Skip to content

armgit / avibe

Public Imported from GitHub
0

Make your computer a home for your AI partner. Avibe is the local-first Agent OS: Vibey, its built-in AI partner, works out of the box, and Claude Code, Codex & OpenCode run unchanged beside it — reach them from your browser, phone, or any chat app.

docs(readme): refresh highlights for Model Hub, voice, and sharing (#2288)

Call White604285c1ce38
diff --git a/README.md b/README.md
index 7f93475fe..aa5af83c4 100644
--- a/README.md
+++ b/README.md
@@ -4,16 +4,15 @@
# Avibe
-### The local-first Agent OS — your AI partner lives on your own machine.
+### The local-first Agent OS — your own AI partner.
-**Own the agent. Reach it from anywhere. Never get locked in.**
+**Every agent. Every subscription. Running on your machine, steered from your pocket.**
[![GitHub Stars](https://img.shields.io/github/stars/avibe-bot/avibe?color=ffcb47&labelColor=black&style=flat-square)](https://github.com/avibe-bot/avibe/stargazers)
-[![Python](https://img.shields.io/badge/python-3.9%2B-3776AB?labelColor=black&style=flat-square)](https://www.python.org/)
+[![Release](https://img.shields.io/github/v/release/avibe-bot/avibe?labelColor=black&style=flat-square)](https://github.com/avibe-bot/avibe/releases/latest)
+[![Python](https://img.shields.io/badge/python-3.10%2B-3776AB?labelColor=black&style=flat-square)](https://www.python.org/)
[![License](https://img.shields.io/badge/license-MIT-green?labelColor=black&style=flat-square)](LICENSE)
-<a href="https://www.producthunt.com/products/vibe-remote?embed=true&utm_source=badge-featured&utm_medium=badge&utm_campaign=badge-vibe-remote" target="_blank" rel="noopener noreferrer"><img alt="Avibe — the local-first Agent OS | Product Hunt" width="250" height="54" src="https://api.producthunt.com/widgets/embed-image/v1/featured.svg?post_id=1104967&theme=light&t=1774450119248"></a>
-
[Docs](https://docs.avibe.bot) · [English](README.md) · [中文](README_ZH.md)
**Drives** ![Claude Code](https://img.shields.io/badge/Claude%20Code-D4A27F?style=flat-square&logo=anthropic&logoColor=white) ![OpenCode](https://img.shields.io/badge/OpenCode-00B4D8?style=flat-square) ![Codex](https://img.shields.io/badge/Codex-412991?style=flat-square)
@@ -28,23 +27,24 @@
---
-## Your AI agent is brilliant — and stuck
+## Your agent is a genius — chained to a terminal
-Claude Code, Codex, OpenCode are incredible. But:
+Claude Code, Codex, and OpenCode are absurdly capable. But:
-- 🖥️ **Trapped on one machine.** Your agent lives in a terminal. Close the laptop and it stops.
-- 📵 **Out of reach.** Away from your desk, you can't see what it's doing — let alone steer it.
-- 🔒 **Locked in.** Every tool wants to be the whole stack: its app, its cloud, its subscription, your code uploaded to someone else's box.
+- 🖥️ **Close the lid, lose the agent.** It lives in one terminal window. Walk away and the work stops with you.
+- 📵 **Out of sight, out of control.** Away from your desk, you can't see what it's doing — let alone steer it.
+- 💸 **Paying for several plans, stuck on one.** One hits its cap mid-task while another sits idle, and you have no idea which one is earning its keep.
+- 🔒 **Every tool wants to own you.** Its app, its cloud, its subscription — and your code on someone else's box.
-## Avibe flips that
+## Avibe sets it free
-**One command turns your own machine into the home your AI partner lives in.** You drive the *official* Claude Code, Codex, and OpenCode — from a browser or any chat app — while your code and keys stay on your machine, and avibe.bot never sees your data.
+**One command, and your machine becomes the home your AI partner lives in.** Drive the *official* Claude Code, Codex, and OpenCode from a browser, your phone, or the chat app you already have open. Your Claude, ChatGPT, Gemini, Kimi, and xAI subscriptions and your API keys all land behind one local gateway. Your code, keys, and agent processes stay on your machine — `avibe.bot` signs you in and opens a secure tunnel, and never touches your workspace.
```bash
bash -o pipefail -c 'curl -fsSL https://avibe.bot/install.sh | bash -s -- --launch'
```
-The browser opens, you follow a short wizard, and your machine becomes an Agent OS you can reach from anywhere.
+The browser pops open. A three-step wizard wires up your models, finds your agents (one click installs any that are missing), and drops you straight into the Workbench.
> Open source — read the [install script](https://github.com/avibe-bot/avibe/blob/master/install.sh) first if you like. The short URL is a 307 redirect to that file.
@@ -60,68 +60,81 @@ We recommend WSL on Windows for the best compatibility — see [Run Avibe with W
## What you get
-### 💬 A Workbench, not another chat box
+### 💬 Not a chat box. A whole Workbench.
-Chat, inspect files, edit code, and run terminal commands in one windowed workspace. Built-in apps and the Show Pages your agent creates open side by side, so the browser feels like an Agent OS instead of a settings dashboard.
+Files, editor, and terminal open as windows beside your chat, all in one browser tab — with the apps and Show Pages your agent builds opening right next to them. It stops feeling like a web page and starts feeling like an operating system.
<img src="assets/screenshots/v4/apps-library-en.png" alt="Avibe's multi-window Workbench with Files, Terminal, and Editor open together" />
-### 🧠 Its own timeline — the Agent Harness
+### 🔀 Model Hub — every plan you pay for, finally on the same team
-Most AI tools only act when you type. Avibe gives your agent durable primitives — **run, schedule, watch, and inspect** — so it can start work, wait for the right moment, run in the background, and come back with results. Task and watch details expose their trigger, Agent and session binding, delivery configuration, and run history; standalone definitions can intentionally omit a conversation delivery target.
+Sign in each subscription and add each API key once — official vendors, relays, aggregators, self-hosted endpoints, all welcome. Switch an agent to gateway mode and Avibe's local gateway routes its models through them in the order you set.
-<img src="assets/screenshots/v4/harness-tasks-en.png" alt="Agent Harness task list with schedule, Agent ownership, session, and delivery details" />
+- **Set up once, use it in every agent.** One gateway-held sign-in or key feeds Claude Code, Codex, and OpenCode at the same time — no copying credentials into three configs. (A subscription you leave to its own CLI serves only that CLI.)
+- **Mix and match models.** GPT inside Claude Code, Claude inside Codex — put any gateway-held source's models on any agent's list, and the gateway translates between protocols.
+- **Plan runs dry? The next one picks it up.** If a source hits a quota, rate limit, or network failure before it starts answering, the next source in your route takes the request — no retry, no babysitting. Once your first choice recovers, the next turn goes back to it.
+- **Route it your way.** Give each agent a default source order, then pin any single model to its own route chain.
+- **Move in with one click.** Already signed in to a supported CLI? Bring that login over with your explicit OK. Credentials and routes stay on your machine.
-<img src="assets/screenshots/v4/harness-watches-en.png" alt="Agent Harness watch list with an expanded watch detail panel showing its command, Agent, session, delivery, and follow-up message" />
+*Four sources feeding three agents: GPT and Claude models mixed in every agent, and relay-gpt already covering for Codex's unavailable subscription.*
-### 🤖 Bring the agents you already trust
+<img src="assets/screenshots/v4/model-hub-routes-en.png" alt="Model Hub routes view: four upstream sources wired to Claude Code, Codex, and OpenCode, each agent mixing GPT and Claude models, with one Codex model taken over by a backup source" />
-Run the official Claude Code, Codex, and OpenCode CLIs behind one Agent registry. Choose models and reasoning per Agent, route projects to the right specialist, and keep reusable Skills shared across backends.
+**Know exactly what your plans are worth.** Usage trends by model and source, and your usage priced at official API rates. For Claude and ChatGPT subscriptions, add live 5-hour and weekly quotas — down to how many times over each plan has already paid for itself.
-<img src="assets/screenshots/v4/agents-en.png" alt="Avibe Agents registry with Claude Code, Codex, and OpenCode backends" />
+*One real day on the maintainer's machine: ~680M tokens, worth ~$540 at official API prices.*
-Open **Runs** to see that collaboration as a graph: who started each background Session, where it reports back, and the execution history behind every node.
+<img src="assets/screenshots/v4/model-hub-usage-en.png" alt="Model Hub usage view with 24-hour token totals, request count, cache share, API-price value, and an hourly usage trend" />
-<img src="assets/screenshots/v4/agents-graph-en.png" alt="Avibe Agent run graph showing a parent Session delegating work to Claude Code, Codex, and OpenCode background Sessions" />
+### ⏰ It works while you sleep — the Agent Harness
+
+Most AI tools freeze the moment you stop typing. Avibe gives your agent four durable primitives — **run, schedule, watch, and inspect** — so it starts work on its own, waits for the right moment, grinds away in the background, and comes back when there's something worth your attention. Scheduled shell commands stay silent when they pass; a failure leaves a notice — or, if you opt in, goes straight to an Agent.
-### 🎨 Show Pages — point, comment, and iterate
+No flags to learn. Just say it:
+
+- *"Watch this PR and come back when there's actionable review feedback."*
+- *"Run the deployment check every weekday morning and post the summary here."*
+- *"Start a separate investigation session for this incident, but report the conclusion to this channel."*
+- *"If CI fails, summarize the logs; if it passes, tell me whether the PR is mergeable."*
-When a picture beats a paragraph, your agent hands you a live web page — a flowchart, dashboard, diff, report, or small app. Comment on an element or screenshot, and the agent can rework the page or answer exactly where you pointed.
+<img src="assets/screenshots/v4/harness-tasks-en.png" alt="Agent Harness task list with schedule, Agent ownership, session, and delivery details" />
-In Chat, click **Visualize** to switch the current session to its Show Page. Hover the button to open the page in a new app window or browser tab. You can also drag the button downward and release to place a new app window at the pointer, or drop it on **Apps** to pin that Show Page to the Dock.
+### 🤖 The real agents, not knock-offs
-See [Show Pages](docs/SHOW_PAGES.md) for synchronous API timeout behavior and
-guidance for longer background refreshes.
+Avibe runs the official Claude Code, Codex, and OpenCode CLIs — the same ones you'd install yourself — behind one Agent registry. Give each Agent its own model and reasoning effort, and hand each project or channel to the right specialist. Write a Skill once; all three backends load it the same way.
-<img src="assets/screenshots/v4/show-page-en.png" alt="Show Page review with anchored comments and Agent replies on the page" />
+When your agents start delegating to each other, **Runs** draws the whole chain as a graph: who started each background Session, where it reports back, and the history behind every node.
-### 🔐 Vaults — secret values stay out of Vault responses
+<img src="assets/screenshots/v4/agents-graph-en.png" alt="Avibe Agent run graph showing a parent Session delegating work to Claude Code, Codex, and OpenCode background Sessions" />
-Add a **Standard** secret once and approve Agent requests from the browser. Avibe's Vault responses never include the secret value; commands receiving it must still avoid printing it. **Protected**, passkey-gated custody remains pre-launch until the sandbox integrity gate is enforced.
+### 🎨 Show Pages — your agent answers with a web page
-### 🌐 Browser and messaging are equal surfaces
+Ask for a dashboard, flowchart, diff, report, or small app, and get a live page instead of a wall of text. Click an element, box a region, or mark up a screenshot with numbered notes, say what you want, and the agent reworks the page or answers right where you pointed.
-Use the Workbench when you want windows and rich interaction; use Slack, Discord, Telegram, WeChat, or Lark / Feishu when chat is the fastest route. They reach the same Agents and sessions.
+Keep a page private, or pair with `avibe.bot` to share it with specific people or publish it at a link you'll actually remember. Pin your favorites to the Dock and they open like apps.
-<img src="assets/screenshots/v4/platforms-en.png" alt="Avibe messaging platform connections for Slack, Discord, Telegram, WeChat, and Lark" />
+<img src="assets/screenshots/v4/show-page-en.png" alt="Show Page review with anchored comments and Agent replies on the page" />
-### 📱 In your pocket
+### 📱 Leave your desk. Keep your agent.
<img src="assets/screenshots/v4/workbench-mobile-en.png" alt="Avibe Workbench on mobile" width="270" align="right" />
-Your machine runs the work; you don't have to sit in front of it. Install the Workbench as a desktop or mobile app, use built-in voice-to-text, and get a push notification when a job needs you. Run `vibe remote` and the same local Workbench becomes reachable from any browser through a secure `avibe.bot` tunnel — no VPN or port forwarding.
-
-You're on a plane, at a café, on a borrowed laptop. The agent pings that a job needs you. Open the link, steer it, walk away again.
+Your machine does the work; you don't have to babysit it. Open the Workbench when you want windows, or fire off a message in Slack, Discord, Telegram, WeChat, or Lark / Feishu when that's faster — same Agents, same machine.
-- 🌍 **Your own `you-app.avibe.bot`** — 30-second sign-in, your slug for life
-- 🔒 **Fail-closed at every join** — auth, routing, and host checks default to "deny"
-- 📱 **Mobile-aware UI** — thumb-friendly, built for borrowed screens
+- 🔔 **It taps you on the shoulder.** Install the Workbench on your phone or desktop and get a push the moment a job needs you.
+- 🎙️ **Talk, don't type.** Pair with `avibe.bot` and realtime voice input writes as you speak, then tidies the transcript in place. On desktop, press ⌥Z (Alt+Z on Windows and Linux).
+- 🌍 **Your own `you-app.avibe.bot`.** Pair once with `vibe remote` and your Workbench opens in any browser, anywhere — no VPN, no port forwarding.
+- 🔒 **Your door, your guest list.** Only the people your access policy allows can sign in — by default, just you. Auth, routing, and host checks all fail closed.
-**Your data plane stays on your machine**; `avibe.bot` only carries the control-plane handshake.
+On a plane. At a café. On a borrowed laptop. Your agent pings you; you open the link, steer, and walk away again.
<br clear="all"/>
-**Plus** — per-channel Agent routing · resumable sessions (thread = session) · interactive prompts · voice-to-text · file attachments · push and completion notifications.
+### 🔐 Vaults — secrets by name, never by value
+
+Add an API key or token once — or let an Agent request a missing one for you to fill in from the browser. Agents use it by name, and Avibe hands it to the command, authenticated request, or signing operation that needs it. Vault responses never include the secret value; commands receiving it must still avoid printing it. Passkey-protected custody stays pre-launch until the sandbox integrity gate is enforced.
+
+**Plus** — thread = session, picks up where you left off · interactive buttons and forms when the agent needs a decision · rich attachments with in-app preview and inline audio · keyboard shortcuts · one-click backend updates.
---
@@ -129,32 +142,32 @@ You're on a plane, at a café, on a borrowed laptop. The agent pings that a job
| | |
|---|---|
-| **Local-first, and yours** | Your AI partner, its execution, your keys, and your data stay on your machine. `avibe.bot` only issues identity and a secure tunnel — it never proxies your data. |
-| **One substrate, every first-party agent** | Drive the *official* Claude Code, Codex, and OpenCode. Bring your own subscription or keys, switch per task, and never get locked into one vendor's silo. |
-| **Browser and chat, both first-class** | Operate from the browser Workbench, or from Slack, Discord, Telegram, WeChat, and Lark / Feishu. Same agent, same sessions. |
-| **No middleman** | No extra reasoning loop sits between you and your agent. Tokens go straight to the agent you chose. |
+| **Local-first, actually yours** | Your AI partner, its execution, your keys, and your code stay on your machine. `avibe.bot` issues identity and a secure tunnel; it never proxies your workspace. |
+| **Every first-party agent, one home** | Drive the *official* Claude Code, Codex, and OpenCode. Switch per task, per project, or per channel — no vendor silo. |
+| **Every subscription, one team** | Model Hub pools your plans and keys behind one gateway, lets agents mix models, fails over on its own, and shows what each one is really worth. |
+| **Browser and chat, both first-class** | Workbench, phone, Slack, Discord, Telegram, WeChat, and Lark / Feishu — same agents, same machine. |
+| **No middleman tax** | No extra reasoning loop sits between you and your agent. Every token goes straight to the agent you chose. |
---
## How it works
```
-┌──────────────┐ ┌──────────────┐ ┌──────────────┐
-│ You │ browser │ │ stdio │ Claude Code │
-│ (anywhere) │ Slack │ Avibe │ ─────────▶ │ OpenCode │
-│ │ Discord │ (your machine)│ ◀───────── │ Codex │
-│ │ Telegram │ │ │ │
-│ │ WeChat │ │ │ │
-│ │ Lark │ │ │ │
-└──────────────┘ └──────────────┘ └──────────────┘
+ You, anywhere Your machine Your models
+┌──────────────┐ ┌──────────────────────────────────────┐ ┌──────────────┐
+│ Browser │ │ Avibe │ │ Subscriptions│
+│ Phone (PWA) │ ───▶ │ Claude Code ─┐ │ │ API keys │
+│ Slack │ │ Codex ─┼─▶ Model Hub ────────┼────▶ │ Relays │
+│ Discord … │ ◀─── │ OpenCode ─┘ │ │ Self-hosted │
+└──────────────┘ └──────────────────────────────────────┘ └──────────────┘
```
-1. **You type** — in the browser or a chat app: *"Add dark mode to the settings page."*
-2. **Avibe routes** to your configured agent, in the right project.
-3. **The agent** reads your local codebase, writes code, and streams back.
-4. **You review** in the same surface, iterate in the thread, and resume later from anywhere.
+1. **You ask** — in the browser or a chat app: *"Add dark mode to the settings page."*
+2. **Avibe routes** the message to the right Agent, in the right project.
+3. **The agent** reads your local codebase, writes code, and streams its work back. In gateway mode, Model Hub picks the source for each call; in direct mode, the agent uses its own login.
+4. **You review** in the same surface, iterate in the thread, and pick it up later from anywhere.
-**Your code stays on your machine.** Avibe runs locally and connects out via Slack Socket Mode, Discord Gateway, Telegram long-polling, WeChat polling, or Lark WebSocket — no public inbound ports for normal chat control. Prompts go only to the AI provider you choose.
+Avibe connects out via Slack Socket Mode, Discord Gateway, Telegram long-polling, WeChat polling, or Lark WebSocket — no public inbound ports for chat control. Agent prompts go only to the model sources you configure.
---
@@ -163,79 +176,60 @@ You're on a plane, at a café, on a borrowed laptop. The agent pings that a job
| | Avibe | OpenClaw |
|---|---|---|
| **Setup** | One command + web wizard. Done in minutes. | Gateway + channels + JSON config. Expect an afternoon. |
-| **Security** | Local-first. Socket Mode / WebSocket only. No public inbound ports, minimal attack surface. | Gateway exposes ports. More moving parts, more surface. |
+| **Security** | Local-first. Outbound Socket Mode / WebSocket / long polling only. No public inbound ports, minimal attack surface. | Gateway exposes ports. More moving parts, more surface. |
| **Token cost** | No extra reasoning loop in between. Tokens go straight to your chosen agent. | Every message carries a long persona/orchestration context. Tokens burn on overhead before your task starts. |
-| **Lock-in** | Drives the official agent CLIs; bring your own keys; switch per task. | Tied to its own assistant loop. |
+| **Lock-in** | Drives the official agent CLIs; bring your own subscriptions and keys; switch per task. | Tied to its own assistant loop. |
-OpenClaw is an always-on personal assistant — great for casual chat, expensive for real work. Avibe is a **local-first Agent OS** for the agents you already trust: the agent stays itself, your data stays local, and the colleague experience comes from putting the agent into the flow where your work already happens.
+OpenClaw is an always-on personal assistant. Avibe is the **Agent OS** for the agents you already trust: the agent stays itself, your data stays local, and the colleague experience comes from putting the agent where your work already happens.
---
## FAQ
+<details>
+<summary><b>Can I use my existing Claude or ChatGPT subscription?</b></summary>
+
+Yes. Claude Code and Codex can keep their own logins, or you can add supported subscriptions to Model Hub as gateway sources to get quota tracking, usage analytics, and automatic takeover. Each vendor's terms still apply; Avibe warns you before a sign-in that may carry account restrictions. See [Model Hub](https://docs.avibe.bot/concepts/model-hub).
+</details>
+
<details>
<summary><b>Does it run local models?</b></summary>
-Local-first here means your **code, data, and execution** stay on your machine — not the model weights. Avibe drives the agent you configure (Claude Code, Codex, OpenCode); OpenCode can point at local or OpenAI-compatible endpoints if you want inference local too.
+Local-first here means your **code, data, and execution** stay on your machine — not necessarily the model weights. Model Hub can point at any compatible self-hosted endpoint if you want inference local too.
</details>
<details>
<summary><b>Where do my code and data go?</b></summary>
-Your code, keys, and agent processes stay on your own machine. `avibe.bot` only issues identity and a secure tunnel — it never proxies or stores your data. Prompts go only to the AI provider you chose.
+Your code, keys, and agent processes stay on your machine. Agent prompts go to the model sources you configure. `avibe.bot` handles sign-in and the Remote Web UI tunnel without proxying your workspace; Voice input is the exception that goes through `avibe.bot` for transcription. The [boundary inventory](https://docs.avibe.bot/concepts/local-first#what-leaves-your-machine) lists every network path that carries your content.
</details>
<details>
<summary><b>Do I have to pay for Avibe?</b></summary>
-Avibe is open source (MIT) and free to run. You bring your own agent subscription or API keys and pay your provider directly — no markup, no second subscription.
+Avibe is open source (MIT) and free to run. You bring your own agent subscriptions or API keys and pay your provider directly — no markup, no second subscription.
</details>
<details>
<summary><b>Which agents and platforms are supported?</b></summary>
-**Agents:** the official Claude Code, Codex, and OpenCode CLIs. **Surfaces:** a built-in browser Workbench plus Slack, Discord, Telegram, WeChat, and Lark / Feishu.
+**Agents:** the official Claude Code, Codex, and OpenCode CLIs. **Surfaces:** a built-in browser Workbench (installable on desktop and mobile) plus Slack, Discord, Telegram, WeChat, and Lark / Feishu.
</details>
<details>
<summary><b>Is remote access secure?</b></summary>
-`vibe remote` opens a Cloudflare tunnel; browser traffic reaches your machine only after sign-in. Auth, routing, and host checks are **fail-closed**, and there are no public inbound ports for normal chat control.
+`vibe remote` runs a Cloudflare tunnel from your machine; browser traffic reaches it only after sign-in (except Show Pages you choose to publish), and only for the people your access policy allows (by default, just you). Auth, routing, and host checks are **fail-closed**, and there are no public inbound ports for chat control.
</details>
<details>
<summary><b>How is Avibe different from OpenClaw or Hermes?</b></summary>
-OpenClaw and Hermes are *agents* — a gateway-style assistant and a self-improving agent. Avibe is a different layer: the **Agent OS**. It gives any agent a unified world model — agents, sessions, Show Pages, the Harness — so it can schedule itself, build its own loops, and reach you through a real interaction layer, then runs the official Claude Code, Codex, and OpenCode you bring (an Avibe-native agent is [on the roadmap](#roadmap)). See the [comparison table](#avibe-vs-openclaw) above for OpenClaw specifics.
+OpenClaw and Hermes are *agents* — a gateway-style assistant and a self-improving agent. Avibe is a different layer: the **Agent OS**. It gives every agent it runs a unified world model — agents, sessions, Show Pages, the Harness, Model Hub — so it can schedule itself, build its own loops, and reach you through a real interaction layer, then runs the official Claude Code, Codex, and OpenCode you bring (an Avibe-native agent is [on the roadmap](#roadmap)). See the [comparison table](#avibe-vs-openclaw) above for OpenClaw specifics.
</details>
---
-## Talk to it like a colleague
-
-Ask in plain language and the Harness composes the commands behind the scenes:
-
-- *"Watch this PR and come back when there's actionable review feedback."*
-- *"Run the deployment check every weekday morning and post the summary here."*
-- *"Start a separate investigation session for this incident, but report the conclusion to this channel."*
-- *"If CI fails, summarize the logs; if it passes, tell me whether the PR is mergeable."*
-
-**Switch agents mid-conversation** — just prefix your message:
-
-```
-Plan: design a new caching layer for the API
-```
-
-**Route per project** — different work, different agent:
-
-```
-frontend → OpenCode (fast iteration)
-backend → Claude Code (complex logic)
-prototypes → Codex (quick experiments)
-```
-
----
-
## Meet Vibey
<div align="center">
@@ -256,12 +250,14 @@ Forgets nothing. Holds opinions. Says thanks when you fix its bugs.
vibe # Start Avibe and open the Workbench
vibe status # Check service and configuration status
vibe stop # Stop the local service
-vibe doctor # Diagnose common setup issues; use "vibe doctor repair" for explicit safe fixes
+vibe upgrade # Upgrade to the latest release
+vibe doctor # Diagnose common setup issues; "vibe doctor repair" applies safe fixes
vibe remote # Reach your Workbench from any device via avibe.bot
-vibe agent # Run and manage Avibe agents
+vibe agent # Run and manage Avibe Agents
vibe task # Schedule time-based work (cron / one-off)
vibe watch # Wait on a condition, then act
vibe runs # Inspect agent run history
+vibe show # Create, inspect, and publish Show Pages
vibe vault # Manage secrets, requests, authenticated fetches, and signing
```
@@ -269,43 +265,43 @@ vibe vault # Manage secrets, requests, authenticated fetches, and signing
|---|---|
| Mention the bot | Start a task or ask a question |
| Reply in thread | Continue the same agent session |
-| `/stop` | Stop the current session |
+| `/stop` | Interrupt the current run |
Full references: [Commands](docs/COMMANDS.md) · [CLI](docs/CLI.md)
---
-## Prerequisites
+## Agents
-You need at least one agent installed:
+The setup wizard detects Claude Code, Codex, and OpenCode on your machine and installs any you choose that are missing. To install one yourself:
<details>
-<summary><b>OpenCode</b> (recommended)</summary>
+<summary><b>Claude Code</b></summary>
```bash
-curl -fsSL https://opencode.ai/install | bash
-```
-
-Add to `~/.config/opencode/opencode.json` to skip permission prompts:
-
-```json
-{ "permission": "allow" }
+bash -o pipefail -c 'curl -fsSL https://claude.ai/install.sh | bash'
```
</details>
<details>
-<summary><b>Claude Code</b></summary>
+<summary><b>Codex</b></summary>
```bash
-npm install -g @anthropic-ai/claude-code
+npm install -g @openai/codex
```
</details>
<details>
-<summary><b>Codex</b></summary>
+<summary><b>OpenCode</b></summary>
```bash
-npm install -g @openai/codex
+bash -o pipefail -c 'curl -fsSL https://opencode.ai/install | bash'
+```
+
+OpenCode may pause tool calls for approval unless `~/.config/opencode/opencode.json` allows them; the setup wizard can set this for you:
+
+```json
+{ "permission": "allow" }
```
</details>
@@ -315,20 +311,27 @@ npm install -g @openai/codex
- **Local-first** — Avibe runs on your machine; your code and agent processes stay there.
- **No public inbound ports** — Socket Mode / WebSocket / long-polling only for chat control.
-- **Your keys, your data** — stored under `~/.avibe/`, sent only to the AI provider you chose. Existing installs keep `~/.vibe_remote/` as a compatibility path.
-- **Fail-closed remote access** — `avibe.bot` only brokers identity and the tunnel, never your data.
+- **Your keys, your data** — stored under `~/.avibe/`; agent prompts go only to the model sources you configure. Existing installs keep `~/.vibe_remote/` as a compatibility path.
+- **Fail-closed remote access** — `avibe.bot` brokers identity and the tunnel, never your workspace.
---
## Uninstall
+The snippet removes the first `vibe` on your `PATH`. If `which -a vibe` lists more than one Avibe launcher, remove the others too.
+
```bash
-vibe stop
avibe_home="${AVIBE_HOME:-$HOME/.avibe}"
avibe_home="${avibe_home/#\~/$HOME}"
+vibe_bin="$(command -v vibe)" # capture Avibe's launcher before uninstalling
+if "$vibe_bin" version 2>/dev/null | grep -Eq '^(avibe-os|vibe-remote) '; then
+ "$vibe_bin" stop
+else
+ echo "Skipping ${vibe_bin}: not Avibe's launcher"; vibe_bin=""
+fi
uv tool uninstall avibe-os
-uv tool uninstall vibe-remote # legacy installs
-vibe_bin="$(command -v vibe)" && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
+uv tool uninstall vibe-remote # legacy install
+[ -n "$vibe_bin" ] && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
rm -rf "$avibe_home/runtime/install-generations"
rm -rf "$avibe_home" ~/.vibe_remote
```
@@ -339,12 +342,13 @@ rm -rf "$avibe_home" ~/.vibe_remote
What's coming next:
+- **A native desktop app** — macOS and Windows builds with signed self-updates are in testing now.
- **Deeper interaction-first workflows** — more direct manipulation, structured decisions, and in-place collaboration between chat, apps, and Show Pages.
- **Protected Vault custody** — finish the immutable sandbox integrity gate before enabling passkey-protected secrets for real users.
- **SaaS mode** — one-click hosted onboarding with a cloud relay, while execution still stays on your own machine.
- **An Avibe-native agent** — a first-party agent tuned for this runtime, alongside the official CLIs you bring.
-Shipped recently: multi-window Apps, Show Page annotations, the Agent Harness, Standard Vaults, voice-to-text, and the Skills manager.
+Shipped recently: Model Hub with automatic takeover, quota tracking, and usage analytics · realtime voice input · Show Page sharing · Skills unified across all three backends · the three-step setup wizard · the Agent Harness and run graph · multi-window Apps · Standard Vaults.
---
@@ -352,7 +356,8 @@ Shipped recently: multi-window Apps, Show Page annotations, the Agent Harness, S
- **[Official Docs](https://docs.avibe.bot)** — quickstart, concepts, platform & agent guides, troubleshooting
- **[What is Avibe](https://docs.avibe.bot/concepts/agent-os)** — the Agent OS model
-- **[CLI Reference](docs/CLI.md)** · **[Commands](docs/COMMANDS.md)**
+- **[Model Hub](https://docs.avibe.bot/concepts/model-hub)** — sources, routes, takeover, usage, and quota
+- **[CLI Reference](docs/CLI.md)** · **[Commands](docs/COMMANDS.md)** · **[Show Pages](docs/SHOW_PAGES.md)**
- **[Install via AI agent](docs/INSTALL_FOR_AI.md)** — hand this to Claude Code, Codex, or OpenCode for guided setup
- **[Slack](docs/SLACK_SETUP.md)** · **[Discord](docs/DISCORD_SETUP.md)** · **[Telegram](docs/TELEGRAM_SETUP.md)** setup guides
@@ -360,8 +365,8 @@ Shipped recently: multi-window Apps, Show Page annotations, the Agent Harness, S
<div align="center">
-**Own the agent. Reach it from anywhere.**
+**Own your agents. Take them everywhere.**
-[Install Now](#avibe-flips-that) · [Docs](https://docs.avibe.bot) · [Report a bug](https://github.com/avibe-bot/avibe/issues) · [Follow @alex_metacraft](https://x.com/alex_metacraft)
+[Install Now](#avibe-sets-it-free) · [Docs](https://docs.avibe.bot) · [Report a bug](https://github.com/avibe-bot/avibe/issues) · [Follow @alex_metacraft](https://x.com/alex_metacraft)
</div>
diff --git a/README_ZH.md b/README_ZH.md
index e047926a3..27ce809ba 100644
--- a/README_ZH.md
+++ b/README_ZH.md
@@ -4,17 +4,16 @@
# Avibe
-### Avibe 是本地优先的 Agent OS——你的 AI 伙伴,住在你自己的机器上。
+### Avibe 是本地优先的 Agent OS——你的专属 AI 伙伴。
-**拥有你的 agent。随处都能找到它。永不被锁死。**
+**所有 Agent,所有订阅,随时随地调用,始终跑在你自己的机器上。**
[![GitHub Stars](https://img.shields.io/github/stars/avibe-bot/avibe?color=ffcb47&labelColor=black&style=flat-square)](https://github.com/avibe-bot/avibe/stargazers)
-[![Python](https://img.shields.io/badge/python-3.9%2B-3776AB?labelColor=black&style=flat-square)](https://www.python.org/)
+[![Release](https://img.shields.io/github/v/release/avibe-bot/avibe?labelColor=black&style=flat-square)](https://github.com/avibe-bot/avibe/releases/latest)
+[![Python](https://img.shields.io/badge/python-3.10%2B-3776AB?labelColor=black&style=flat-square)](https://www.python.org/)
[![License](https://img.shields.io/badge/license-MIT-green?labelColor=black&style=flat-square)](LICENSE)
-<a href="https://www.producthunt.com/products/vibe-remote?embed=true&utm_source=badge-featured&utm_medium=badge&utm_campaign=badge-vibe-remote" target="_blank" rel="noopener noreferrer"><img alt="Avibe — 本地优先的 Agent OS | Product Hunt" width="250" height="54" src="https://api.producthunt.com/widgets/embed-image/v1/featured.svg?post_id=1104967&theme=light&t=1774450119248"></a>
-
-[文档](https://docs.avibe.bot) · [English](README.md) · [中文](README_ZH.md)
+[文档](https://docs.avibe.bot/zh) · [English](README.md) · [中文](README_ZH.md)
**驱动** ![Claude Code](https://img.shields.io/badge/Claude%20Code-D4A27F?style=flat-square&logo=anthropic&logoColor=white) ![OpenCode](https://img.shields.io/badge/OpenCode-00B4D8?style=flat-square) ![Codex](https://img.shields.io/badge/Codex-412991?style=flat-square)
@@ -28,23 +27,24 @@
---
-## 你的 AI agent 很强——但被困住了
+## 你的 Agent 是个天才——却被拴在终端里
-Claude Code、Codex、OpenCode 都很能打。但是:
+Claude Code、Codex、OpenCode 都强得离谱。可是:
-- 🖥️ **困在一台机器上。** 它活在终端里,合上笔记本它就停了。
-- 📵 **够不着。** 离开工位,你连它在干什么都看不到,更别说指挥。
-- 🔒 **被锁死。** 每个工具都想当整个栈:它的 app、它的云、它的订阅,你的代码还得传到别人的盒子里。
+- 🖥️ **合上盖子,它就歇了。** 它活在一个终端窗口里,你一走开,活儿也跟着停。
+- 📵 **离开工位,就失联。** 它在干什么你看不到,更别提指挥它。
+- 💸 **付着好几份订阅,却卡在一份额度上。** 一个套餐干到一半撞墙,另一个在吃灰;哪个真值回票价,你根本说不清。
+- 🔒 **每个工具都想把你圈起来。** 它的 app、它的云、它的订阅——还有你那份跑到别人服务器上的代码。
-## Avibe 把这件事反过来
+## Avibe 给它松绑
-**一条命令,把你自己的机器变成 AI 伙伴的家。** 你驱动的是*官方*的 Claude Code、Codex、OpenCode——从浏览器或任意聊天软件——而代码与密钥都留在你的机器上,avibe.bot 也看不到你的数据。
+**一条命令,你的机器就成了 AI 伙伴的家。** 用浏览器、手机,或者你本来就开着的聊天软件,驱动*官方*的 Claude Code、Codex、OpenCode。你的 Claude、ChatGPT、Gemini、Kimi、xAI 订阅,加上各家 API Key,统统汇进同一个本地网关。代码、密钥和 agent 进程都留在你的机器上——`avibe.bot` 负责登录和安全隧道,从不碰你的工作区。
```bash
bash -o pipefail -c 'curl -fsSL https://avibe.bot/install.sh | bash -s -- --launch'
```
-浏览器自动打开,跟着简短向导走完,你的机器就成了一个随处可达的 Agent OS。
+浏览器自动弹开,三步设置向导帮你接好模型、找到你已有的 agent(缺的点一下就装上),直接把你送进 Workbench。
> 开源——想看可以先读一遍[安装脚本](https://github.com/avibe-bot/avibe/blob/master/install.sh)。短链只是到这个文件的 307 重定向。
@@ -60,65 +60,81 @@ Windows 上推荐用 WSL,兼容性最好——见 [从零用 WSL 跑 Avibe](do
## 你能得到什么
-### 💬 一个真正的 Workbench,不是又一个 Chat 框
+### 💬 不是又一个聊天框,是一整个 Workbench
-在同一个窗口化工作区里对话、查看文件、编辑代码、运行终端命令。内置应用和 Agent 做出的 Show Page 可以并排打开,让浏览器真正像一个 Agent OS,而不是设置面板。
+文件、编辑器、终端开成窗口,和对话并排摆在同一个浏览器标签页里;agent 做出来的应用和 Show Page 就开在旁边。用着用着你会忘了这是网页——它更像一个操作系统。
<img src="assets/screenshots/v4/apps-library-zh.png" alt="Avibe 多窗口 Workbench,同时打开文件、终端和编辑器" />
-### 🧠 它有自己的时间线——Agent Harness
+### 🔀 Model Hub——你付费的每一份订阅,终于能组队干活
-大多数 AI 工具只在你打字时才动。Avibe 给 Agent 几个持久化基础能力——**运行、定时、监听、查历史**——让它能发起工作、等到合适的时机、后台跑完回来汇报。任务和 Watch 的详情会展示触发条件、Agent 与会话绑定、投递配置和运行历史;独立定义也可以有意不设置对话投递目标。
+订阅登录一次、API Key 添加一次——官方厂商、中转站、聚合平台、自建服务,来者不拒。把 Agent 切到网关模式,Avibe 的本地网关就按你排好的顺序,把它的模型请求送过去。
-<img src="assets/screenshots/v4/harness-tasks-zh.png" alt="Agent Harness 定时任务,展示调度、Agent、会话和投递详情" />
+- **配一次,所有 Agent 都能用。** 交给模型网关管理的一次登录、一个 Key,Claude Code、Codex、OpenCode 同时取用——不用把凭据抄进三份配置。(留给 CLI 自己管理的订阅,只供那个 CLI 用。)
+- **模型随便混搭。** 在 Claude Code 里用 GPT,在 Codex 里用 Claude——只要来源交给模型网关管理,它的模型就能加进任何 agent 的列表,协议差异由网关负责转换。
+- **额度见底?下一个已经接上了。** 某个来源在开始回答前撞上额度、限流或网络故障,路由里的下一个来源直接接手这次请求——你不用重试,也不用盯着。首选来源恢复后,下一轮对话自动切回去。
+- **路由想怎么排就怎么排。** 每个 agent 一套默认的来源顺序,任何一个模型还能单独固定自己的路由。
+- **一键搬家。** 已经在支持的 CLI 里登录过?你点头,登录就迁过来。凭据和路由都留在你的机器上。
+
+*四个来源供给三个 agent:每个 agent 都混用 GPT 和 Claude 模型,Codex 的订阅暂不可用,relay-gpt 已经顶上。*
-<img src="assets/screenshots/v4/harness-watches-zh.png" alt="Agent Harness Watch 列表与展开的详情面板,展示命令、Agent、会话、投递和跟进消息" />
+<img src="assets/screenshots/v4/model-hub-routes-zh.png" alt="Model Hub 路由视图:四个上游来源连到 Claude Code、Codex 和 OpenCode,每个 agent 混用 GPT 和 Claude 模型,其中一个 Codex 模型已由备用来源自动接管" />
-### 🤖 带上你已经信任的 Agent
+**每份套餐值多少,一清二楚。** 按模型和来源看用量趋势,按官方 API 价格把用量折成钱。Claude 和 ChatGPT 订阅还能实时盯住 5 小时和每周额度——直接算出每个套餐已经回本几倍。
-在一套 Agent 注册表后面运行官方 Claude Code、Codex 和 OpenCode CLI。为不同 Agent 选择模型和推理强度,把项目路由给合适的专家,同时让可复用的 Skills 跨后端共享。
+*维护者机器上真实的一天:24 小时约 6.8 亿 token,按官方 API 价格折合约 $540。*
-<img src="assets/screenshots/v4/agents-zh.png" alt="Avibe Agent 注册表,统一管理 Claude Code、Codex 和 OpenCode 后端" />
+<img src="assets/screenshots/v4/model-hub-usage-zh.png" alt="Model Hub 用量视图:24 小时 token 总量、请求数、缓存占比、折合 API 价格和按小时的用量趋势" />
-打开**运行**,就能把协作关系看成一张图:谁启动了每个后台 Session、结果向哪里汇报,以及每个节点背后的执行历史。
+### ⏰ 你睡觉,它上班——Agent Harness
-<img src="assets/screenshots/v4/agents-graph-zh.png" alt="Avibe Agent 运行图,展示主 Session 向 Claude Code、Codex 和 OpenCode 后台 Session 委派工作" />
+大多数 AI 工具,你一停手它就停。Avibe 给 agent 四个持久化基础能力——**运行、定时、监听、查历史**——让它能自己开工、等准时机、在后台闷头干完,有值得你看的东西再回来找你。定时 shell 命令成功时一声不吭,失败时留下一条通知——你开启的话,也能直接交给 Agent 处理。
-### 🎨 Show Pages——指着评论,就地迭代
+不用背参数,直接说:
-当一张图胜过一段话,Agent 直接给你一个实时网页——流程图、仪表盘、diff、报告或小应用。你可以对元素或截图原位评论,Agent 也能在你指的位置回答或直接改好页面。
+- *"盯着这个 PR,有可执行的 review 意见了再回来找我。"*
+- *"每个工作日早上跑一次部署检查,把总结发到这里。"*
+- *"为这个故障单独开个调查会话,结论汇报回这个频道。"*
+- *"CI 挂了就总结日志;过了就告诉我这个 PR 能不能合。"*
-在 Chat 中点击**可视化**,即可把当前 Session 切换到它的 Show Page。将鼠标悬停在按钮上,可以改为在新的应用窗口或浏览器标签页中打开;也可以向下拖拽后松开,在指针位置放置一个新窗口,或拖到**应用**按钮上,把该 Show Page 固定到应用栏。
+<img src="assets/screenshots/v4/harness-tasks-zh.png" alt="Agent Harness 定时任务,展示调度、Agent、会话和投递详情" />
+
+### 🤖 正版 Agent,不是山寨复刻
-<img src="assets/screenshots/v4/show-page-zh.png" alt="Show Page 审阅界面,包含原位评论与 Agent 页面内回复" />
+Avibe 运行的是官方 Claude Code、Codex、OpenCode CLI——就是你自己会装的那几个——统一收进一套 Agent 注册表。每个 Agent 单独选模型和推理强度,每个项目、每个频道都交给最合适的专家。Skill 写一次,三个后端用同一种方式加载。
-### 🔐 Vaults——Vault 响应不暴露密钥值
+agent 之间开始互相派活时,**Runs** 把整条协作链画成一张图:谁发起了哪个后台会话、结果汇报给谁、每个节点背后的执行历史。
-添加一次**标准档**密钥,在浏览器里处理 Agent 的授权请求。Avibe 的 Vault 响应不会返回密钥值;接收密钥的命令仍然不得自行打印它。需要 passkey 的**保护档**在 sandbox 完整性门禁启用前仍处于上线前阶段。
+<img src="assets/screenshots/v4/agents-graph-zh.png" alt="Avibe Agent 运行关系图,展示父会话把工作委派给 Claude Code、Codex 和 OpenCode 后台会话" />
-### 🌐 浏览器和通讯平台都是一等界面
+### 🎨 Show Pages——agent 直接甩给你一个网页
-需要窗口和丰富交互时用 Workbench;聊天最快时用 Slack、Discord、Telegram、微信或 Lark / 飞书。它们连接的是同一批 Agent 和会话。
+要仪表盘、流程图、diff、报告,还是一个小应用?agent 交给你的是一个活的网页,而不是一大段文字。点一个元素、框一块区域、在截图上标几条编号批注,说一句你想要什么,agent 就改页面,或者就在你指的地方回答。
-<img src="assets/screenshots/v4/platforms-zh.png" alt="Avibe 通讯平台连接:Slack、Discord、Telegram、微信和 Lark" />
+页面可以私有;和 `avibe.bot` 配对后,还能只分享给指定的人,或者发布成一个好记的链接。常用的固定到应用栏,点开就是一个 app。
-### 📱 装进口袋
+<img src="assets/screenshots/v4/show-page-zh.png" alt="Show Page 评审:评论锚定在页面元素上,Agent 回复在页面上" />
-<img src="assets/screenshots/v4/workbench-mobile-zh.png" alt="移动端 Avibe Workbench" width="270" align="right" />
+### 📱 人离开工位,agent 照样在手
-机器在干活,你不用守着它。把 Workbench 装成桌面或手机 app,用内置语音转文字交代任务,有活儿需要你时直接收到推送。运行 `vibe remote`,同一个本地 Workbench 通过安全的 `avibe.bot` 隧道,从任意浏览器都能打开——不用 VPN 或端口转发。
+<img src="assets/screenshots/v4/workbench-mobile-zh.png" alt="手机上的 Avibe Workbench" width="270" align="right" />
-你在飞机上、在咖啡馆、用着借来的电脑。agent 提示有个活儿需要你。点开链接,指挥两句,再走开。
+活儿在你的机器上跑,你不用守着它。想开窗口就用 Workbench;想快就在 Slack、Discord、Telegram、微信或飞书里发一条消息——连到的是同一批 Agent、同一台机器。
-- 🌍 **你自己的 `you-app.avibe.bot`**——30 秒登录,slug 跟你一辈子
-- 🔒 **每道关卡默认拒绝**——鉴权、路由、host 校验都是 fail-closed
-- 📱 **移动端友好**——为借来的小屏设计
+- 🔔 **需要你时,它会拍拍你。** 把 Workbench 装成手机或桌面 app,任务需要你的那一刻就收到推送。
+- 🎙️ **动嘴,不动手。** 和 `avibe.bot` 配对后,实时语音输入边说边出字,说完自动就地整理干净。桌面上按 ⌥Z(Windows 和 Linux 上是 Alt+Z)。
+- 🌍 **你专属的 `you-app.avibe.bot`。** `vibe remote` 配对一次,你的 Workbench 在任何地方、任何浏览器里都能打开——不用 VPN,不用端口转发。
+- 🔒 **你的门,你定名单。** 只有你的访问策略允许的人能登录——默认只有你自己。认证、路由、主机校验全部默认拒绝。
-**数据面留在你的机器上**;`avibe.bot` 只负责控制面的握手。
+在飞机上,在咖啡馆,用着借来的电脑。agent 叫你一声,你打开链接指挥两句,转身接着忙。
<br clear="all"/>
-**还有** —— 按频道路由 Agent · 可恢复会话(thread = session)· 交互式提问 · 语音转文字 · 文件附件 · 推送与完成通知。
+### 🔐 Vaults——按名字用密钥,永远不碰密钥值
+
+API Key 或 token 添加一次——缺了哪个,Agent 也能发请求让你在浏览器里补上。Agent 按名字取用,Avibe 把它递给需要它的命令、认证请求或签名操作。Vault 的响应永远不包含密钥值;接收密钥的命令仍需避免把它打印出来。passkey 保护托管在 sandbox 完整性门禁落地前仍未上线。
+
+**还有**——thread 即会话,随时接着聊 · agent 需要你拍板时弹按钮和表单 · 丰富的附件,应用内直接预览,音频就地播放 · 键盘快捷键 · 后端一键升级。
---
@@ -126,32 +142,32 @@ Windows 上推荐用 WSL,兼容性最好——见 [从零用 WSL 跑 Avibe](do
| | |
|---|---|
-| **本地优先,且属于你** | AI 伙伴、它的执行、你的密钥和数据,都留在你的机器上。`avibe.bot` 只签发身份和一条安全隧道,从不中转你的数据。 |
-| **一套底座,通吃所有第一方 agent** | 驱动*官方*的 Claude Code、Codex、OpenCode。自带订阅或 key,按任务切换,不被任何一家的 silo 绑死。 |
-| **浏览器和聊天,都是一等入口** | 用浏览器 Workbench,或 Slack、Discord、Telegram、微信、飞书。同一个 agent,同一批会话。 |
-| **没有中间人** | 你和 agent 之间没有第二层推理。token 直接花在你选的 agent 上。 |
+| **本地优先,真正归你** | AI 伙伴、它的执行、你的密钥和代码都留在你的机器上。`avibe.bot` 签发身份、提供安全隧道,从不代理你的工作区。 |
+| **所有第一方 agent,一个家** | 驱动*官方*的 Claude Code、Codex、OpenCode。按任务、按项目、按频道切换,不被任何一家锁死。 |
+| **所有订阅,组成一支队** | Model Hub 把你的套餐和 Key 汇进同一个网关,模型可以跨 agent 混搭,自动故障接管,还告诉你每一份到底值多少。 |
+| **浏览器和聊天,都是一等公民** | Workbench、手机、Slack、Discord、Telegram、微信、飞书——同一批 agent,同一台机器。 |
+| **没有中间商抽成** | 你和 agent 之间没有额外的推理循环,每个 token 都直接给到你选的 agent。 |
---
## 它怎么工作
```
-┌──────────────┐ ┌──────────────┐ ┌──────────────┐
-│ 你 │ 浏览器 │ │ stdio │ Claude Code │
-│ (任意地点) │ Slack │ Avibe │ ─────────▶ │ OpenCode │
-│ │ Discord │ (你的机器) │ ◀───────── │ Codex │
-│ │ Telegram │ │ │ │
-│ │ 微信 │ │ │ │
-│ │ 飞书 │ │ │ │
-└──────────────┘ └──────────────┘ └──────────────┘
+ 你,在任何地方 你的机器 你的模型
+┌──────────────┐ ┌──────────────────────────────────────┐ ┌──────────────┐
+│ Browser │ │ Avibe │ │ Subscriptions│
+│ Phone (PWA) │ ───▶ │ Claude Code ─┐ │ │ API keys │
+│ Slack │ │ Codex ─┼─▶ Model Hub ────────┼────▶ │ Relays │
+│ Discord … │ ◀─── │ OpenCode ─┘ │ │ Self-hosted │
+└──────────────┘ └──────────────────────────────────────┘ └──────────────┘
```
-1. **你说一句**——在浏览器或聊天软件里:*“给设置页加个深色模式。”*
-2. **Avibe 路由**到你配置的 agent,落在正确的项目里。
-3. **agent** 读你本地的代码、写代码、把结果流式发回来。
-4. **你审阅**,在同一个界面里迭代,之后随时随地接着干。
+1. **你开口**——在浏览器或聊天软件里:*"给设置页加个深色模式。"*
+2. **Avibe 路由**到对的 Agent、对的项目。
+3. **agent** 读你本地的代码、写代码、把过程流式回传。网关模式下,每次调用走哪个来源由 Model Hub 决定;直连模式下,agent 用它自己的登录。
+4. **你在同一个界面里 review**,在 thread 里迭代,之后在任何地方接着干。
-**你的代码留在你的机器上。** Avibe 在本地运行,通过 Slack Socket Mode、Discord Gateway、Telegram 长轮询、微信轮询或飞书 WebSocket 向外连接——正常聊天控制不需要任何公网入站端口。Prompt 只会发给你选的 AI 提供商。
+Avibe 通过 Slack Socket Mode、Discord Gateway、Telegram 长轮询、微信轮询或飞书 WebSocket 主动向外连接——聊天控制不需要任何公网入站端口。agent 的 prompt 只发给你配置的模型来源。
---
@@ -159,80 +175,61 @@ Windows 上推荐用 WSL,兼容性最好——见 [从零用 WSL 跑 Avibe](do
| | Avibe | OpenClaw |
|---|---|---|
-| **设置** | 一条命令 + Web 向导,几分钟搞定。 | Gateway + channels + JSON 配置,准备搭一下午。 |
-| **安全** | 本地优先,只走 Socket Mode / WebSocket。没有公网入站端口,攻击面小。 | Gateway 暴露端口,组件更多、面更大。 |
-| **Token 开销** | 中间没有额外推理 loop,token 直接花在你选的 agent 上。 | 每条消息都背着一长串人设/编排上下文,活儿还没开始 token 先烧。 |
-| **锁定** | 驱动官方 agent CLI,自带 key,按任务切换。 | 绑在它自己的助手循环上。 |
+| **上手** | 一条命令 + 网页向导,几分钟搞定。 | Gateway + channels + JSON 配置,准备花一个下午。 |
+| **安全** | 本地优先,只用 Socket Mode / WebSocket / 长轮询主动外连,无公网入站端口,攻击面极小。 | Gateway 要暴露端口,组件更多,面更大。 |
+| **Token 成本** | 中间没有额外的推理循环,token 直接给到你选的 agent。 | 每条消息都带着长长的人设/编排上下文,任务还没开始 token 就先烧在开销上。 |
+| **锁定** | 驱动官方 agent CLI,自带订阅和 Key,按任务切换。 | 绑死在它自己的助手循环里。 |
-OpenClaw 是个 always-on 的个人助手——闲聊很好,干真活儿就贵。Avibe 是给你已经信任的那些 agent 用的**本地优先 Agent OS**:agent 还是它自己,数据留在本地,同事感来自把 agent 放进你本来就在用的工作流里。
+OpenClaw 是一个常驻的个人助手。Avibe 是给你已经信任的 agent 用的 **Agent OS**:agent 还是它自己,数据留在本地,而"同事感"来自把 agent 放进你本来就在工作的地方。
---
## 常见问题
+<details>
+<summary><b>能用我现有的 Claude 或 ChatGPT 订阅吗?</b></summary>
+
+能。Claude Code 和 Codex 可以继续用各自的登录;也可以把支持的订阅作为网关来源加进 Model Hub,获得额度追踪、用量分析和自动接管。各厂商的条款依然适用;如果某种登录方式可能带来账号限制,Avibe 会在登录前提醒你。详见 [Model Hub](https://docs.avibe.bot/zh/concepts/model-hub)。
+</details>
+
<details>
<summary><b>能跑本地模型吗?</b></summary>
-这里的「本地优先」指你的**代码、数据、执行**留在你自己的机器上——不是模型权重。Avibe 驱动你配置的 agent(Claude Code、Codex、OpenCode);如果你也想本地推理,OpenCode 可以指向本地或 OpenAI 兼容端点。
+这里的"本地优先"指你的**代码、数据和执行**留在你的机器上——不一定是模型权重。如果你也想让推理在本地,Model Hub 可以指向任何兼容的自建服务。
</details>
<details>
-<summary><b>我的代码和数据去哪了?</b></summary>
+<summary><b>我的代码和数据会去哪?</b></summary>
-你的代码、密钥、agent 进程都留在你自己的机器上。`avibe.bot` 只签发身份和一条安全隧道——从不代理或存储你的数据。Prompt 只发给你选择的 AI 提供商。
+代码、密钥和 agent 进程都留在你的机器上。agent 的 prompt 发给你配置的模型来源。`avibe.bot` 负责登录和远程 Web UI 隧道,但不代理你的工作区;例外是语音输入,它会经 `avibe.bot` 做转写。每一条承载你内容的网络路径都列在[边界清单](https://docs.avibe.bot/zh/concepts/local-first#哪些东西会离开你的机器)里。
</details>
<details>
-<summary><b>用 Avibe 要付费吗?</b></summary>
+<summary><b>Avibe 要付费吗?</b></summary>
-Avibe 开源(MIT)、免费运行。你自带 agent 订阅或 API key,直接付给你的提供商——没有加价,也没有第二份订阅。
+Avibe 开源(MIT),免费运行。你自带 agent 订阅或 API Key,直接付给服务商——不加价,也不用再多订一份。
</details>
<details>
<summary><b>支持哪些 agent 和平台?</b></summary>
-**Agent:** 官方 Claude Code、Codex、OpenCode CLI。**界面:** 内置浏览器 Workbench,外加 Slack、Discord、Telegram、微信、飞书 / Lark。
+**Agent:** 官方 Claude Code、Codex、OpenCode CLI。**界面:** 内置浏览器 Workbench(可安装到桌面和手机),以及 Slack、Discord、Telegram、微信、飞书。
</details>
<details>
<summary><b>远程访问安全吗?</b></summary>
-`vibe remote` 会开一条 Cloudflare 隧道;浏览器流量在登录之后才能到你的机器。鉴权、路由、host 校验全部 **fail-closed**,常规聊天控制没有任何公网入站端口。
+`vibe remote` 从你的机器上跑一条 Cloudflare 隧道;浏览器流量必须先登录(你主动公开发布的 Show Page 除外),而且只放行你的访问策略允许的人(默认只有你)。认证、路由、主机校验全部 **fail-closed**,聊天控制也不需要任何公网入站端口。
</details>
<details>
<summary><b>Avibe 和 OpenClaw、Hermes 有什么不同?</b></summary>
-OpenClaw 和 Hermes 是 *agent*——一个偏网关式助手,一个会自我进化。Avibe 是完全不同的一层:**Agent OS**。它给任何 agent 一套统一的世界模型——agents、sessions、Show Pages、Harness——让它能自己调度、自己搭 loop、通过真正的人机交互层来找你,再去驱动你自带的官方 Claude Code、Codex、OpenCode(Avibe 原生 agent 已在[路线图](#路线图)上)。OpenClaw 的细节见上方的[对比表](#avibe-vs-openclaw)。
+OpenClaw 和 Hermes 是 *agent*——一个是网关式助手,一个是会自我进化的 agent。Avibe 在另一层:**Agent OS**。它给自己运行的每个 agent 一个统一的世界模型——Agent、会话、Show Page、Harness、Model Hub——让 agent 能自己排期、搭自己的循环、通过真正的交互层找到你;然后它运行的是你自带的官方 Claude Code、Codex、OpenCode(Avibe 原生 agent [在路线图上](#路线图))。OpenClaw 的具体对比见上面的[对比表](#avibe-vs-openclaw)。
</details>
---
-## 像同事一样跟它说话
-
-用人话说,Harness 在背后自己组合命令:
-
-- *“盯着这个 PR,有 actionable review 再回来处理。”*
-- *“每个工作日上午跑一遍部署检查,把摘要发到这里。”*
-- *“为这个 incident 新开一个调查会话,但把结论发回这个频道。”*
-- *“CI 挂了就总结日志;过了就告诉我这个 PR 能不能合。”*
-
-**对话中途换 agent**——加个前缀就行:
-
-```
-Plan: 给 API 设计一个新的缓存层
-```
-
-**按项目路由**——不同的活儿,不同的 agent:
-
-```
-frontend → OpenCode (快速迭代)
-backend → Claude Code (复杂逻辑)
-prototypes → Codex (快速试验)
-```
-
----
-
## 认识云团子(Vibey)
<div align="center">
@@ -253,12 +250,14 @@ prototypes → Codex (快速试验)
vibe # 启动 Avibe 并打开 Workbench
vibe status # 查看服务和配置状态
vibe stop # 停止本地服务
-vibe doctor # 诊断常见安装问题;用 "vibe doctor repair" 显式执行安全修复
+vibe upgrade # 升级到最新版本
+vibe doctor # 诊断常见安装问题;"vibe doctor repair" 执行安全修复
vibe remote # 通过 avibe.bot 从任意设备访问 Workbench
vibe agent # 运行和管理 Avibe agent
vibe task # 安排定时工作(cron / 一次性)
vibe watch # 等一个条件成立,然后行动
vibe runs # 查看 agent 运行历史
+vibe show # 创建、查看和发布 Show Page
vibe vault # 管理密钥、请求、认证请求与签名
```
@@ -266,43 +265,43 @@ vibe vault # 管理密钥、请求、认证请求与签名
|---|---|
| @ 一下 bot | 开一个任务或提问 |
| 在 thread 里回复 | 继续同一个 agent 会话 |
-| `/stop` | 停止当前会话 |
+| `/stop` | 中断当前执行 |
完整参考:[命令](docs/COMMANDS_ZH.md) · [CLI](docs/CLI_ZH.md)
---
-## 前置条件
+## Agent
-至少装一个 agent:
+设置向导会检测你机器上的 Claude Code、Codex 和 OpenCode,缺哪个、你选了哪个,就帮你装上。想自己装的话:
<details>
-<summary><b>OpenCode</b>(推荐)</summary>
+<summary><b>Claude Code</b></summary>
```bash
-curl -fsSL https://opencode.ai/install | bash
-```
-
-在 `~/.config/opencode/opencode.json` 里加上这一段,跳过权限弹窗:
-
-```json
-{ "permission": "allow" }
+bash -o pipefail -c 'curl -fsSL https://claude.ai/install.sh | bash'
```
</details>
<details>
-<summary><b>Claude Code</b></summary>
+<summary><b>Codex</b></summary>
```bash
-npm install -g @anthropic-ai/claude-code
+npm install -g @openai/codex
```
</details>
<details>
-<summary><b>Codex</b></summary>
+<summary><b>OpenCode</b></summary>
```bash
-npm install -g @openai/codex
+bash -o pipefail -c 'curl -fsSL https://opencode.ai/install | bash'
+```
+
+除非 `~/.config/opencode/opencode.json` 放行工具调用,否则 OpenCode 可能会停下来等你审批;设置向导可以帮你写好:
+
+```json
+{ "permission": "allow" }
```
</details>
@@ -310,20 +309,31 @@ npm install -g @openai/codex
## 安全
-- **本地优先**——Avibe 跑在你的机器上,代码和 agent 进程都留在本地。
-- **没有公网入站端口**——聊天控制只走 Socket Mode / WebSocket / 长轮询。
-- **你的 key、你的数据**——存在 `~/.avibe/` 下,只发给你选的 AI 提供商。老安装保留 `~/.vibe_remote/` 作为兼容路径。
-- **远程访问默认拒绝**——`avibe.bot` 只代理身份和隧道,从不碰你的数据。
+- **本地优先**——Avibe 跑在你的机器上,代码和 agent 进程都留在那。
+- **无公网入站端口**——聊天控制只用 Socket Mode / WebSocket / 长轮询。
+- **你的密钥,你的数据**——存放在 `~/.avibe/`;agent 的 prompt 只发给你配置的模型来源。已有安装会保留 `~/.vibe_remote/` 作为兼容路径。
+- **远程访问默认拒绝**——`avibe.bot` 负责身份和隧道,从不经手你的工作区。
---
## 卸载
+这段命令只删除 `PATH` 上排在最前面的 `vibe`。如果 `which -a vibe` 列出了不止一个 Avibe 启动器,其余的也要一并删掉。
+
```bash
-vibe stop
+avibe_home="${AVIBE_HOME:-$HOME/.avibe}"
+avibe_home="${avibe_home/#\~/$HOME}"
+vibe_bin="$(command -v vibe)" # 卸载前先记下 Avibe 的启动器
+if "$vibe_bin" version 2>/dev/null | grep -Eq '^(avibe-os|vibe-remote) '; then
+ "$vibe_bin" stop
+else
+ echo "跳过 ${vibe_bin}:不是 Avibe 的启动器"; vibe_bin=""
+fi
uv tool uninstall avibe-os
uv tool uninstall vibe-remote # 旧版安装
-rm -rf ~/.avibe ~/.vibe_remote
+[ -n "$vibe_bin" ] && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
+rm -rf "$avibe_home/runtime/install-generations"
+rm -rf "$avibe_home" ~/.vibe_remote
```
---
@@ -332,20 +342,22 @@ rm -rf ~/.avibe ~/.vibe_remote
接下来要做的:
-- **更深入的交互式工作流** —— 在 Chat、应用与 Show Page 之间加入更多直接操作、结构化决策和原位协作。
-- **保护档 Vault 托管** —— 完成不可变 sandbox 完整性门禁后,再向真实用户开放 passkey 保护的密钥。
-- **SaaS 模式** —— 一键托管上手 + 云端中继,而执行依然留在你自己的机器上。
-- **Avibe 原生 agent** —— 为这个运行时调校的第一方 agent,与你自带的官方 CLI 并存。
+- **原生桌面 app**——带签名自动更新的 macOS 和 Windows 版本正在测试中。
+- **更深入的交互式工作流**——在 Chat、应用与 Show Page 之间加入更多直接操作、结构化决策和原位协作。
+- **保护档 Vault 托管**——完成不可变 sandbox 完整性门禁后,再向真实用户开放 passkey 保护的密钥。
+- **SaaS 模式**——一键托管上手 + 云端中继,而执行依然留在你自己的机器上。
+- **Avibe 原生 agent**——为这个运行时调校的第一方 agent,与你自带的官方 CLI 并存。
-近期已上线:多窗口应用、Show Page 标注、Agent Harness、标准档 Vaults、语音转文字与 Skills 管理。
+近期已上线:支持自动接管、额度追踪和用量分析的 Model Hub · 实时语音输入 · Show Page 分享 · 三个后端统一的 Skills · 三步设置向导 · Agent Harness 与运行关系图 · 多窗口应用 · 标准档 Vaults。
---
## 文档
-- **[官方文档](https://docs.avibe.bot)**——快速上手、概念、平台与 agent 指南、排障
+- **[官方文档](https://docs.avibe.bot/zh)**——快速上手、概念、平台与 agent 指南、排障
- **[Avibe 是什么](https://docs.avibe.bot/zh/concepts/agent-os)**——Agent OS 模型
-- **[CLI 参考](docs/CLI_ZH.md)** · **[命令](docs/COMMANDS_ZH.md)**
+- **[Model Hub](https://docs.avibe.bot/zh/concepts/model-hub)**——来源、路由、自动接管、用量与额度
+- **[CLI 参考](docs/CLI_ZH.md)** · **[命令](docs/COMMANDS_ZH.md)** · **[Show Pages](docs/SHOW_PAGES.md)**
- **[让 AI agent 帮你装](docs/INSTALL_FOR_AI_ZH.md)**——把它丢给 Claude Code、Codex 或 OpenCode,引导式安装
- **[Slack](docs/SLACK_SETUP_ZH.md)** · **[Discord](docs/DISCORD_SETUP_ZH.md)** · **[Telegram](docs/TELEGRAM_SETUP_ZH.md)** 设置指南
@@ -353,8 +365,8 @@ rm -rf ~/.avibe ~/.vibe_remote
<div align="center">
-**拥有你的 agent。随处都能找到它。**
+**Agent 归你。走到哪,带到哪。**
-[立即安装](#avibe-把这件事反过来) · [文档](https://docs.avibe.bot) · [报 bug](https://github.com/avibe-bot/avibe/issues) · [关注 @alex_metacraft](https://x.com/alex_metacraft)
+[立即安装](#avibe-给它松绑) · [文档](https://docs.avibe.bot/zh) · [报 bug](https://github.com/avibe-bot/avibe/issues) · [关注 @alex_metacraft](https://x.com/alex_metacraft)
</div>
diff --git a/assets/screenshots/v4/model-hub-routes-en.png b/assets/screenshots/v4/model-hub-routes-en.png
new file mode 100644
index 000000000..a26029f75
Binary files /dev/null and b/assets/screenshots/v4/model-hub-routes-en.png differ
diff --git a/assets/screenshots/v4/model-hub-routes-zh.png b/assets/screenshots/v4/model-hub-routes-zh.png
new file mode 100644
index 000000000..de2a9193c
Binary files /dev/null and b/assets/screenshots/v4/model-hub-routes-zh.png differ
diff --git a/assets/screenshots/v4/model-hub-usage-en.png b/assets/screenshots/v4/model-hub-usage-en.png
new file mode 100644
index 000000000..be2dea878
Binary files /dev/null and b/assets/screenshots/v4/model-hub-usage-en.png differ
diff --git a/assets/screenshots/v4/model-hub-usage-zh.png b/assets/screenshots/v4/model-hub-usage-zh.png
new file mode 100644
index 000000000..515f0a80a
Binary files /dev/null and b/assets/screenshots/v4/model-hub-usage-zh.png differ
diff --git a/docs/INSTALL_FOR_AI.md b/docs/INSTALL_FOR_AI.md
index 74d714221..a16138b8b 100644
--- a/docs/INSTALL_FOR_AI.md
+++ b/docs/INSTALL_FOR_AI.md
@@ -191,14 +191,20 @@ does not need an IM chat to complete this smoke test.
## Uninstall
-Only run this if the user asks to remove Avibe:
+Only run this if the user asks to remove Avibe. It removes the first `vibe` on `PATH`; if `which -a vibe` lists more than one Avibe launcher, remove the others too:
```bash
-vibe stop
avibe_home="${AVIBE_HOME:-$HOME/.avibe}"
avibe_home="${avibe_home/#\~/$HOME}"
-uv tool uninstall vibe-remote
-vibe_bin="$(command -v vibe)" && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
+vibe_bin="$(command -v vibe)" # capture Avibe's launcher before uninstalling
+if "$vibe_bin" version 2>/dev/null | grep -Eq '^(avibe-os|vibe-remote) '; then
+ "$vibe_bin" stop
+else
+ echo "Skipping ${vibe_bin}: not Avibe's launcher"; vibe_bin=""
+fi
+uv tool uninstall avibe-os
+uv tool uninstall vibe-remote # legacy install
+[ -n "$vibe_bin" ] && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
rm -rf "$avibe_home/runtime/install-generations"
rm -rf "$avibe_home" ~/.vibe_remote
```
diff --git a/docs/INSTALL_FOR_AI_ZH.md b/docs/INSTALL_FOR_AI_ZH.md
index becbc391f..7325dbc95 100644
--- a/docs/INSTALL_FOR_AI_ZH.md
+++ b/docs/INSTALL_FOR_AI_ZH.md
@@ -166,14 +166,20 @@ vibe doctor
## 卸载
-只有用户明确要删除 Avibe 时才运行:
+只有用户明确要删除 Avibe 时才运行。它只删除 `PATH` 上排在最前面的 `vibe`;如果 `which -a vibe` 列出了不止一个 Avibe 启动器,其余的也要一并删掉:
```bash
-vibe stop
avibe_home="${AVIBE_HOME:-$HOME/.avibe}"
avibe_home="${avibe_home/#\~/$HOME}"
-uv tool uninstall vibe-remote
-vibe_bin="$(command -v vibe)" && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
+vibe_bin="$(command -v vibe)" # 卸载前先记下 Avibe 的启动器
+if "$vibe_bin" version 2>/dev/null | grep -Eq '^(avibe-os|vibe-remote) '; then
+ "$vibe_bin" stop
+else
+ echo "跳过 ${vibe_bin}:不是 Avibe 的启动器"; vibe_bin=""
+fi
+uv tool uninstall avibe-os
+uv tool uninstall vibe-remote # 旧版安装
+[ -n "$vibe_bin" ] && rm -f "$vibe_bin" "$(dirname "$vibe_bin")/.vibe.avibe-generation"
rm -rf "$avibe_home/runtime/install-generations"
rm -rf "$avibe_home" ~/.vibe_remote
```