Skip to content
Back to search

Vibe_Assistant

A cross-platform desktop application for managing AI Agents and MCP Servers. 一款用于管理AI智能体与MCP服务器的跨平台桌面应用程序。

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/Zafer-Liu/Vibe_Assistant with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/Zafer-Liu/Vibe_Assistant'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/zafer-liu--vibe-assistant/badge.svg)](/listings/zafer-liu--vibe-assistant)

HTML

<a href="/listings/zafer-liu--vibe-assistant"><img src="/api/listings/zafer-liu--vibe-assistant/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Download a prebuilt package (recommended)

Grab the latest from [Releases](https://github.com/Zafer-Liu/Vibe_Assistant/releases):

| Platform | Package | |----------|---------| | Windows x64 | `-x64-setup.exe` (NSIS installer) | | macOS Apple Silicon | `-aarch64.dmg` | | macOS Intel | `-x64.dmg` | | Linux x64 | `-amd64.AppImage` / `-amd64.deb` |

Build from source

**Prerequisites**: [Node.js](https://nodejs.org) ≥ 18, [Rust](https://rustup.rs) stable, and the [Tauri prerequisites](https://tauri.app/start/prerequisites/) (VS C++ Build Tools on Windows).

git clone https://github.com/Zafer-Liu/Vibe_Assistant.git
cd Vibe_Assistant

# install dependencies
npm install
cd frontend && npm install && cd ..

# development
npm run dev

# production build
npm run build

Bundles land in `src-tauri/target/release/bundle/`.

---

<a id="quickstart"></a>

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 541 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. <img src="./Logo/Banner.png" alt="智管-Vibe Assistant Banner" width="100%" />
  3. # 智管-Vibe Assistant
  4. **One workspace for memory, Skills, and MCP.**
  5. [![Version](https://img.shields.io/badge/Version-v1.1.0-blue.svg)](https://github.com/Zafer-Liu/Vibe_Assistant/releases)
  6. [![License](https://img.shields.io/badge/License-Apache%202.0-yellow.svg)](LICENSE)
  7. [![Stars](https://img.shields.io/github/stars/Zafer-Liu/Vibe_Assistant?style=flat-square)](https://github.com/Zafer-Liu/Vibe_Assistant/stargazers)
  8. [![CI](https://img.shields.io/github/actions/workflow/status/Zafer-Liu/Vibe_Assistant/ci.yml?style=flat-square&label=CI)](https://github.com/Zafer-Liu/Vibe_Assistant/actions)
  9. [![Platform](https://img.shields.io/badge/Platform-Windows%20%7C%20macOS%20%7C%20Linux-blue?style=flat-square)](#install)
  10. [![Tauri](https://img.shields.io/badge/Tauri-2-24C8DB?style=flat-square&logo=tauri)](https://tauri.app)
  11. [![React](https://img.shields.io/badge/React-19-61DAFB?style=flat-square&logo=react)](https://react.dev)
  12. [![Rust](https://img.shields.io/badge/Rust-stable-CE422B?style=flat-square&logo=rust)](https://www.rust-lang.org)
  13. [![Sponsor 88API](https://img.shields.io/badge/Sponsor-88API-FF6B35?style=flat-square&logo=OpenAI&logoColor=white)](https://88api.ai/sign-up?aff=VNNV)
  14. [✨ Highlights](#highlights) · [🧠 Memory Center](#memory) · [⚙️ Install](#install) · [🚀 Quick Start](#quickstart) · [❓ FAQ](#faq)
  15. **[简体中文](./README_ZH.md)** | English
  16. </div>
  17. ---
  18. <a id="sponsors"></a>
  19. ## 💖 Sponsors
  20. Thanks to **[88API](https://88api.ai/sign-up?aff=VNNV)** for sponsoring this project! 88API is a one-stop multi-model API platform operated by a Hong Kong company, built for developers, creators, and AI app users. It exposes a unified interface to text, image, voice, and video models — covering AI programming, smart translation, content creation, dubbing, image generation, and video generation in common workflows. You can add and switch providers directly inside Vibe Assistant, with international payment methods, invoices, and enterprise-grade reliability. **[Sign up via this link](https://88api.ai/sign-up?aff=VNNV)** to get an exclusive bonus credit!
  21. <p align="center">
  22. <a href="https://88api.ai/sign-up?aff=VNNV"><img src="frontend/public/88api-banner.jpg" alt="88API — One API for text, image, voice, video" width="600"></a>
  23. </p>
  24. ---
  25. ## ✨ Highlights
  26. **Vibe Assistant** is a local Tauri 2 + React 19 + Rust desktop workspace focused on:
  27. - **Memory management**: capture and search conversations, maintain working memory and a long-term Profile, and inject context into connected clients.
  28. - **Skill management**: browse, publish, and deploy skills, review local differences, and reuse equipment profiles.
  29. - **MCP management**: manage services and client configurations, discover tools, and inspect calls and errors.
  30. - Supporting tools: token usage, ports, networking, environment variables, backups, and cloud memory sync; Chinese/English and light/dark themes.
  31. Process launching, embedded terminals and Web UIs, collaboration tasks, and proxy publishing have been removed. Existing clients such as Claude and Codex still use their memory, Skill, and MCP integrations. Saved memories, configurations, and historical output files are preserved.
  32. ---
  33. ## 🎬 How It Works
  34. Three systems, one goal: give every agent the same memory, skills, and tools without copying config or losing track of who said what.
  35. | System | The problem it solves | Watch it work |
  36. |--------|----------------------|---------------|
  37. | **🧬 Memory** | Injecting full history blows up context; injecting nothing loses continuity | [L0→L3 pipeline](#memory-deep-dive) |
  38. | **🔌 MCP** | Each agent wants its own config format; debugging calls across 10 clients is hell | [Unified management](#mcp-deep-dive) |
  39. | **📚 Skills** | Copying SKILL.md files between agents; version drift you can't see | [Shared library](#skill-deep-dive) |
  40. ---
  41. ## 📑 Table of Contents
  42. - [How It Works](#how-it-works)
  43. - [Memory Center](#memory)
  44. - [Memory deep dive: L0–L3](#memory-deep-dive)
  45. - [Skill management](#skills)
  46. - [Skill deep dive: scan → publish → equip](#skill-deep-dive)
  47. - [MCP Server Center](#mcp)
  48. - [MCP deep dive: one config, ten clients](#mcp-deep-dive)
  49. - [System Toolkit](#system)
  50. - [Install](#install)
  51. - [Quick Start](#quickstart)
  52. - [Model configuration](#llm-config)
  53. - [Data Storage](#data-paths)
  54. - [Telemetry and hooks](#telemetry)
  55. - [FAQ](#faq)
  56. - [Contributing](#contributing)
  57. - [License](#license)
  58. ---
  59. <a id="mcp"></a>
  60. ## 🔌 MCP Server Center
  61. A single catalog for MCP servers on your machine — manage them once, distribute everywhere.
  62. <a id="mcp-deep-dive"></a>
  63. ### 🎬 How MCP flows: one config → ten clients
  64. ![How MCP flows: one config to ten clients](Images/animations/mcp-system.gif)
  65. One catalog defines your MCP servers; Vibe Assistant writes each client's native config format (Claude JSON, Codex TOML, dsh YAML, and seven others). A stdio relay sits between agent and server to observe call boundaries (server, tool, success, latency — never arguments or content) and reports to a local telemetry ledger. Config changes hot-reload into running clients without restarts.
  66. **Why it matters**: stop hand-editing ten config files; debug "which tool failed and when" across every agent in one place; your prompts and tool outputs never leave your machine.
  67. ![MCP Server Center](Images/mcpcenter.png)
  68. **Local server management**
  69. - **Local scan**: auto-detects npm-global MCP packages
  70. - **Smart parsing**: paste any text (docs, install instructions) and AI extracts the config
  71. - **Manual entry**: stdio / SSE configurations
  72. **Catalog & distribution**
  73. - Keep frequently used MCP servers as catalog entries
  74. - One-click install into Claude Code, Claude Desktop, Codex CLI, Codex Desktop, Qoder, WorkBuddy, MiniMax Code, Kimi, ZCode, and more
  75. - Import from Agents' existing configs; per-host install status shown live (installed / config drift)
  76. ---
  77. <a id="system"></a>
  78. ## 🧰 System Toolkit
  79. <a id="ports"></a>
  80. ### Port Manager
  81. ![Port Manager](Images/portcenter.png)
  82. - Every listening port with protocol, PID, and process name
  83. - Kill the process occupying a port in one click — the fastest way to diagnose "port already in use" startup failures
  84. <a id="network"></a>
  85. ### Network Manager
  86. ![Network Manager](Images/Internetcenter.png)
  87. - **Live throughput** per network interface, with a rolling download/upload chart
  88. - **Interface list** with IPv4 / IPv6 addresses, MAC, and MTU
  89. - **Public IP** for both direct and proxy egress (detects your proxy automatically)
  90. - **Bandwidth speed test** — latency, download, and upload over both direct and proxy paths
  91. <a id="env"></a>
  92. ### Environment Variables
  93. ![Environment Variables](Images/environmentvariable.png)
  94. The equivalent of Windows' "Edit environment variables" dialog, without opening the registry editor:
  95. - **User and system** scopes in separate tabs (system changes need an administrator-run app)
  96. - **PATH-style variables are edited one entry per line** — add, remove, and reorder paths visually instead of fighting a semicolon blob
  97. - Values containing `%REFERENCES%` are stored as `REG_EXPAND_SZ` with an expanded-value preview
  98. - Changes are broadcast (`WM_SETTINGCHANGE`) so newly started processes pick them up immediately
  99. - Critical variables (PATH, PATHEXT, WINDIR, …) are protected from accidental deletion
  100. ---
  101. <a id="memory"></a>
  102. ## 🧬 Memory Center — One Brain for All Agents
  103. Every coding Agent you use remembers your preferences, decisions, and current work — and every other Agent benefits from it.
  104. <a id="memory-deep-dive"></a>
  105. ### 🎬 How memory flows: L0 → L3
  106. ![How memory flows: L0 to L3](Images/animations/memory-system.gif)
  107. Every conversation an agent completes becomes raw events (L0). An LLM extracts 1–3 typed records per session (L1). Every 30 days of L1 evidence consolidates into working memory (L2). A human-reviewed draft of L2 publishes as the long-term profile (L3) that every connected agent receives at session start. The split keeps context budgets small while preserving an audit trail back to the original conversation.
  108. **Why it matters**: durable memory without the "paste 500K tokens of history" cost; a human gate prevents model mistakes from polluting long-term preferences.
  109. ![Memory Center](Images/memorycentre.png)
  110. ### The L0–L3 memory model
  111. | Layer | Name | Source | Role |
  112. |-------|------|--------|------|
  113. | L0 | Event ledger | Hook events / local transcript scans | Audit source; never injected as a whole |
  114. | L1 | Retrievable memories | 1–3 facts / decisions / constraints / preferences extracted at session end | Semantic + keyword recall on demand |
  115. | L2 | 30-day working memory | Map-reduce compression of recent L1 evidence | Injected at session start |
  116. | L3 | Long-term Profile | Drafted from L2, gated by review before publishing | Injected every turn — stable preferences and constraints |
  117. ### 📥 Automatic capture (Agents → memory)
  118. - Supports **Codex, Claude Code, Qoder, WorkBuddy, MiniMax Code, Kimi, GitHub Copilot, ZCode, and DeepSeek Harness** — via hooks (the first four) or local transcript scanning (all nine; DeepSeek Harness zstd-compressed sessions are decoded in place). DeepSeek Harness also gets hook-equivalent automatic injection through its cordis pre-step plugin
  119. - Sessions land in a local SQLite ledger first, extraction runs async — nothing is lost when offline or before a model is configured; failed sessions can be re-run manually
  120. - **Pending Memories** and **Organized Conversations** panels show full conversations, extraction progress, and failure reasons
  121. - L1 cleaning & dedup ("dreaming"): local BGE semantic candidates + LLM arbitration, with rollback checkpoints
  122. - **Import from a folder**: point the app at any transcript folder and extract memories from it directly
  123. ### 🔎 Search, ranking & the memory engine
  124. - **Semantic search** across all L1 memories (hybrid semantic + keyword) right in the Memory Center
  125. - **Importance ranking**: every memory gets a score with its evidence (supporting sessions, Agents, recall count); pin what matters
  126. - **Memory engine**: the optional local semantic sidecar (status visible in the Memory Center) can be started / stopped on demand
  127. - Inline edit / delete of any memory, plus manually added entries
  128. - Reset-all-for-re-extraction when you want a clean rebuild of L1
  129. ### ⏱️ Scheduled L2/L3 refresh
  130. The published L2 working memory and L3 Profile no longer depend on clicking buttons:
  131. - Checks every 30 minutes while the app runs; rebuilds **L2** when new memories arrived and the configured interval (default **7 days**) elapsed
  132. - Rebuilds and **auto-publishes L3** when L2 changed and its interval (default **30 days**) elapsed
  133. - Intervals and the on/off switch live in **Settings → Scheduled memory refresh**; changes apply immediately
  134. - Every background run is logged on the Memory Injection page; manual consolidation still works anytime
  135. ### 💉 Memory injection (memory → Agents, dual channel)
  136. | Channel | Mechanism | Best for |
  137. |---------|-----------|----------|
  138. | Hook injection | SessionStart injects L2 once; UserPromptSubmit injects L3 every turn (content-fingerprint gated — unchanged turns cost zero tokens) | Working context at start, durable preferences throughout |
  139. | Shared-memory MCP | Agents call `recall_memory` for semantic search on demand | Task-specific recall without bloating context |
  140. **One-click setup**: install / uninstall the shared-memory MCP into **Claude Code, Claude Desktop, Codex CLI, Codex Desktop, Qoder, WorkBuddy, MiniMax Code, Kimi, ZCode, DeepSeek Harness**, and manage capture/injection hooks for the four hook-capable Agents (Claude Code, Codex, Qoder, WorkBuddy). Per-device data directory overrides are supported.
  141. ### 🔗 DeepSeek Harness bridge plugin
  142. [dsh-vibe-assistant-bridge](https://github.com/Zafer-Liu/dsh-vibe-assistant-bridge) is the official bridge plugin that turns DeepSeek Harness into a first-class Vibe Assistant citizen — one bundle, six capabilities:
  143. | Capability | What it does |
  144. |------------|--------------|
  145. | **Memory injection** | L2/L3 shared memory folded into every session start (offline snapshot fallback) |
  146. | **MCP observation** | Tool call boundaries reported to the local ledger (no arguments, no content) |
  147. | **Token usage** | Per-message usage reported in real time, deduplicated against transcript scans |
  148. | **Collection trigger** | Session end signals immediate L0→L1 extraction (no waiting for the scan cycle) |
  149. | **Recall tools** | `recall_memory` / `get_user_profile` as native dsh tools |
  150. | **Skill tools** | `list_shared_skills` / `read_shared_skill` as native dsh tools |
  151. **Install**: click **一键接入** (one-click setup) in Vibe Assistant's MCP library, or in the dsh desktop app go to **Plugins → Add plugin** and paste `dsh-vibe-assistant-bridge`. Requires Vibe Assistant running locally (loopback HTTP on port 9420).
  152. **Source & development**: [Zafer-Liu/dsh-vibe-assistant-bridge](https://github.com/Zafer-Liu/dsh-vibe-assistant-bridge) (MIT). Also discoverable in the community plugin directory via the `dsh-plugin` topic.
  153. ### ✍️ Custom memories
  154. Hand-written entries under the working-memory or long-term cards — only you can edit or delete them, automatic cleaning never touches them, and they are always injected.
  155. <a id="skills"></a>
  156. ### 📚 Shared Skill library & marketplace
  157. <a id="skill-deep-dive"></a>
  158. ### 🎬 How skills flow: scan → publish → equip
  159. ![How skills flow: scan, publish, equip](Images/animations/skill-system.gif)
  160. Every agent's `SKILL.md` files scan into one shared library, keyed by content hash. Before any sync you preview exactly what will be added (green), updated (yellow), or conflicted (red). "Publish + equip" writes approved skills back to every agent's directory in one step. Version drift is visible per agent; one click rolls back the latest publish or adopts a local edit as the new shared version. The marketplace installs curated skills from official OpenAI and Anthropic repositories with size bounds and path validation.
  161. **Why it matters**: no more "which agent has the latest version of this skill"; conflicts are explicit choices, not silent overwrites; marketplace skills are audited before they touch disk.
  162. ![Skill Library](Images/skillcenter.png)
  163. - **Skill library**: scans each Agent's `SKILL.md` files into a shared library; content-hash previews of adds / updates / conflicts before you confirm a sync; "publish + equip" in one step, with bulk operations, per-Agent version drift detection, one-click rollback of the latest publish, and "adopt local" to take an Agent's local edit as the new shared version
  164. - **Marketplace**: browse and install curated skills from the official **OpenAI** and **Anthropic** skill repositories — downloads are size-bounded, every relative path validated, and downloaded scripts are never executed
  165. ### 📊 Token usage analytics
  166. ![Token usage](Images/tokencenter.png)
  167. Aggregates real input / output / cache usage from local transcripts across Agents — coverage includes Codex, Claude Code, Qoder, WorkBuddy, MiniMax Code, Kimi, GitHub Copilot, Gemini CLI, OpenCode, OpenClaw, and more — aligned with how providers bill:
  168. - Summary dashboard plus per-session usage records and live session activity
  169. - Injection dashboard: per-source session/prompt injection counts, fingerprint-gate hit rate, and estimated token cost
  170. ### ☁️ Cloud Memory Vault (optional)
  171. Sync published L2/L3 documents (with their source provenance), custom memories, and usage stats across machines — one brain everywhere. On WebDAV, S3 and Baidu Netdisk targets, **all extracted L1 memories — including session-scoped ones — sync as per-device read-only mirrors**: each machine writes only its own aggregate, deletions and re-organization propagate along, receiver-side retagging and pin/recall decisions merge back across devices, and the original scope labels are preserved.
  172. - **AES-256-GCM end-to-end encryption**: keys derived from your sync password, never stored; the server only ever sees ciphertext
  173. - CAS conflict detection, incremental pulls, tombstone deletions; local SQLite stays authoritative, offline sync failures never block injection
  174. - **Baidu Netdisk** is available alongside the self-hosted, WebDAV and S3 targets. In Settings → Cloud Memory Vault, select Baidu Netdisk, authorize it, paste the refresh token and choose **Deploy in one click**. Select a private tools folder for OpenList, then save settings. Authorize the same account on each device and sync them in turn using the same sync password. The encrypted file is `/VibeAssistant/agent-manager/vault.amv`.
  175. - Single-file self-hostable server, open source in [`vault/`](vault/README.md) — Docker / Railway / Orange Pi + Tailscale
  176. ---
  177. <a id="install"></a>
  178. ## ⚙️ Install
  179. ### Download a prebuilt package (recommended)
  180. Grab the latest from [Releases](https://github.com/Zafer-Liu/Vibe_Assistant/releases):
  181. | Platform | Package |
  182. |----------|---------|
  183. | Windows x64 | `-x64-setup.exe` (NSIS installer) |
  184. | macOS Apple Silicon | `-aarch64.dmg` |
  185. | macOS Intel | `-x64.dmg` |
  186. | Linux x64 | `-amd64.AppImage` / `-amd64.deb` |
  187. ### Build from source
  188. **Prerequisites**: [Node.js](https://nodejs.org) ≥ 18, [Rust](https://rustup.rs) stable, and the [Tauri prerequisites](https://tauri.app/start/prerequisites/) (VS C++ Build Tools on Windows).
  189. ```bash
  190. git clone https://github.com/Zafer-Liu/Vibe_Assistant.git
  191. cd Vibe_Assistant
  192. # install dependencies
  193. npm install
  194. cd frontend && npm install && cd ..
  195. # development
  196. npm run dev
  197. # production build
  198. npm run build
  199. ```
  200. Bundles land in `src-tauri/target/release/bundle/`.
  201. On Windows, `npm run dev` sets the Cargo build directory to Medium integrity so executables do not inherit a sandbox checkout's Low label and fail to write AppData with a SQLite read-only error. This preflight only adjusts build output, preserving database and checkout permissions.
  202. ---
  203. <a id="quickstart"></a>
  204. ## 🚀 Quick Start
  205. 1. Read the Memory, Skill, and MCP introduction on first launch, or reopen it from Settings. Memory Center is the default page.
  206. 2. Add a project detail or preference manually. For automatic extraction, configure and verify a memory model in Settings, then check capture and injection.
  207. 3. Read a local Skill and publish it to selected clients. Review deployment and differences in Published.
  208. 4. Import or add an MCP service, check its command or URL, select clients, and inspect tools and call records.
  209. No CLI import or process launch is required.
  210. ---
  211. <a id="llm-config"></a>
  212. ## 🤖 LLM Configuration
  213. MCP parsing and memory extraction need an LLM. Add one under **Settings → Memory Extraction**:
  214. | Field | Meaning | Example |
  215. |-------|---------|---------|
  216. | Name | Anything you like | DeepSeek |
  217. | Base URL | Any OpenAI-compatible API | `https://api.deepseek.com/v1` |
  218. | API Key | Provider API key | `sk-xxx` |
  219. | Model | Model name | `deepseek-chat` |
  220. Hit **Test Connection** — green means good.
  221. **Built-in presets**: DeepSeek (`deepseek-chat`), OpenAI (`gpt-4o-mini`), or any custom OpenAI-compatible endpoint. Keys are encrypted at rest (AES-256-GCM, machine-bound).
  222. **Local models (Ollama)**: the settings page has a built-in Ollama module — connect, list pulled models with size / quantization, and promote one to a custom provider or the memory-extraction model in one click.
  223. ---
  224. <a id="data-paths"></a>
  225. ## 📁 Data Storage Paths
  226. | Data | Windows | macOS / Linux |
  227. |------|---------|---------------|
  228. | Legacy runtime configs (preserved, unused) | `%APPDATA%\agent-manager\agents.json` | `~/Library/Application Support/agent-manager/` (macOS), `~/.local/share/agent-manager/` (Linux) |
  229. | LLM providers | `%APPDATA%\agent-manager\llm_config.json` | same directory |
  230. | Memory & telemetry DB | `%APPDATA%\agent-manager\telemetry.sqlite3` | same directory |
  231. | Agent source overrides | `%APPDATA%\agent-manager\agent_source_paths.json` | same directory |
  232. | Historical Caddyfile (unused) | `%APPDATA%\agent-manager\Caddyfile` | same directory |
  233. | Claude Desktop MCP config | `%APPDATA%\Claude\claude_desktop_config.json` | `~/Library/Application Support/Claude/claude_desktop_config.json` |
  234. The Settings page offers one-click **export / import** of a full config backup for migration and disaster recovery.
  235. ---
  236. <a id="telemetry"></a>
  237. ## 🔌 Telemetry & External Access (experimental)
  238. The memory pipeline uses local SQLite as its event ledger: hook events are persisted first and extracted asynchronously, so an offline model never loses captured sessions.
  239. - **Codex, Claude Code, Qoder**: per-Agent hooks install/uninstall from the Memory Center; port and auth follow the external-trigger settings
  240. - **WorkBuddy**: never writes to its Claude Code config (no double capture); can submit aggregated session / usage data via the standard telemetry endpoint
  241. - **Token accounting**: Codex, Claude, and WorkBuddy read real usage from local transcripts; sources without complete provider usage show clearly-marked local estimates. Stray hook events are audit-only and never double-counted; adapters can overwrite estimates with real final numbers via `session_usage`
  242. Standard endpoint: `POST http://127.0.0.1:<hook-port>/telemetry/events/{codex|workbuddy|claude|qoder}`
  243. ```json
  244. {
  245. "session_id": "stable-session-id",
  246. "event": "session_usage",
  247. "cwd": "D:/Github-repo/example",
  248. "usage": { "input_tokens": 120, "output_tokens": 48, "cached_tokens": 60 },
  249. "usage_scope": "session"
  250. }
  251. ```
  252. `input_tokens` must be the session's full input total; `cached_tokens` is display-only detail (do not add it into totals). Re-submissions for the same `source + session_id` overwrite instead of accumulate.
  253. <a id="whats-new"></a>
  254. ## 🗺️ What's New
  255. > **Current version `v1.1.0`**
  256. **v1.1.0**
  257. - ✅ **Cloud vault sync** — WebDAV / S3-compatible storage join the built-in vault server for encrypted memory sync
  258. - ✅ **88API built-in provider** — sponsor entry bundled, model lists fetched automatically, no manual entry for new models
  259. - ✅ **Cancellable memory extraction** — pending extractions can be cancelled in the Pending Memories panel; failed sessions can still re-run
  260. - ✅ **Light theme by default** — fresh installs start in light mode; dark mode remains available in settings
  261. - ✅ **Feature-focus slimming** — process launcher, embedded terminal / Web UI, collaboration tasks, and agent publishing retired; memory, settings, and historical artifacts are fully preserved
  262. **v1.0.x**
  263. - ✅ **Project renamed to Vibe Assistant** (智管-Vibe Assistant) — display name, prompts, docs, and window title; protocol identifiers and data paths unchanged, so existing hooks, MCP installs, and local data keep working
  264. - ✅ **Environment Variables manager** — user/system scopes, PATH edited one entry per line, `REG_EXPAND_SZ` previews, protected critical variables
  265. - ✅ **Scheduled L2/L3 memory refresh** — configurable intervals (default L2 7 days / L3 30 days), model calls only when new memories exist, L3 auto-publish
  266. - ✅ **Network Manager** — live per-interface throughput, public-IP (direct + proxy egress) and dual-path bandwidth speed tests
  267. - ✅ **Skill Marketplace** — install curated skills from official OpenAI / Anthropic repositories into the shared Skill library
  268. - ✅ **Workflow feature removed** — the visual workflow builder, task board, and run history were retired to keep the app focused (see release notes for history)
  269. - ✅ **L1 extraction hardening** — truncated-JSON salvage, contract-drift tolerance, and model-output excerpts in error messages for diagnosable failures
  270. 📖 [Full changelog](https://github.com/Zafer-Liu/Vibe_Assistant/releases)
  271. ---
  272. <a id="faq"></a>
  273. ## ❓ FAQ
  274. <details>
  275. <summary><b>🧬 Memory Center</b></summary>
  276. <details>
  277. <summary><b>A session never became a memory?</b></summary>
  278. Check in order:
  279. 1. Memory extraction needs an LLM: confirm at least one provider passes **Test Connection** in **Settings → Memory Extraction** (or an Ollama model is attached)
  280. 2. Open **Memory Center → Pending Memories** — is the session queued? What is the failure reason? (Failures quote the model's raw output for diagnosis.)
  281. 3. Confirm the Agent's capture hook is installed, or its transcript directory is configured correctly
  282. 4. Failed sessions can be re-run from the panel
  283. </details>
  284. <details>
  285. <summary><b>Edits to the L3 Profile don't take effect?</b></summary>
  286. L3 uses a draft → publish gate: a draft must be **published** before it is injected. Published profiles can be edited directly — changes apply immediately. If retrieval via MCP returns stale content, confirm the host's shared-memory MCP is connected.
  287. </details>
  288. <details>
  289. <summary><b>How often are L2 / L3 refreshed?</b></summary>
  290. On schedule, without any clicks: every 30 minutes the app checks whether new memories arrived and whether the configured interval (default L2 7 days, L3 30 days) elapsed. Adjust or disable under **Settings → Scheduled memory refresh**. Every run is logged on the Memory Injection page.
  291. </details>
  292. <details>
  293. <summary><b>Cloud Vault sync conflict?</b></summary>
  294. Local SQLite is always authoritative; a unreachable vault never blocks injection. On conflict the app asks which side to keep. The server only stores AES-256-GCM ciphertext and cannot read or alter content.
  295. </details>
  296. </details>
  297. <details>
  298. <summary><b>⚙️ Install & Runtime</b></summary>
  299. <details>
  300. <summary><b>Windows: "Unknown publisher" warning?</b></summary>
  301. Click **More info → Run anyway** — the installer is not Microsoft code-signed.
  302. </details>
  303. <details>
  304. <summary><b>macOS: "can't be opened because it is from an unidentified developer"?</b></summary>
  305. ```bash
  306. xattr -d com.apple.quarantine /Applications/智管-Vibe\ Assistant.app
  307. ```
  308. Or right-click the app → **Open** → **Open** again.
  309. </details>
  310. <details>
  311. <summary><b>npm run dev says the port is in use?</b></summary>
  312. The dev server uses port **1420**. Find and kill the holder:
  313. ```powershell
  314. netstat -ano | findstr :1420
  315. taskkill /PID <PID> /F
  316. ```
  317. </details>
  318. <details>
  319. <summary><b>I renamed/updated the app and the shared-memory MCP stopped working?</b></summary>
  320. A renamed install changes the executable path recorded in Agents' MCP configs. Open **Memory Center → Memory Injection** and reconnect the affected hosts once — after that, regular in-place updates keep working.
  321. </details>
  322. </details>
  323. ---
  324. <a id="contributing"></a>
  325. ## 🤝 Contributing
  326. PRs and issues are welcome:
  327. 1. **Fork** the repository
  328. 2. Create a feature branch (`git checkout -b feature/amazing-feature`)
  329. 3. Commit (`git commit -m 'feat: add amazing feature'`)
  330. 4. Push (`git push origin feature/amazing-feature`)
  331. 5. Open a **Pull Request**
  332. Bugs and ideas: [Issues](https://github.com/Zafer-Liu/Vibe_Assistant/issues). For a dev setup see [Build from source](#install).
  333. ---
  334. <a id="license"></a>
  335. ## 📄 License
  336. [Apache 2.0](LICENSE)
  337. ---
  338. <div align="center">
  339. ## ⭐ The Goal
  340. **Keep useful context, reusable skills, and connected tools in one workspace.**
  341. </div>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.