{"id":44324570,"url":"https://github.com/scouzi1966/vesta-mac-dist","last_synced_at":"2026-05-09T05:01:52.614Z","repository":{"id":313835061,"uuid":"1053047178","full_name":"scouzi1966/vesta-mac-dist","owner":"scouzi1966","description":"Vesta macOS Distribution - Official releases and downloads.Vesta AI Chat Assistant for macOS - Built with SwiftUI, Swift MLX  and Apple Intelligence using Apple's on device model on MacOs Tahoe (MacOS 26). Now with side-by-side Qwen3-VL for vison","archived":false,"fork":false,"pushed_at":"2026-05-09T03:34:49.000Z","size":9303,"stargazers_count":80,"open_issues_count":4,"forks_count":8,"subscribers_count":3,"default_branch":"main","last_synced_at":"2026-05-09T04:07:26.922Z","etag":null,"topics":["ai","apple","apple-llm","apple-llm-integration","applefoundationalmodels","appleintelligence","chat-application","finetuning-llms","llamacpp","localllm","lora","macos","mcp","mcp-server","mlx","neural-networks","privateai","qwen3-vl","tahoe"],"latest_commit_sha":null,"homepage":"","language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/scouzi1966.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-09-08T23:17:56.000Z","updated_at":"2026-05-09T02:00:59.000Z","dependencies_parsed_at":"2026-05-09T04:03:29.483Z","dependency_job_id":null,"html_url":"https://github.com/scouzi1966/vesta-mac-dist","commit_stats":null,"previous_names":["scouzi1966/vesta-mac-dist"],"tags_count":19,"template":false,"template_full_name":null,"purl":"pkg:github/scouzi1966/vesta-mac-dist","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scouzi1966%2Fvesta-mac-dist","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scouzi1966%2Fvesta-mac-dist/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scouzi1966%2Fvesta-mac-dist/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scouzi1966%2Fvesta-mac-dist/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/scouzi1966","download_url":"https://codeload.github.com/scouzi1966/vesta-mac-dist/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/scouzi1966%2Fvesta-mac-dist/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":32807861,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-08T08:22:46.396Z","status":"online","status_checked_at":"2026-05-09T02:00:06.633Z","response_time":123,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","apple","apple-llm","apple-llm-integration","applefoundationalmodels","appleintelligence","chat-application","finetuning-llms","llamacpp","localllm","lora","macos","mcp","mcp-server","mlx","neural-networks","privateai","qwen3-vl","tahoe"],"created_at":"2026-02-11T07:16:24.614Z","updated_at":"2026-05-09T05:01:52.608Z","avatar_url":"https://github.com/scouzi1966.png","language":null,"funding_links":[],"categories":["Rising projects"],"sub_categories":[],"readme":"# Vesta AI Explorer for macOS - macOS Tahoe 26\n\n### [vesta-mac.pages.dev](https://vesta-mac.pages.dev)\n\n## Install with Homebrew (recommended)\n\n```bash\nbrew tap scouzi1966/afm\nbrew install --cask scouzi1966/afm/vesta-mac\n```\n\n## Direct Download\n\n| Channel | Version | Download | Date |\n|---------|---------|----------|------|\n| **Stable** | 0.9.7 | [Vesta-0.9.7.dmg](https://github.com/scouzi1966/vesta-mac-dist/releases/download/v0.9.7/Vesta-0.9.7.dmg) | 2026-05-08 |\n| Nightly | Vesta-next | [Vesta-next.dmg](https://github.com/scouzi1966/vesta-mac-dist/releases/download/Vesta-next/Vesta-next.dmg) | Vesta-next |\n\n### Nightly via Homebrew\n\n```bash\nbrew install --cask scouzi1966/afm/vesta-mac-next\n```\n\nVesta is a multi-backend AI chat application for macOS that runs models locally on Apple Silicon. It supports five AI backends simultaneously -- Apple Intelligence, MLX, llama.cpp, HuggingFace Inference API, and OpenAI-compatible API servers -- with runtime switching, vision understanding, text-to-speech, speech-to-text, image generation, video generation, and a full MCP server for programmatic control.\n\n### Demo\n\n[![Vesta Demo](demo-preview.gif)](https://vesta-mac.pages.dev/mcp-demo.mp4)\n\n*Click to watch the full demo video*\n\nNote that there is a known bug in Qwen3-VL models where it will loop indefinitely in some cases and never complete. In this case stop the generation and tweak your prompt. This is a model behavior, not the app.\nReference: https://github.com/QwenLM/Qwen3-VL/issues/1611#issuecomment-3639175711\n\n---\n\n## Vesta 0.9.7 (May 8, 2026)\n\nMLX model support sync with maclocal-api and release pipeline hardening.\n\n### Changes\n- **Qwen 3.6 support** -- imported maclocal-api architecture mapping for Qwen 3.6 MLX models.\n- **Gemma 4 support** -- aligned Gemma 4 model handling and chat template fallback behavior with maclocal-api.\n- **Crash fix** -- corrected Gemma 4 VL attention reshape handling for prompt generation.\n- **Release publishing** -- rebuilt from scratch in Release, notarized, stapled, and published as a signed DMG.\n\n---\n\n## Vesta 0.9.6 (Mar 4, 2026)\n\nStability and release pipeline improvements.\n\n### Changes\n- **Stacked download button** -- version and date displayed vertically on the website download button\n- **Publish date on stable button** -- stable download button now shows the release date\n- **Expanded test suite** -- automated test pipeline expanded to 4 stages with full coverage\n- **Pre-test validation** -- release skill checks for running Vesta instance before tests\n\n---\n\n## Vesta 0.9.5 (Feb 11, 2026)\n\n**94 commits, 2 months of development since v0.9.0**\n\nThis is a major release that transforms Vesta from a two-backend chat app into a full multi-modal AI platform with five backends, MCP integration, and media generation capabilities. Explore the world of AI beyond ChatGPT!\n\n### New in 0.9.5\n\n#### Multi-Backend Architecture\n- **HuggingFace Explorer** -- Leverage the HuggingFace Pro subscription and inference providers partners for ImageGen, Videogen, Image edit, Whisper etc. I am not affiliated with HuggingFace but the pro sub is worth it.\n- **5 simultaneous backends** -- Apple Intelligence, MLX, llama.cpp, HuggingFace Explorer, and External AI (OpenAI-compatible) all active at once (Single chat routing)\n- **Runtime backend switching** -- switch between backends without restarting the app\n- **Per-backend settings** -- each backend has its own generation parameters, model selection, and configuration\n- **HuggingFace Browser** -- Direct browse and download models from HuggingFace in-app\n\n#### MCP Server (Model Context Protocol) - Agentic Sidekick!\n- **Full MCP server** running on TCP loopback with token-based authentication\n- **Agentic Sidekick** Vesta detects Claude Code when you enable MCP - Claude Code acts as an agent with a NLI (Natural Language Interface) to the app. Ask Claude to set things up, have a conversation with any other model!\n- **41+ tools** -- backend management, chat, model download/load/unload, vision analysis, settings, conversation history search, diagnostics, UI navigation\n- **6 resources** -- app state, models, conversation, settings, logs, system info\n- **7 prompts** -- guides for Vesta, MLX, llama.cpp, HuggingFace, and common workflows\n- **AI Sidekick** -- Claude Code integration for programmatic Vesta control\n- **Conversation history search** -- full-text search and read-only SQL queries against the SQLite message database\n\n#### HuggingFace Explorer (New Backend)\n- **Cloud inference** via 16+ providers (Cerebras, Groq, Together, Fireworks, SambaNova, Nebius, Replicate, and more)\n- **Text-to-image generation** -- FLUX.1 Schnell/Dev, FLUX.2, Stable Diffusion 3.x, SDXL with configurable size, guidance, and steps\n- **Image editing** -- instruction-based editing with FLUX Kontext and FLUX.2 via Replicate\n- **Video generation** -- Wan 2.2 T2V (text-to-video) with async polling\n- **Speech-to-text transcription** -- OpenAI Whisper models via HuggingFace Inference API with 14 languages\n- **Vision/VLM** -- Qwen2.5-VL, Qwen3-VL and other vision-language models via cloud\n- **Model browser** -- search and discover models from HuggingFace Hub with download counts, likes, and gated model detection\n- **Reasoning display** -- chain-of-thought rendering for models that emit `\u003cthink\u003e` tags (DeepSeek R1, QwQ, etc.)\n- **7-tab settings panel** -- Chat, Vision, Image, Edit, Transcribe, Video, Settings\n\n#### Text-to-Speech (TTS)\n- **Kokoro** (82M) -- 46+ voices across 10+ languages, fast and high-quality\n- **Marvis** (100M/250M) -- conversational TTS with voice cloning support via reference audio\n- Models download on first use from HuggingFace\n\n#### Speech-to-Text (STT) -- WhisperKit\n- **On-device transcription** via WhisperKit CoreML -- zero network required\n- **6 model sizes** -- Tiny (39M) through Large V3 (1.5B) and Large V3 Turbo (809M)\n- **28+ languages** with auto-detect\n- Per-segment timing and speed ratio reporting\n\n#### Jinja Template Support (minja)\n- Full Jinja2 template parsing for GGUF models via llama.cpp's minja library\n- Correctly renders chat templates embedded in model metadata\n- Falls back to llama_chat_apply_template() for non-Jinja templates\n\n#### GGUF Model Browser\n- Search and browse GGUF models from HuggingFace Hub\n- Capability badges: Vision, Tool Use, Reasoning, Coding, Math, Multilingual\n- Split/multipart file detection\n- Automatic mmproj detection for vision models\n\n#### MLX Improvements\n- **Qwen3-VL M-RoPE patch** -- +81% performance improvement for vision inference (auto-applied via script)\n- **KV cache controls** exposed -- max KV size, quantization bits, prefill step size\n- **MLX benchmark tool** (mlx-bench) for standalone performance testing\n- **Wired memory** set to 90% of GPU recommended working set for large model performance\n\n### Improvements\n- **Per-message metrics** -- token count and tokens/sec stored with each message in the database\n- **Green parameter labels** -- visual indicator when a generation parameter matches the model's configured default\n- **Binary voice format** -- Kokoro voice files converted from JSON to binary (144 MB down to 27 MB)\n- **Conversation history view** with backend filtering and pagination\n- **One-command build** -- `build-from-scratch.sh` handles submodules, patching, llama.cpp library build, and Xcode build\n- **Distribution pipeline** -- automated DMG creation, notarization, and GitHub release via `build-vesta-mac-dist.sh`\n- **Automated testing framework** -- 44+ UI tests via MCP-based test runner\n\n### Bug Fixes\n- Fix AVKit VideoPlayer crash during SwiftUI transitions (disabled transition animations)\n- Fix download progress stuck at 0% and crash in llama.cpp streaming\n- Fix TTS mode hijacking text generation when both TTS and LLM models are loaded\n- Fix O(N^2) reasoning parser performance\n- Fix reasoning parser stripMarkers bug and chat history contamination\n- Fix streaming throttle not kicking in when content scrolls off-screen\n- Fix llama.cpp default context size (2048 changed to 16384 for Qwen3-VL)\n- Fix Continuity Camera Swift 6 concurrency crash in Release builds (Objective-C workaround)\n- Fix NSHostingView constraint crash in MLX settings window (non-observing wrapper)\n- Fix MoE warmup crash for models with 32+ experts (reduced warmup batch size)\n- Fix mxfp4 MoE Metal shader crash (skip warmup for mxfp4 models)\n- Fix ESpeakNG unsealed contents causing notarization failure\n- Fix GGUF vision model image handling and model deduplication\n\n---\n\n## Vesta 0.9.0 (Dec 10, 2025)\n\n### New in 0.9.0\n- **Vision capabilities** with Qwen3-VL model (describe images, analyze screenshots)\n- **Continuity Camera** input (capture photos directly from iPhone/iPad)\n- **Code syntax highlighting** for 20+ programming languages\n- **Edit responses** inline after generation\n- **HTML preview** for rendered content\n- **Enhanced LaTeX** math rendering in blockquotes\n- Improved rendering engine with real-time code block highlighting\n\n---\n\n## Features\n\n- **Apple Intelligence** -- on-device AI via Foundation Models framework (always available)\n- **MLX Backend** -- Apple Silicon optimized inference with mlx-swift (Qwen3-VL vision models)\n- **llama.cpp Backend** -- GGUF model support with full Metal GPU acceleration and Jinja templates\n- **HuggingFace Explorer** -- cloud inference, image/video generation, transcription via 16+ providers\n- **External AI** -- connect to any OpenAI-compatible API server (LM Studio, Ollama, etc.)\n- **Vision** -- image understanding via Qwen3-VL (MLX, llama.cpp, or HuggingFace)\n- **Text-to-Speech** -- Kokoro, Marvis (with voice cloning), and Orpheus TTS engines\n- **Speech-to-Text** -- WhisperKit on-device transcription (Tiny through Large V3)\n- **MCP Server** -- 41+ tools for programmatic control, model management, and AI Sidekick integration\n- **GitHub Flavored Markdown** -- tables, task lists, strikethrough via remark/rehype pipeline\n- **LaTeX Math** -- inline and block math rendering with KaTeX\n- **Code Highlighting** -- 20+ languages with real-time streaming highlight\n- **Liquid Glass UI** -- native macOS Tahoe design\n- **App Sandbox** -- Developer ID signed and Apple notarized\n\n## Verify\n\n```bash\n# Check SHA256\nshasum -a 256 ~/Downloads/Vesta-*.dmg\n\n# Check code signature\ncodesign --verify --deep --strict /Applications/Vesta.app\n\n# Check notarization\nspctl --assess --type execute /Applications/Vesta.app\n```\n\n## Requirements\n\n- **macOS 26.0** (Tahoe) or later\n- **Apple Silicon Mac** (M1/M2/M3/M4)\n- Microphone access for voice input and STT\n- Internet access for HuggingFace backend and model downloads (on-device backends work offline after model download)\n\n## Security \u0026 Privacy\n\n- Signed with **Developer ID Application: Soprano Technologies Inc.**\n- **Notarized by Apple**\n- **App Sandbox enabled**\n- On-device backends (Apple Intelligence, MLX, llama.cpp) process everything locally -- no data sent to servers\n- HuggingFace and External AI backends require network access for inference\n- API tokens stored in macOS Keychain\n\n## Related\n\n- **Source Code**: https://github.com/scouzi1966/vesta-mac\n- **Distribution**: https://github.com/scouzi1966/vesta-mac-dist (this repo)\n- **Website**: https://vesta-mac.pages.dev\n- **CLI Alternative**: https://github.com/scouzi1966/maclocal-api\n\n## Support\n\n- [Report an Issue](https://github.com/scouzi1966/vesta-mac-dist/issues/new)\n- [Request a Feature](https://github.com/scouzi1966/vesta-mac-dist/issues/new)\n- [Browse Issues](https://github.com/scouzi1966/vesta-mac-dist/issues)\n\n## License\n\n(c) 2025-2026 Soprano Technologies Inc. All rights reserved.\n\n## Built With\n\n- **Apple Intelligence** -- Foundation Models framework\n- **MLX** -- mlx-swift + mlx-swift-lm for Apple Silicon inference\n- **llama.cpp** -- GGUF inference with Metal acceleration\n- **WhisperKit** -- CoreML-based Whisper speech-to-text\n- **mlx-audio** -- Kokoro/Marvis/Orpheus TTS\n- **SwiftUI** -- native macOS interface\n- **KaTeX** -- math rendering\n- **highlight.js** -- code syntax highlighting\n- **remark/rehype** -- markdown processing pipeline\n\n---\n\nBuilt with automated distribution pipeline | Notarized and code-signed | Apple Silicon native\n\n## Star History\n\n[![Star History Chart](https://api.star-history.com/svg?repos=scouzi1966/vesta-mac-dist\u0026type=Date)](https://star-history.com/#scouzi1966/vesta-mac-dist\u0026Date)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fscouzi1966%2Fvesta-mac-dist","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fscouzi1966%2Fvesta-mac-dist","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fscouzi1966%2Fvesta-mac-dist/lists"}