Ekorbia v0.1 — first release of a local LLM client

Ekorbia released v0.1.0, the first public build of a native macOS desktop client for local AI models that operates entirely offline without cloud services, API keys, or telemetry. The open-source application features multi-tab chat with streaming responses, local RAG capabilities with file and folder attachments, a quick-query overlay, and screenshot capture, all built with Tauri 2, Rust, React, and SQLite. The release is available as an unsigned DMG download from GitHub, with Windows and Linux builds planned for future development.

← All posts ./index.html Ekorbia v0.1 — first release of a local LLM client Ekorbia 0.1.0 is out. It's the first public build of a native macOS desktop client for local AI models, running entirely on your desktop. No cloud. No API keys. No telemetry. Built with Tauri 2 + Rust backend, plain React no bundler on the front-end, and SQLite for state. If you've been wanting a feature filled Mac-native chat app for your local models, Ekorbia is for you. What's in 0.1.0 Chat - Multi-tab chat with independent conversation history per tab. - Private ephemeral chats — nothing touches the database; conversation lives in memory only and disappears when the tab closes. - Streaming responses with a mid-generation Stop button that freezes partial output. - Markdown rendering for assistant replies with syntax-highlighted code blocks and per-block Copy buttons. - Edit & retry any past message; retry the last assistant reply to regenerate. Chat export to Markdown or JSON. Attachments & local RAG - File attachments .txt , .md , .pdf , images via paperclip . - Folder attachments — chunked, embedded locally with nomic-embed-text or any embedding model you pull , retrieved per query. - Incremental folder re-index — modify one file in a 500-file folder, only that file is re-embedded. - Citations : replies emit N markers that map to a Sources footer of clickable file chips. Shift-click any chip to reveal in Finder. - Vision routing : images route through vision-capable models as base64; a VISION badge indicates the active model can see them. Prompts, overlay, watches, files, memory - Prompt library backed by a folder of .md files with YAML frontmatter — git-friendly, shareable, editable in any editor. 28 built-in prompts ship with the app. - Quick-query overlay ⌘⇧Space — Spotlight-style panel over any app, streams from your local model, never steals focus. - Screenshot capture ⌘⇧1 — native region selector, opens a fresh chat with the image attached as a vision attachment. - Watch for folders, RSS feeds, and URLs with optional snapshot/diff modes and CSS selectors — pipes sources into chat as citable context with optional OS notifications. - Files panel — tool-capable models can call a write file tool to save outputs into a per-chat sandboxed output directory. - Memory file — a single editable markdown file the app injects as system context across chats. - Full-text search across chat history BM25-ranked, SQLite FTS5 . - Five themes — One Dark, One Light, Ayu Dark, Ayu Mirage, Ayu Light. What's not in 0.1.0 - Windows / Linux builds. This is on the TODO list. Linux is likely first. - Code signing. The .dmg isn't signed with an Apple Developer ID, so the first launch needs a xattr -dr com.apple.quarantine incantation or a right-click → Open . The README has the details https://github.com/ekorbia/ekorbia-desktop install . How to grab it Head to the releases page https://github.com/ekorbia/ekorbia-desktop/releases/latest , download the .dmg , drag it into Applications. If you'd rather not run unsigned binaries, you can build from source https://github.com/ekorbia/ekorbia-desktop build-from-source — locally-built binaries are signed with your machine's ad-hoc identity, which Gatekeeper trusts without the workaround.