<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: İbrahim Kalaycı</title>
    <description>The latest articles on DEV Community by İbrahim Kalaycı (@shura5860).</description>
    <link>https://dev.to/shura5860</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4258312%2F4e2b1d97-9ee0-420a-883f-3734b4771f12.png</url>
      <title>DEV Community: İbrahim Kalaycı</title>
      <link>https://dev.to/shura5860</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/shura5860"/>
    <language>en</language>
    <item>
      <title>Building an Offline-First Local AI Coding Studio with Rust (PyO3), FastAPI, and Tauri 2.0)</title>
      <dc:creator>İbrahim Kalaycı</dc:creator>
      <pubDate>Sun, 22 Nov 2026 05:56:41 +0000</pubDate>
      <link>https://dev.to/shura5860/building-an-offline-first-local-ai-coding-studio-with-rust-pyo3-fastapi-and-tauri-20-1d10</link>
      <guid>https://dev.to/shura5860/building-an-offline-first-local-ai-coding-studio-with-rust-pyo3-fastapi-and-tauri-20-1d10</guid>
      <description>&lt;p&gt;If you work on proprietary codebases or regulated systems, sending your source code to cloud LLM APIs is often a non-starter. I wanted a local, offline-capable coding assistant that goes beyond a basic chat window—something that can actually index a repository, reason across files, propose line-by-line diffs, and let me review or revert changes safely.&lt;/p&gt;

&lt;p&gt;So I built &lt;strong&gt;AURA&lt;/strong&gt; (&lt;strong&gt;AURA-CORE&lt;/strong&gt; + &lt;strong&gt;AURA Studio&lt;/strong&gt;), and recently removed the dual-licensing layer to release the entire project as &lt;strong&gt;100% open source under GNU AGPLv3&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;🔗 &lt;strong&gt;GitHub Repository:&lt;/strong&gt; &lt;a href="https://github.com/Shura5860/LocalAIProject" rel="noopener noreferrer"&gt;https://github.com/Shura5860/LocalAIProject&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9nud51nlcwb4tlgey6px.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9nud51nlcwb4tlgey6px.png" alt="AURA Studio Desktop Preview" width="800" height="506"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;(Tip: If you record a short offline demo clip with Wi-Fi switched off, embed it right here!)&lt;/em&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  Why Build Another Local AI Tool?
&lt;/h2&gt;

&lt;p&gt;Most local setups either stop at a chat wrapper around an inference daemon, or require spinning up separate Docker containers for vector databases, proxies, and UIs.&lt;/p&gt;

&lt;p&gt;AURA takes a single-machine, embedded architecture approach:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;In-Process Rust Vector Engine (PyO3):&lt;/strong&gt; Instead of running an external vector database over the network, the HNSW vector index lives directly inside the Python process via PyO3 C-FFI bindings, backed by a Write-Ahead Log (WAL) and atomic disk snapshots.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Hybrid RAG 2.0:&lt;/strong&gt; Combines dense vector search (Rust HNSW) with sparse lexical search (Okapi BM25) using &lt;strong&gt;Reciprocal Rank Fusion (RRF)&lt;/strong&gt;, plus AST-based repo mapping and Caller → Callee call-graph extraction.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Local GGUF Inference with KV Prefix Caching:&lt;/strong&gt; Runs quantized GGUF models locally via &lt;code&gt;llama-cpp-python&lt;/code&gt; with &lt;code&gt;LlamaRAMCache&lt;/code&gt; prefix caching, automatic context compaction, and Linux CPU thread affinity (&lt;code&gt;sched_setaffinity&lt;/code&gt;) so the desktop UI stays responsive.&lt;/li&gt;
&lt;/ol&gt;




&lt;h2&gt;
  
  
  Composer-Style Diff Review, HITL Gates, and 1-Click Undo
&lt;/h2&gt;

&lt;p&gt;Letting an autonomous agent edit files directly on disk without guardrails is risky. In &lt;strong&gt;AURA Studio&lt;/strong&gt; (built with &lt;strong&gt;Tauri 2.0&lt;/strong&gt;), file-modifying operations follow an explicit review workflow:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Visual Unified Diff Preview:&lt;/strong&gt; Before &lt;code&gt;apply_diff&lt;/code&gt;, &lt;code&gt;write_file&lt;/code&gt;, or &lt;code&gt;apply_multi_diff&lt;/code&gt; touches disk, AURA simulates the patch and renders a line-numbered red/green unified diff with &lt;code&gt;+additions / -deletions&lt;/code&gt; counts.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Human-in-the-Loop (HITL) Approval:&lt;/strong&gt; Critical actions (file edits, shell commands, Python execution, git rollbacks) pause in a pending state until you click &lt;strong&gt;Apply&lt;/strong&gt; or &lt;strong&gt;Reject&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Snapshot-Backed 1-Click Undo:&lt;/strong&gt; Every applied change captures a pre-edit snapshot of affected files (and tracks newly created files). You can revert any individual edit—or all changes from the session—directly from the &lt;strong&gt;Changed Files&lt;/strong&gt; bar above the composer input.&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  Under the Hood: Agentic Core &amp;amp; Safety
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Autonomous ReAct Agent:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Semantic Tool Pruning:&lt;/strong&gt; Selects only the relevant tools per query instead of stuffing every tool schema into the prompt.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Disk Guard:&lt;/strong&gt; Verifies target paths against the bound workspace before executing file operations to catch hallucinated paths early.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Levenshtein Fuzzy Diff Matching:&lt;/strong&gt; Handles minor whitespace or indentation drift when applying code patches.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Flight Recorder:&lt;/strong&gt; Writes step-by-step execution traces to &lt;code&gt;.aura/traces/*.auratrace.json&lt;/code&gt; for offline debugging.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multi-Agent Swarm Mode:&lt;/strong&gt; Splits coding tasks across &lt;strong&gt;Architect&lt;/strong&gt; (planning), &lt;strong&gt;Coder&lt;/strong&gt; (diff patching), and &lt;strong&gt;Verifier&lt;/strong&gt; (compiler/syntax checks with rollback on failure).&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Defense in Depth:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;Strict workspace &lt;code&gt;realpath&lt;/code&gt; boundary checks (blocks path traversal and symlink escapes).&lt;/li&gt;
&lt;li&gt;Linux &lt;strong&gt;Bubblewrap (&lt;code&gt;bwrap&lt;/code&gt;)&lt;/strong&gt; sandboxing for terminal commands.&lt;/li&gt;
&lt;li&gt;SSRF protection on web fetching and XML quarantine envelopes for untrusted external content.&lt;/li&gt;
&lt;li&gt;Localhost-bound API (&lt;code&gt;127.0.0.1:8000&lt;/code&gt;) with restricted CORS and Tauri CSP.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ecosystem Standards:&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;OpenAI-Compatible API&lt;/strong&gt; (&lt;code&gt;/v1/chat/completions&lt;/code&gt;, &lt;code&gt;/v1/models&lt;/code&gt;, &lt;code&gt;/v1/embeddings&lt;/code&gt;) so you can also connect tools like Continue, Aider, or Open-WebUI.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Model Context Protocol (MCP)&lt;/strong&gt; stdio client bridge for external tools.&lt;/li&gt;
&lt;li&gt;Automatic discovery of &lt;code&gt;.cursorrules&lt;/code&gt; and &lt;code&gt;.aura/rules.md&lt;/code&gt;.&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ul&gt;




&lt;h2&gt;
  
  
  Getting Started (Linux)
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/Shura5860/LocalAIProject.git
&lt;span class="nb"&gt;cd&lt;/span&gt; LocalAIProject

python3 -m venv .venv
&lt;span class="nb"&gt;source&lt;/span&gt; .venv/bin/activate
pip install --upgrade pip
pip install maturin numpy &lt;span class="s2"&gt;"httpx[http2]"&lt;/span&gt; duckduckgo_search sentence-transformers psutil fastapi uvicorn llama-cpp-python

&lt;span class="c"&gt;# Build the in-process Rust vector engine&lt;/span&gt;
&lt;span class="nb"&gt;cd&lt;/span&gt; Rust-Python-Vector-Engine/vector_engine
maturin develop --release
&lt;span class="nb"&gt;cd&lt;/span&gt; ../..

&lt;span class="c"&gt;# Launch AURA-CORE + AURA Studio&lt;/span&gt;
./start_assistant.sh
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;






&lt;h2&gt;
  
  
  Looking for Feedback &amp;amp; Contributors
&lt;/h2&gt;

&lt;p&gt;The project is tested on Linux (Arch, Ubuntu, Fedora) and licensed under &lt;strong&gt;GNU AGPLv3&lt;/strong&gt;. I would love feedback from developers who work in air-gapped or privacy-strict environments:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;What is still missing before you would trust a local agent on a real repository?&lt;/li&gt;
&lt;li&gt;Which local models and workflows are you using right now?&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Repo: &lt;a href="https://github.com/Shura5860/LocalAIProject" rel="noopener noreferrer"&gt;https://github.com/Shura5860/LocalAIProject&lt;/a&gt; — issues, PRs, and architectural critiques are very welcome!&lt;/p&gt;

</description>
      <category>rust</category>
      <category>opensource</category>
      <category>ai</category>
      <category>python</category>
    </item>
    <item>
      <title>Building a Rootless Linux Sandbox in Pure Rust — Without Docker or a Daemon</title>
      <dc:creator>İbrahim Kalaycı</dc:creator>
      <pubDate>Sun, 22 Nov 2026 05:54:30 +0000</pubDate>
      <link>https://dev.to/shura5860/building-a-rootless-linux-sandbox-in-pure-rust-without-docker-or-a-daemon-21be</link>
      <guid>https://dev.to/shura5860/building-a-rootless-linux-sandbox-in-pure-rust-without-docker-or-a-daemon-21be</guid>
      <description>&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9ejnqc6gd8bdf1bqkcby.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F9ejnqc6gd8bdf1bqkcby.png" alt=" " width="800" height="1060"&gt;&lt;/a&gt;&lt;br&gt;
If you ever need to run untrusted scripts, test builds, or execute AI-generated code on Linux, the usual options are either spinning up a full Docker container or trusting the process on your host.&lt;/p&gt;

&lt;p&gt;I wanted something lighter: a single binary and embeddable Rust library that isolates a process using Linux's built-in kernel features — &lt;strong&gt;no root privileges, no daemon, and no C dependencies&lt;/strong&gt;.&lt;/p&gt;

&lt;p&gt;That turned into &lt;strong&gt;&lt;a href="https://github.com/Shura5860/rust-sandbox" rel="noopener noreferrer"&gt;rust-sandbox (&lt;code&gt;rsb&lt;/code&gt;)&lt;/a&gt;&lt;/strong&gt; (&lt;code&gt;v0.1.2&lt;/code&gt;).&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Folder-level access control:&lt;/strong&gt; Choose exactly which directories a command can read or write, or run it in a temporary copy-on-write (&lt;code&gt;--cow&lt;/code&gt;) mode that leaves your real folder completely untouched and reports what changed on exit.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Domain-level network filtering:&lt;/strong&gt; Keep networking off by default, or allow outbound traffic only to specific domains and ports while blocking private networks and cloud metadata IPs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Resource &amp;amp; timeout limits:&lt;/strong&gt; Enforce timeouts, output size caps, and optional Cgroups v2 memory/CPU/PID quotas.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Auto-generated profiles:&lt;/strong&gt; Run &lt;code&gt;rsb learn&lt;/code&gt; once on a command to trace what it needs and save a reusable &lt;code&gt;sandbox.toml&lt;/code&gt; config.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Embeddable everywhere:&lt;/strong&gt; Use it as a standalone CLI, a Rust crate (&lt;code&gt;rsb-core&lt;/code&gt;), or via the included C and Python FFI bindings (including a warm Python worker pool for fast repeated runs).&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Quick look
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Allow writing only to ./work, with no internet access&lt;/span&gt;
rsb run --rw ./work -- python3 untrusted.py

&lt;span class="c"&gt;# Run destructively without touching the real folder&lt;/span&gt;
rsb ai-exec --cow ./work -- /bin/sh -c &lt;span class="s1"&gt;'rm -rf ./work/*; echo test &amp;gt; ./work/a.txt'&lt;/span&gt;

&lt;span class="c"&gt;# Allow HTTPS only to example.com:443&lt;/span&gt;
rsb run --transparent-tap --allow-domain example.com:443 -- curl -sI https://example.com
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Scope &amp;amp; Notes
&lt;/h2&gt;

&lt;p&gt;It's a Linux-only tool that shares the host kernel (namespace, Landlock, and Seccomp-BPF isolation — not a virtual machine). All of the architectural decisions and known boundaries are documented in &lt;a href="https://github.com/Shura5860/rust-sandbox/blob/main/ARCHITECTURE.md" rel="noopener noreferrer"&gt;&lt;code&gt;ARCHITECTURE.md&lt;/code&gt;&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;If you're into Rust, Linux internals, or sandboxing AI agents, feel free to check out the repo and share feedback or test it on your distro:&lt;/p&gt;

&lt;p&gt;👉 &lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/Shura5860/rust-sandbox" rel="noopener noreferrer"&gt;https://github.com/Shura5860/rust-sandbox&lt;/a&gt;&lt;/p&gt;

</description>
      <category>rust</category>
      <category>linux</category>
      <category>security</category>
      <category>opensource</category>
    </item>
    <item>
      <title>Building Corten Desktop: A Pure-Rust Wayland Desktop Environment with Smithay 0.7 &amp; Iced 0.13</title>
      <dc:creator>İbrahim Kalaycı</dc:creator>
      <pubDate>Sun, 22 Nov 2026 05:52:00 +0000</pubDate>
      <link>https://dev.to/shura5860/building-corten-desktop-a-pure-rust-wayland-desktop-environment-with-smithay-07-iced-013-3l64</link>
      <guid>https://dev.to/shura5860/building-corten-desktop-a-pure-rust-wayland-desktop-environment-with-smithay-07-iced-013-3l64</guid>
      <description>&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8l1vyyki5td12wne5pu0.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F8l1vyyki5td12wne5pu0.jpg" alt=" " width="800" height="450"&gt;&lt;/a&gt;Over the past few months I’ve been building &lt;strong&gt;Corten Desktop&lt;/strong&gt; (&lt;code&gt;v0.5.8&lt;/code&gt;), an open-source, process-isolated Wayland desktop environment written in pure Rust—structured as a single 30-crate Cargo workspace with no &lt;code&gt;wlroots&lt;/code&gt; C wrappers, no Electron/WebKit shells, and no Python/Bash glue scripts.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;GitHub Repository:&lt;/strong&gt; &lt;a href="https://github.com/Shura5860/corten-desktop" rel="noopener noreferrer"&gt;https://github.com/Shura5860/corten-desktop&lt;/a&gt;&lt;br&gt;&lt;br&gt;
&lt;strong&gt;Latest Release (&lt;code&gt;v0.5.8&lt;/code&gt;):&lt;/strong&gt; &lt;a href="https://github.com/Shura5860/corten-desktop/releases/tag/v0.5.8" rel="noopener noreferrer"&gt;https://github.com/Shura5860/corten-desktop/releases/tag/v0.5.8&lt;/a&gt;&lt;/p&gt;


&lt;h2&gt;
  
  
  Why Build a Desktop Environment from Scratch in Rust?
&lt;/h2&gt;

&lt;p&gt;Linux desktops usually fall into two camps: massive C/C++ monoliths, or DIY tiling window manager setups held together by dozens of unrelated utilities and fragile scripts. &lt;strong&gt;Corten Desktop&lt;/strong&gt; aims for a cohesive, memory-safe middle ground where the compositor, dock, launcher, notifications, file manager, settings, portals, and app store all share one workspace, one theme crate, and one IPC layer.&lt;/p&gt;
&lt;h2&gt;
  
  
  Architecture Highlights
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Compositor (&lt;code&gt;corten-compositor&lt;/code&gt;):&lt;/strong&gt; Built directly on &lt;strong&gt;Smithay 0.7&lt;/strong&gt; with both a standalone DRM/KMS (&lt;code&gt;udev&lt;/code&gt; + &lt;code&gt;libseat&lt;/code&gt; + &lt;code&gt;libinput&lt;/code&gt; + &lt;code&gt;gbm&lt;/code&gt; + &lt;code&gt;glow&lt;/code&gt;) backend and a nested Winit backend for development. Includes floating, master-stack, BSP, monocle, and scrollable-strip layouts, direct hardware scanout, &lt;code&gt;linux-drm-syncobj-v1&lt;/code&gt;, adaptive tearing (&lt;code&gt;wp_tearing_control_v1&lt;/code&gt;), XWayland, and custom GLSL ES Dual-Kawase blur/halo shaders.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Process-Isolated Shell (&lt;code&gt;Iced 0.13&lt;/code&gt; + &lt;code&gt;tiny-skia&lt;/code&gt;):&lt;/strong&gt; Every component runs as its own binary (&lt;code&gt;corten-panel&lt;/code&gt;, &lt;code&gt;corten-launcher&lt;/code&gt;, &lt;code&gt;corten-notify&lt;/code&gt;, &lt;code&gt;corten-widgets&lt;/code&gt;, &lt;code&gt;corten-lock&lt;/code&gt;, &lt;code&gt;corten-greeter&lt;/code&gt;, &lt;code&gt;corten-polkit&lt;/code&gt;, &lt;code&gt;corten-portal&lt;/code&gt;, &lt;code&gt;corten-wallpaper&lt;/code&gt;, &lt;code&gt;corten-settings&lt;/code&gt;, &lt;code&gt;corten-files&lt;/code&gt;, &lt;code&gt;corten-monitor&lt;/code&gt;, &lt;code&gt;corten-snipper&lt;/code&gt;, &lt;code&gt;corten-forge&lt;/code&gt;, &lt;code&gt;corten-ai&lt;/code&gt;). Rendering shell utilities with &lt;code&gt;tiny-skia&lt;/code&gt; keeps GPU VRAM dedicated to user apps and games.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;IPC &amp;amp; Supervision (&lt;code&gt;corten-ipc&lt;/code&gt;, &lt;code&gt;corten-session&lt;/code&gt;):&lt;/strong&gt; Components communicate via &lt;code&gt;zbus&lt;/code&gt; (D-Bus) and Unix JSON-RPC sockets, supervised by &lt;code&gt;corten-session&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Corten Forge &amp;amp; OTA (&lt;code&gt;corten-forge&lt;/code&gt;, &lt;code&gt;corten-update&lt;/code&gt;):&lt;/strong&gt; Built-in software center with live Flathub API v2 browsing/search, Flatpak sandbox inspection, Arch repository support, and SHA256-verified GitHub Release OTA updates.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Optional Local AI (&lt;code&gt;corten-llm&lt;/code&gt;, &lt;code&gt;corten-vector&lt;/code&gt;, &lt;code&gt;corten-ai&lt;/code&gt;):&lt;/strong&gt; Pure-Rust &lt;code&gt;mmap&lt;/code&gt; GGUF inference + HNSW vector index for an offline desktop assistant.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2&gt;
  
  
  Try It Out (&lt;code&gt;v0.5.8&lt;/code&gt;)
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Prebuilt x86_64 binary installer:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;curl -fsSL https://raw.githubusercontent.com/Shura5860/corten-desktop/main/scripts/install-release-bin.sh &lt;span class="o"&gt;|&lt;/span&gt; sudo bash
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Fresh Arch Linux ISO (&lt;code&gt;archinstall&lt;/code&gt; profile):&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;archinstall --config https://raw.githubusercontent.com/Shura5860/corten-desktop/main/packaging/arch/archinstall-corten.json
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;Build from source:&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/Shura5860/corten-desktop.git
&lt;span class="nb"&gt;cd&lt;/span&gt; corten-desktop
cargo &lt;span class="nb"&gt;test&lt;/span&gt; --workspace
cargo build --release --workspace
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Feedback, bug reports, and contributions are very welcome—especially from anyone testing on different GPUs or distros!&lt;/p&gt;

</description>
      <category>rust</category>
      <category>linux</category>
      <category>wayland</category>
      <category>opensource</category>
    </item>
  </channel>
</rss>
