Chevey339

@Kubai087 求一个 🙏 我做的 Kelivo，开源的全平台 LLM 客户端，支持 MCP、技能、长期记忆，手机上还内置 Linux 沙盒，模型能直接读写文件、跑命令，数据全留在本地。图标一直是自己凑合的，想换个像样点的。 Chevey339/kelivo

Kelivo: An open-source LLM client for mobile and desktop (4.3k stars) What it does All text below is quoted directly from the project README. >Kelivo is a cross-platform LLM client built with Flutter for Android, iOS, macOS, Windows and Linux. Connect your own API keys or sign in with a supported subscription, and switch between OpenAI, Gemini, Claude, DeepSeek, OpenRouter and any OpenAI-compatible service without switching apps. Features: - Native protocols: OpenAI Chat Completions and Responses API, Google Gemini and Vertex AI, and Anthropic Claude, plus any OpenAI-compatible endpoint, including self-hosted models. - Account sign-in: use a ChatGPT (Codex), Grok or Kimi Code account directly. Kelivo syncs the available models and shows your usage limits. - Multiple API keys per provider, with round-robin, priority, least-used or random load balancing and automatic error tracking. - Workspaces: bind a conversation to a workspace to give the model seven tools: shell, read_file, write_file, edit_file, list_dir, glob and grep. Command output streams live, file edits appear as diffs, and shell commands can require approval, which you grant per command or for the whole session. - Native on desktop: on macOS, Windows and Linux, tools run in your system shell, in an app-managed folder or a folder you link from your computer. - MCP: connect Model Context Protocol servers over Streamable HTTP, SSE or STDIO, with OAuth sign-in, JSON import, per-tool approval and a built-in fetch server. STDIO servers run natively on desktop and inside the Linux sandbox on mobile. - Scheduled tasks: run a prompt with a chosen assistant on a schedule, such as a morning briefing. Results are saved as conversations. Available on Android and desktop. - World books: entries triggered by keywords or regular expressions, injected at a configurable position, role and depth. >Conversations, settings and files are stored locally. Kelivo has no account system of its own; back up to WebDAV or S3-compatible storage whenever you choose. Repository Chevey339/kelivo Repository stats |Metric|Value| |:-|:-| |Stars|4,278| |Forks|488| |Last release|v1.3.1 (2026-10-09)| |Languages|Dart 95.8%, Kotlin 1.7%, Swift 0.8%| |Contributors|27| |AI contributor accounts (Claude / Codex / Cursor / Copilot)|claude: 16, cursoragent: 2| |AI co-authored commits|Claude: 16, Cursor: 2| Main contributor |Field|Value| |:-|:-| |Name|Psyche| |GitHub|@Chevey339| |Commits to this repo|2,063 (top contributor)| |Followers|81| ────────────────────────────── 📌 How many great GitHub repos have you discovered and completely forgotten about? Don't let another useful tool disappear into your Reddit feed. Get the community's best repos, organized by category, plus key takeaways from real developer discussions, delivered to your inbox twice a week. 📬 Never lose track of great repos again. Subscribe free.

Kelivo GitHub Explained: Local AI Workspace Security and Prompt Injection Risks kelivo by chevey339: chevey339/kelivo Kelivo is an open-source local LLM client whose workspace tools connect conversational intent to files and native desktop commands. This overview traces the security boundary from a local inference endpoint and disposable workspace through chat, file-preview, and command-approval layers. It also explains why desktop execution carries the active user's privileges, how indirect prompt injection can turn an untrusted document into a dangerous tool request, and why MCP integrations extend the same trust model. The focus is architectural: local privacy and capability come with a direct responsibility to review every proposed command. TimeStamps: 0:00 Operational AI agency and the security tradeoff 0:56 Kelivo's local-only architecture 1:29 Disposable workspaces and desktop privilege boundaries 2:39 The chat boundary: intent becomes a file operation 3:38 File previews and interruptible diffs 4:34 Shell access and indirect prompt injection 5:45 Command approval and approval fatigue 6:19 MCP integrations extend the risk surface 6:53 The chat, preview, and approval triad 7:25 Local capability means transferred security responsibility 🧠 Local AI turns conversation into operating-system actions 🗂️ Disposable workspaces narrow the model's file context 🛡️ Human approval remains the desktop execution gate 🔌 MCP connections inherit the same trust boundary Kelivo's architecture is most useful when local data sovereignty matters and the operator can act as a careful security gatekeeper. The practical takeaway is to keep workspace scope narrow, treat file content as potentially untrusted, and review shell or MCP requests individually rather than trading safety for unattended autonomy. #Kelivo #LocalAI #PromptInjection [de-DE] YNAS ynp_6167190a4cb96923 Cinematic Privater automatisierter Staging-Upload. Paketidentität: ynp_6167190a4cb96923; Ausgabe: cinematic. [es-US] YNAS ynp_6167190a4cb96923 cinemático Carga de entorno de prueba de automatización privada. Identidad del paquete: ynp_6167190a4cb96923; salida: cinematográfica. [fr-FR] YNAS ynp_6167190a4cb96923 cinématique Téléchargement de l'environnement de test d'automatisation privé. Identifiant du package : ynp_6167190a4cb96923 ; sortie : cinématique. [hi] YNAS ynp_6167190a4cb96923 सिनेमैटिक निजी स्वचालन स्टेजिंग अपलोड। पैकेज पहचान: ynp_6167190a4cb96923; आउटपुट: सिनेमाई। [id] YNAS ynp_6167190a4cb96923 sinematik Unggahan staging otomatisasi pribadi. Identitas paket: ynp_6167190a4cb96923; output: sinematik. [it] YNAS ynp_6167190a4cb96923 cinematografico Caricamento di automazione privata. Identità del pacchetto: ynp_6167190a4cb96923; output: cinematografico. [pt-BR] YNAS ynp_6167190a4cb96923 cinemático Upload de teste de automação privada. Identificação do pacote: ynp_6167190a4cb96923; saída: cinemática.