update mudler/vllm.cpp to
8af8a3e907330847d35cf223d93c2ba66a1c9389by @localai-org-maint-bot in #12810 - chore: ⬆️ Update vllm-metal (darwin) to
v0.32.0by @localai-org-maint-bot in #12811 - chore: ⬆️ Update 0xShug0/audio.cpp to
41bb3f0ffb6b46b7a3bb3c02fcea52900d88001bby @localai-org-maint-bot in #12812 - feat(addons): installable add-ons foundation and the Tables add-on by @mudler-agent in #12702
- feat(backends): --disable-backend-install disables backend downloads. by @blackd in #12754
- feat(addons): list and install backends from a library by @mudler-agent in #12816
- [httpclient] validate redirect targets and pin dialed addresses for user URLs by @mudler-agent in #12853
- refactor(ui): remove the built-in copies of the add-on pages by @mudler-agent in #12852
- ci(gallery): have the index bots write the model library by @mudler-agent in #12849
- chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12846
- chore: ⬆️ Update 0xShug0/audio.cpp to
80e66d2cfedc9c113c7c8034ff4030afd43ea33fby @localai-org-maint-bot in #12843 - docs: ⬆️ update docs version mudler/LocalAI by @localai-org-maint-bot in #12839
- fix(ci): make master green after the library merge by @mudler-agent in #12855
- chore: ⬆️ Update mudler/vllm.cpp to
c9e6bb7b3ae7f7676ea35f570d5a54af12a6a73bby @localai-org-maint-bot in #12842 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
c12b5ac33be74ea4fd09a5beed371725ddf2afe1by @localai-org-maint-bot in #12841 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12874
- chore: ⬆️ Update ikawrakow/ik_llama.cpp to
804ee29c28a5e892df6e84941cb78628ed4c266cby @localai-org-maint-bot in #12866 - fix(ci): repair master failures in CLI bind test, addons UI build and e2e by @mudler-agent in #12876
- test(nodes): give the DistributedBackendManager fixtures a system state by @mudler-agent in #12877
- fix(ci): give tables a platform so add-on packages publish by @mudler-agent in #12879
- ci(addons): publish every add-on to one repository by @mudler-agent in #12881
- test(addons): refuse an unlisted platform in the packaging contract by @mudler-agent in #12884
- test(ui): stub the pin routes in the add-on e2e fixtures by @mudler-agent in #12882
- chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12886
- feat(gallery): add Nemotron Bolt E4B variants by @localai-org-maint-bot in #12880
- ci(addons): run the add-on packages workflow on packaging test changes by @mudler-agent in #12889
- docs: ⬆️ update docs version mudler/LocalAI by @localai-org-maint-bot in #12865
- chore: ⬆️ Update antirez/ds4 to
ee7e8405232c8971000cc7ba2b945b00bf6d20d3by @localai-org-maint-bot in #12870 - chore: ⬆️ Update 0xShug0/audio.cpp to
dd92e32b6e268178ff43f4fa3ca36f0b380491e0by @localai-org-maint-bot in #12869 - chore: ⬆️ Update CrispStrobe/CrispASR to
c08d243de93dc7c5a08d39cf158ebdaa9c3503adby @localai-org-maint-bot in #12844 - chore: ⬆️ Update mudler/vllm.cpp to
2b789b86eb46ced8e2ac4c425ff9597c04bb5348by @localai-org-maint-bot in #12868 - chore: ⬆️ Update PrismML-Eng/llama.cpp to
97ea5aeee195b7c2a8fc48668c37a17c6ecc92ebby @localai-org-maint-bot in #12792 - test(addons): accept either refusal of an unlisted platform in the packaging contract by @mudler-agent in #12894
- test(ui): turn the feature switches on in the navigation specs by @mudler-agent in #12895
- chore: ⬆️ Update ggml-org/llama.cpp to
78a8ae14be559b45f5f13ddee82c188b683d5d8dby @localai-org-maint-bot in #12741 - chore: ⬆️ Update TheTom/llama-cpp-turboquant to
643956891dfdf51aecdbd05ca508dd21e21e8195by @localai-org-maint-bot in #12840 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
a85db96aa5ce2c72bd5aec326f6c5c857839a50aby @localai-org-maint-bot in #12845 - test(ui): derive the settings field count from the schema by @mudler-agent in #12901
- chore: ⬆️ Update CrispStrobe/CrispASR to
119c4031795a005eea4f9a9339351d4d77b5dc95by @localai-org-maint-bot in #12910 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12911
- chore: ⬆️ Update ikawrakow/ik_llama.cpp to
e34d163ad004f0830013240f8c750574425afc71by @localai-org-maint-bot in #12905 - chore(website): refresh the counters by @localai-org-maint-bot in #12915
- chore: ⬆️ Update OpenNMT/CTranslate2 ROCm wheel to
v4.8.4by @localai-org-maint-bot in #12903 - ci(addons): log in to quay.io with the existing registry credentials by @mudler-agent in #12916
- chore: ⬆️ Update PrismML-Eng/llama.cpp to
c96baf4abe24d11a50f326e40b6223b687449d63by @localai-org-maint-bot in #12906 - fix(gallery): correct HTDemucs checksum by @localai-org-maint-bot in #12919
- chore: ⬆️ Update mudler/vllm.cpp to
3090f069bda47a09e6ef809a86b0c8b61a4f21c6by @localai-org-maint-bot in #12909 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
e41fbcf455e0fbb45e08bd535b00d6cba865e921by @localai-org-maint-bot in #12907 - chore: ⬆️ Update 0xShug0/audio.cpp to
082177ecac45f1965929ded33cd81a85c13542f1by @localai-org-maint-bot in #12904 - chore: ⬆️ Update vllm-project/vllm cu130 wheel to
0.32.0by @localai-org-maint-bot in #12760 - chore: ⬆️ Update ggml-org/llama.cpp to
6050b467c07a95c7b87debd4ddac2f8ba8d9b807by @localai-org-maint-bot in #12908 - chore: ⬆️ Update NVIDIA/NeMo-Speech.cpp to
e680fc30ac7404ba3acd58532f713becf8625894by @localai-org-maint-bot in #12417 - ci(addons): use existing registry credentials by @mudler-agent in #12929
- chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12936
- chore: ⬆️ Update mudler/vllm.cpp to
8189b3f544b7f54723baefc2e289f9e054ea245bby @localai-org-maint-bot in #12931 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
a61ad7873c8306fe18123eea9659705cb5cf945dby @localai-org-maint-bot in #12934 - chore: ⬆️ Update ggml-org/llama.cpp to
63b4b303bd87456adc2f09d56c7e1f74e9e26673by @localai-org-maint-bot in #12932 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
be486498eadc29dc3cd234d944c2180ed42b1415by @localai-org-maint-bot in #12935 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12946
- chore: ⬆️ Update ggml-org/llama.cpp to
338fce6511c744893cc0d3d958b275f40552c7a4by @localai-org-maint-bot in #12944 - chore: ⬆️ Update 0xShug0/audio.cpp to
1b8c8122c072293c435ade8dc991405286ec8856by @localai-org-maint-bot in #12930 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12955
- chore: ⬆️ Update ikawrakow/ik_llama.cpp to
1c5ea124098890184b297023cd9fc9655451b1ecby @localai-org-maint-bot in #12954 - chore: ⬆️ Update mudler/vllm.cpp to
3284412c0c45976aaf661115928fcf1c64b83cabby @localai-org-maint-bot in #12952 - chore: ⬆️ Update CrispStrobe/CrispASR to
6f7e812e5a7e05ab56300be8cf96abdd6a25d4a0by @localai-org-maint-bot in #12933 - chore: ⬆️ Update localai-org/kimodo.cpp to
75ed52d2d3ec767a9bac79d11060b65d99f61a89by @localai-org-maint-bot in #12973 - chore: ⬆️ Update mudler/parakeet.cpp to
4fbf4d92b49a44bc4a24bc373ca12cf252919ca6by @localai-org-maint-bot in #12970 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12977
- fix(ui): use token usage for chat throughput by @localai-org-maint-bot in #12923
- chore: ⬆️ Update CrispStrobe/CrispASR to
ecc5a8e7c97aca0835dc4e2864873a68bfc1ab67by @localai-org-maint-bot in #12974 - chore: ⬆️ Update mudler/vllm.cpp to
3c69df47a4fc450ea507b29f2c591b60458c0866by @localai-org-maint-bot in #12971 - chore: ⬆️ Update NVIDIA/NeMo-Speech.cpp to
c540e21167017d2e34444f2fc2d70649065dc500by @localai-org-maint-bot in #12953 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
e239e4b0fd3de49834cec67031a44796ca465abcby @localai-org-maint-bot in #12945 - chore: ⬆️ Update 0xShug0/audio.cpp to
aeb7bcf1a44c2ebe3308e9f206a9225a69740a92by @localai-org-maint-bot in #12951 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
06d729c5aff5aad77c408ad97efb2048475414c4by @localai-org-maint-bot in #12976 - chore: ⬆️ Update PrismML-Eng/llama.cpp to
9d9efeadbdd1d77458d98717a96fbfa3a016e159by @localai-org-maint-bot in #12972 - chore: ⬆️ Update localai-org/privacy-filter.cpp to
3b87bf8793f440a20fe8b2dcd00f6f27e3bfd69dby @localai-org-maint-bot in #12975 - chore: ⬆️ Update NVIDIA/NeMo-Speech.cpp to
20999cc921d2b14aa58708f61756a8b2e787ba20by @localai-org-maint-bot in #12982 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
13cffc006b82e03e4d89cbb8c7327f0255ae3d18by @localai-org-maint-bot in #12984 - chore: ⬆️ Update mudler/vllm.cpp to
8d1eb0daac798d28e08e4b53af5ece6119d16816by @localai-org-maint-bot in #12983 - chore: ⬆️ Update mudler/parakeet.cpp to
2eac2e77d9c7ef8899f7a86a46e693ac5ac1a5cdby @localai-org-maint-bot in #12985 - chore: ⬆️ Update antirez/ds4 to
c97964b8aa77034a8e17afb10ca7a686c40ab511by @localai-org-maint-bot in #12988 - chore: ⬆️ Update 0xShug0/audio.cpp to
350670db7ff64864e5c66c794e6c4b0d3dc98d2fby @localai-org-maint-bot in #12987 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #12989
- chore: ⬆️ Update CrispStrobe/CrispASR to
3345aab0a87241e830780ca2d7d6a729276fa5a7by @localai-org-maint-bot in #12986 - chore: ⬆️ Update ggml-org/whisper.cpp to
482279b8a18855ba741d7b88c30fcb3874691f59by @localai-org-maint-bot in #12950 - chore: ⬆️ Update ggml-org/llama.cpp to
5047748f114564fcf271e6a648f60e2d2eae6264by @localai-org-maint-bot in #12949 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13005
- chore: ⬆️ Update 0xShug0/audio.cpp to
ad14f0c4dbd5b12b7711bb7b46abb8d94f78f067by @localai-org-maint-bot in #13004 - chore: ⬆️ Update CrispStrobe/CrispASR to
1dd0bf6fb9e7a8bf5849cb7a14605b2234e6ee93by @localai-org-maint-bot in #13003 - chore: ⬆️ Update mudler/vllm.cpp to
fb5bfcd70ddd2717cb22908527cd503f8b007143by @localai-org-maint-bot in #13001 - chore: ⬆️ Update ggml-org/llama.cpp to
ddf48f85c933c1f438212a02eb21e7b001d2d37dby @localai-org-maint-bot in #13000 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
c6dfa9def12fd0cb36ac7149ba30a75792bedbf0by @localai-org-maint-bot in #13002 - fix(ci): publish images after large mixed diffs by @mudler-agent in #13011
- chore(website): refresh the counters by @localai-org-maint-bot in #13028
- chore: ⬆️ Update leejet/stable-diffusion.cpp to
290148d2e31ca6ef9757d567905776b5f3646db5by @localai-org-maint-bot in #13018 - chore: ⬆️ Update ggml-org/llama.cpp to
785322d17f2d51e5185cf996a1f4e46d87640841by @localai-org-maint-bot in #13019 - chore: ⬆️ Update mudler/vllm.cpp to
a087e3bf7fd61e838e848ff44dec44f8267751caby @localai-org-maint-bot in #13020 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13021
- fix(ci): use native workflow path filters by @mudler-agent in #13037
- test: update stale tests after the add-on owner and Sandbox skill changes by @mudler-agent in #13045
- fix(ci): repair the failing master checks by @mudler-agent in #13051
- test(studio): give the fence probe turns a bound that fits a slow runner by @mudler-agent in #13056
- refactor(studio): replace the Sandbox special cases with an execution descriptor by @mudler-agent in #13068
- chore: ⬆️ Update OpenNMT/CTranslate2 ROCm wheel to
v4.8.5by @localai-org-maint-bot in #13083 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
656036d98503e154952f6c633cb3febc1a4f08b4by @localai-org-maint-bot in #13089 - test(addons): cover the setup and ownerData fields in the apps spec by @mudler-agent in #13133
- chore: ⬆️ Update 0xShug0/audio.cpp to
8fe585c437c10d61f7bcf4cc21a584508546a41bby @localai-org-maint-bot in #13090 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13093
- chore: ⬆️ Update ggml-org/llama.cpp to
669204ee2b9004845b34f26466a91d003edc98aaby @localai-org-maint-bot in #13088 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
f5df5f8b84f139f1e80496902da90798a0d92896by @localai-org-maint-bot in #13091 - chore: ⬆️ Update NVIDIA/NeMo-Speech.cpp to
536b17a41bef62d2a67a40ed3df61df26a4389feby @localai-org-maint-bot in #13087 - chore: ⬆️ Update antirez/ds4 to
b3f50180ec2c7324892be1f50bc83555ccb8d5e7by @localai-org-maint-bot in #13086 - chore: ⬆️ Update ggml-org/whisper.cpp to
b1535f5fb163f4443804888f7be980621226cb15by @localai-org-maint-bot in #13085 - chore: ⬆️ Update CrispStrobe/CrispASR to
de3ff9cbc6ac0429f057b6598020cf75dde338d8by @localai-org-maint-bot in #13084 - chore: ⬆️ Update mudler/vllm.cpp to
52c291adf170c4b761c7e3f973ad7931305beef0by @localai-org-maint-bot in #13081 - chore: ⬆️ Update Niko1221/Strata to
v0.2.12 (146df3c468739147f520ac3816de806f71232571)by @localai-org-maint-bot in #13138 - fix(ci): get master green again (providerguard, UI e2e runner and spec) by @mudler-agent in #13148
- chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13147
- chore(website): refresh the counters by @localai-org-maint-bot in #13160
- chore: ⬆️ Update 0xShug0/audio.cpp to
f2520dadd6f404d9e9a799d7b9757bcd45958c84by @localai-org-maint-bot in #13139 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13177
- chore: ⬆️ Update ggml-org/llama.cpp to
0a94a3f289687883617b9a2abbc7e0ade2d6b6c4by @localai-org-maint-bot in #13144 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
308fb08df6fb3fa65943d6b0feed7c87d6cc2d07by @localai-org-maint-bot in #13141 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
4e24d0b2122fd34e5817978b00ddefde314df5d9by @localai-org-maint-bot in #13183 - chore: ⬆️ Update OpenNMT/CTranslate2 ROCm wheel to
v4.9.0by @localai-org-maint-bot in #13182 - chore: ⬆️ Update Niko1221/Strata to
v0.2.13 (d2f096475817d4666c364c820b0ca830f751cfc5)by @localai-org-maint-bot in #13175 - chore: ⬆️ Update ggml-org/llama.cpp to
7c2653acc749258dc62067da62d07024d971ccd2by @localai-org-maint-bot in #13184 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
32d516c0197ed5ca31305b8575e14eb1451fad33by @localai-org-maint-bot in #13185 - chore: ⬆️ Update 0xShug0/audio.cpp to
de6fbe3b0e004de79b9ca642e033a35642b56c02by @localai-org-maint-bot in #13176 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13188
- chore: ⬆️ Update ServeurpersoCom/omnivoice.cpp to
dbfc792d6fbe0cc3a776817f56de35a3744d7594by @localai-org-maint-bot in #13142 - chore: ⬆️ Update ggml-org/whisper.cpp to
d2c6f05a46dacf663e357bcf0ed40f542a72825dby @localai-org-maint-bot in #13140 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
883789c25ba93e625bb5266f41f9b73150cf6fa6by @localai-org-maint-bot in #13193 - chore: ⬆️ Update 0xShug0/audio.cpp to
be2cc80ae9f5a1432d15aa53107133e495b46aa8by @localai-org-maint-bot in #13196 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13197
- chore: ⬆️ Update TheTom/llama-cpp-turboquant to
4185b90a0926e77ac72ed6248b0ac3e1812c55b8by @localai-org-maint-bot in #13194 - chore: ⬆️ Update ggml-org/llama.cpp to
4180722eeb3b0335700edb0b1e195d5ca05d0191by @localai-org-maint-bot in #13195 - chore: ⬆️ Update ikawrakow/ik_llama.cpp to
9513b3d44f480fd83fe26fa55eaad77f582019baby @localai-org-maint-bot in #13206 - chore: ⬆️ Update NVIDIA/NeMo-Speech.cpp to
e790e463b2913cff9f989a3df224c587f14b4c5cby @localai-org-maint-bot in #13210 - chore: ⬆️ Update leejet/stable-diffusion.cpp to
b9d7176bf88fe465576e0aafbe47ba253ffcfc2cby @localai-org-maint-bot in #13208 - chore: ⬆️ Update 0xShug0/audio.cpp to
a292a9e36e6dea2b22a788c87affbd604646a6d1by @localai-org-maint-bot in #13209 - chore(model-gallery): ⬆️ update checksum by @localai-org-maint-bot in #13211
- chore: ⬆️ Update ggml-org/llama.cpp to
8ca5a8008b57aa39f2601546c0d07912045080a5by @localai-org-maint-bot in #13207 - ci: open bot pull requests with a GitHub App token by @mudler-agent in #13234
- ci: skip heavy runs for changes that cannot affect them by @mudler-agent in #13238
- feat(swagger): update swagger by @localai-update-bot[bot] in #13240
- chore(website): refresh the counters by @localai-update-bot[bot] in #13257
- Retire models.localai.io: remove the links and the workflow that announced to it by @mudler-agent in #13272
- feat(website): serve the new feature films on the feature pages by @mudler-agent in #13279
- feat(website): download the desktop app for the visitor's system by @mudler-agent in #13281
- fix(oci): swap reversed arguments in image path Contains check by @IgorPolyakov in #13280
- fix(studio): say where a SKILL.md belongs when a chat writes it outside a skills folder by @walcz-de in #13285
- chore(model-gallery): ⬆️ update checksum by @localai-update-bot[bot] in #13289
New Contributors
- @gangna2s-arch made their first contribution in #12826
- @genx7up made their first contribution in #12818
- @mauro-agent made their first contribution in #12917
- @ilsawa made their first contribution in #12921
- @Lil-Clanky made their first contribution in #13162
- @Rutvij97 made their first contribution in #13186
- @zjuncal made their first contribution in #13203
- @localai-update-bot[bot] made their first contribution in #13240
- @IgorPolyakov made their first contribution in #13280
Full Changelog: v4.12.0...v5.0.0
