local-ai

Easy Way to the DeepSeek‑Harness.

A Fluent Windows 11 desktop shell for the official DeepSeek Harness WebUI.

DeepSeek Harness plugin for local MLX-LM models with loopback-only defaults and optional managed server startup

Run DeepSeek Harness (DSH), a local AI agent, inside VS Code — native sidebar panel + editor bridge. A pure protocol client: no rewriting DSH, no second server

Scene-aware vision routing layer for DeepSeek Harness (dsh): decides which engine/backend an image should go to (chat/UI/table/code) before other vision plugins route it. Switch-gated, front-loaded, never touches other plugins' tools. 场景级识图路由层。

Local IndexTTS 2.5 and GPT-SoVITS speech synthesis and playback for DeepSeek Harness.

DSH 本地视觉能力插件：macOS Vision OCR + ollama qwen3-vl 语义描述 + 上传图片桥接

安全版本地语音输入插件 for DeepSeek Harness:同源宿主代理 + 127.0.0.1 + token,SenseVoice-Small INT8 ONNX 本地转写,音频不出网。

DeepSeek Harness 语音输入：本地多语言识别与 Codex Desktop 听写，只写入可编辑草稿。

适配调用JEPA模型的插件，提供一个本地的JEPA模型项目（需要进一步训练持续学习，已完成基本工具调用），可在家用计算机配置上运行。A plugin adapted to call the JEPA model, providing a local JEPA model project (requires further training for continual learning; basic tool calling has been completed) that can run on a home computer CPU.

DSH plugin for local OCR: handwritten Chinese + math → Markdown with LaTeX. GPU / CPU backends.

Local web fetch provider for DeepSeek Harness: web_fetch via local Ollama model (fetch page -> summarize/parse locally -> return to agent)