claude-bot/EJClaw
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

chore: update README for MAGI/MoA/Bun + show model names in MoA dashboard

Browse Source
This commit is contained in:
Eyejoker
2026-03-31 01:51:44 +09:00
parent da27d30079
commit e9f9340ece
3 changed files with 78 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

141
README.md
View File

@@ -2,13 +2,14 @@
![Claude Agent SDK](https://img.shields.io/badge/Claude_Agent_SDK-0.2.81-blueviolet) ![Claude Agent SDK](https://img.shields.io/badge/Claude_Agent_SDK-0.2.81-blueviolet)
![Codex SDK](https://img.shields.io/badge/Codex_SDK-0.115.0-green) ![Codex SDK](https://img.shields.io/badge/Codex_SDK-0.115.0-green)
![Node](https://img.shields.io/badge/Node-20+-339933?logo=nodedotjs&logoColor=white) ![Bun](https://img.shields.io/badge/Bun-1.3+-f9f1e1?logo=bun&logoColor=black)
![Discord](https://img.shields.io/badge/Discord-Bot-5865F2?logo=discord&logoColor=white) ![Discord](https://img.shields.io/badge/Discord-Bot-5865F2?logo=discord&logoColor=white)
Dual-agent AI assistant (Claude Code + Codex) over Discord with autonomous paired review. MAGI-style multi-agent AI assistant (Claude Code + Codex + Arbiter) over Discord with autonomous paired review and Mixture of Agents.
Originally derived from [qwibitai/nanoclaw](https://github.com/qwibitai/nanoclaw), now maintained as EJClaw for personal production use. Originally derived from [qwibitai/nanoclaw](https://github.com/qwibitai/nanoclaw), now maintained as EJClaw for personal production use.
Prompt design inspired by [Q00/ouroboros](https://github.com/Q00/ouroboros), adapted for EJClaw's Discord and dual-service workflow. Prompt design inspired by [Q00/ouroboros](https://github.com/Q00/ouroboros), adapted for EJClaw's Discord and dual-service workflow.
Arbiter system inspired by [Evangelion MAGI](https://evangelion.fandom.com/wiki/MAGI) consensus architecture.
## Overview ## Overview
@@ -16,44 +17,55 @@ A single unified service (`ejclaw`) manages three Discord bots in one process:
- **Codex-main** (`@codex`) — Owner agent. Handles user requests, writes code. - **Codex-main** (`@codex`) — Owner agent. Handles user requests, writes code.
- **Claude** (`@claude`) — Reviewer agent. Critically reviews owner's work, verifies design direction. - **Claude** (`@claude`) — Reviewer agent. Critically reviews owner's work, verifies design direction.
- **Codex-review** (`@codex-review`) — Fallback reviewer. Takes over when Claude hits rate limits (429). - **Codex-review** (`@codex-review`) — Arbiter agent. Summoned on-demand to break deadlocks between owner and reviewer.
Owner and reviewer agent types are configurable via `OWNER_AGENT_TYPE` and `REVIEWER_AGENT_TYPE` in `.env`. All agent types and models are independently configurable per role via `.env`.
## Paired Review ## MAGI 3-Agent System
The core workflow is an autonomous owner-reviewer ping-pong:
``` ```
User message User message
→ Owner responds (implementation, answer, etc.) → Owner responds (implementation, answer, etc.)
→ Reviewer auto-triggered (critical review, design check) → Reviewer auto-triggered (critical review, design check)
→ Verdict: → Verdict:
DONE → Owner gets final turn to finalize → Task completed DONE → Owner finalizes → Task completed
DONE_WITH_CONCERNS → Owner addresses feedback → Reviewer re-reviews DONE_WITH_CONCERNS → Owner addresses feedback → loop
BLOCKED → Escalate to user (needs decision) BLOCKED/NEEDS_CONTEXT
NEEDS_CONTEXT → Escalate to user (missing information) ├─ Arbiter enabled → Arbiter judges → PROCEED/REVISE/RESET/ESCALATE
└─ Arbiter disabled → Escalate to user
→ Deadlock (3+ round trips without progress)
→ Arbiter summoned → binding verdict → loop resumes
``` ```
The system stops autonomously when the reviewer approves or escalates. No manual intervention needed for the happy path. ### Mixture of Agents (MoA)
When enabled, the arbiter collects opinions from external models (Kimi, GLM, etc.) before rendering its verdict:
```
Deadlock detected → MoA reference queries (Kimi + GLM, parallel)
→ Opinions injected into arbiter's prompt
→ SDK arbiter (subscription-based) aggregates all perspectives
→ Final verdict: PROCEED / REVISE / RESET / ESCALATE
```
No extra SDK processes. External references use lightweight API calls (Anthropic-compatible).
## Features ## Features
- **Paired review** — Autonomous owner/reviewer ping-pong with verdict-based control - **MAGI 3-agent system** — Owner/reviewer/arbiter with on-demand deadlock resolution
- **Container-isolated reviewer** — Reviewer runs in persistent Docker container with read-only source mount (kernel-level write protection) - **Mixture of Agents** — External model opinions (Kimi, GLM) enrich arbiter verdicts
- **Post-approval change detection** — Re-triggers review if owner modifies code after reviewer DONE (git tree hash comparison) - **Per-role model selection** — `OWNER_MODEL`, `REVIEWER_MODEL`, `ARBITER_MODEL` + effort + fallback toggle
- **Configurable agent types** — Owner and reviewer roles independently set to `claude-code` or `codex` - **Container-isolated reviewer** — Persistent Docker container with read-only source mount
- **Reviewer fallback** — Claude 429/exhaustion → automatic handoff to codex-review - **Global failover** — Account-level Claude failure → all channels switch to codex, auto-recovers
- **Credential proxy** — API keys never exposed to containers, injected via HTTP proxy - **Post-approval change detection** — Re-triggers review if owner modifies code after approval
- **Voice transcription** — Groq Whisper (primary) / OpenAI Whisper (fallback), shared file cache with dedup - **Voice transcription** — Groq Whisper (primary) / OpenAI Whisper (fallback)
- **Bidirectional images** — Receive Discord attachments as multimodal input, send screenshots back - **Token rotation** — Multi-account Claude/Codex rotation on rate limits
- **Token rotation** — Claude 429 / usage exhaustion → automatic multi-account rotation - **Kimi usage dashboard** — Coding plan 5h/7d usage displayed alongside Claude/Codex
- **MCP integration** — Memento (persistent memory) + EJClaw host tools (send_message, schedule_task, watch_ci, etc.) - **MCP integration** — Memento (persistent memory) + EJClaw host tools
- **Browser automation** — [gstack browse](https://github.com/garrytan/gstack) skill, headless Chromium daemon - **Session persistence** — Separate sessions per role (owner/reviewer/arbiter)
- **Priority queue** — Per-group serialization, global concurrency limit
- **Session persistence** — Resume conversations across restarts (separate sessions per role)
- **Scheduled tasks** — Cron/interval/once via MCP tool - **Scheduled tasks** — Cron/interval/once via MCP tool
- **Mid-turn steering** — Inject follow-up messages while agent is working - **Mid-turn steering** — Inject follow-up messages while agent is working
- **Bun runtime** — Native SQLite (bun:sqlite), fast startup, no native addon builds
## Architecture ## Architecture
@@ -61,27 +73,26 @@ The system stops autonomously when the reviewer approves or escalates. No manual
Discord ──► SQLite (WAL) ──► GroupQueue ──┬──► Owner (host process) Discord ──► SQLite (WAL) ──► GroupQueue ──┬──► Owner (host process)
│ │ │ │
│ ▼ (auto-trigger) │ ▼ (auto-trigger)
└──► Reviewer (Docker container, :ro mount) ├──► Reviewer (Docker container, :ro mount)
│ │ │ │
│ Verdict routing │ Verdict routing
│ ├─ DONE → change detection → Owner finalizes or re-review │ ├─ DONE → change detection → finalize or re-review
│ ├─ BLOCKED/NEEDS_CONTEXT → User │ ├─ BLOCKED → Arbiter (if enabled) or User
│ └─ Feedback → Owner (loop) │ └─ Feedback → Owner (loop)
│ │
IPC polling ◄── follow-up messages (mid-turn steering) ├──► Arbiter (on-demand, fresh session each time)
│ │
│ ┌───┴─── MoA (if enabled) ───┐
│ │ Kimi API ──► opinion │
│ │ GLM API ──► opinion │
│ │ → injected into prompt │
│ └────────────────────────────┘
│ │
│ PROCEED/REVISE/RESET/ESCALATE
│
IPC polling ◄── follow-up messages
│ │
Router ──► Discord (text, images, files) Router ──► Discord (text, images, files)
Owner (host process):
├── Stable per-channel worktree (session persists across tasks)
├── MCP tools, Bash skills, per-group memory
└── Full read-write access
Reviewer (persistent Docker container):
├── Owner workspace mounted read-only (kernel-level protection)
├── Credential proxy (API keys never in container)
├── pnpm store mounted read-only (if applicable)
└── tmpfs for test caches (vitest, jest, npm)
``` ```
## Setup ## Setup
@@ -89,11 +100,11 @@ Reviewer (persistent Docker container):
### Prerequisites ### Prerequisites
- Linux (Ubuntu 22.04+) or macOS - Linux (Ubuntu 22.04+) or macOS
- Node.js 20+ (24 recommended, fnm for version management) - [Bun](https://bun.sh/) 1.3+
- Docker (required for reviewer container isolation) - Docker (required for reviewer container isolation)
- [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) - [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code)
- [Codex CLI](https://github.com/openai/codex) (`npm install -g @openai/codex`) - [Codex CLI](https://github.com/openai/codex) (`npm install -g @openai/codex`)
- Discord bot tokens (3: Claude, Codex-main, Codex-review) - Discord bot tokens (3: Claude, Codex-main, Codex-review/Arbiter)
### Install ### Install
@@ -113,44 +124,38 @@ All configuration in a single `.env` file:
# Discord bots (3 tokens for 3 bots) # Discord bots (3 tokens for 3 bots)
DISCORD_BOT_TOKEN= # Claude bot DISCORD_BOT_TOKEN= # Claude bot
DISCORD_CODEX_BOT_TOKEN= # Codex-main bot (owner) DISCORD_CODEX_BOT_TOKEN= # Codex-main bot (owner)
DISCORD_REVIEW_BOT_TOKEN= # Codex-review bot (fallback reviewer) DISCORD_REVIEW_BOT_TOKEN= # Codex-review bot (arbiter)
# Agent type configuration # Agent types
OWNER_AGENT_TYPE=codex # codex | claude-code OWNER_AGENT_TYPE=codex # codex | claude-code
REVIEWER_AGENT_TYPE=claude-code # claude-code | codex REVIEWER_AGENT_TYPE=claude-code # claude-code | codex
ARBITER_AGENT_TYPE=codex # codex | claude-code (optional, enables 3rd agent)
# Per-role model overrides
OWNER_MODEL=gpt-5.4
REVIEWER_MODEL=claude-opus-4-6
ARBITER_MODEL=gpt-5.4
# API keys # API keys
CLAUDE_CODE_OAUTH_TOKEN= # Claude Code OAuth token CLAUDE_CODE_OAUTH_TOKEN= # Claude Code OAuth token
CLAUDE_CODE_OAUTH_TOKENS= # Comma-separated tokens for multi-account rotation CLAUDE_CODE_OAUTH_TOKENS= # Comma-separated for multi-account rotation
OPENAI_API_KEY= # For Codex
GROQ_API_KEY= # Voice transcription (Groq Whisper) GROQ_API_KEY= # Voice transcription (Groq Whisper)
# Bot names # Mixture of Agents (MoA)
ASSISTANT_NAME=claude # Claude bot trigger name MOA_ENABLED=true
CODEX_ASSISTANT_NAME=codex # Codex bot trigger name MOA_REF_MODELS=kimi,glm
``` MOA_KIMI_MODEL=kimi-k2.5
MOA_KIMI_BASE_URL=https://api.kimi.com/coding
### Authentication MOA_KIMI_API_KEY=sk-kimi-xxx
MOA_KIMI_API_FORMAT=anthropic
Multi-account OAuth token rotation is supported via `CLAUDE_CODE_OAUTH_TOKENS` (comma-separated). When one account hits a rate limit, the system automatically rotates to the next. MOA_GLM_MODEL=glm-5.1
MOA_GLM_BASE_URL=https://open.bigmodel.cn/api/anthropic
Token auto-refresh runs on the Claude service only, refreshing access tokens 30 minutes before expiry using rotating refresh tokens from `~/.claude/.credentials.json` (account 0) and `~/.claude-accounts/{n}/.credentials.json` (account 1+). Generate tokens with `claude setup-token` (account 0) or `CLAUDE_CONFIG_DIR=~/.claude-accounts/{n} claude setup-token` (account 1+). MOA_GLM_API_KEY=xxx
MOA_GLM_API_FORMAT=anthropic
### Systemd Service (Linux)
Single unified service:
```bash
systemctl --user restart ejclaw # Restart
systemctl --user status ejclaw # Check status
systemctl --user enable ejclaw # Enable on boot
journalctl --user -u ejclaw -f # Follow logs
``` ```
### Deploy ### Deploy
Build on server, not locally:
```bash ```bash
ssh clone-ej@100.64.185.108 'cd ~/EJClaw && git pull && bun run build && bun run build:runners && systemctl --user restart ejclaw' ssh clone-ej@100.64.185.108 'cd ~/EJClaw && git pull && bun run build && bun run build:runners && systemctl --user restart ejclaw'
``` ```

5
src/moa.ts
View File

@@ -99,9 +99,8 @@ async function queryModel(
.map((b: { text: string }) => b.text) .map((b: { text: string }) => b.text)
.join(''); .join('');
} else { } else {
content = ( content = (data as { choices?: { message?: { content?: string } }[] })
data as { choices?: { message?: { content?: string } }[] } .choices?.[0]?.message?.content;
).choices?.[0]?.message?.content;
} }
if (!content) throw new Error('Empty response from model'); if (!content) throw new Error('Empty response from model');

4
src/unified-dashboard.ts
View File

@@ -595,7 +595,9 @@ function buildModelConfigSection(): string {
// MoA status // MoA status
const moaConfig = getMoaConfig(); const moaConfig = getMoaConfig();
if (moaConfig.enabled) { if (moaConfig.enabled) {
const refs = moaConfig.referenceModels.map((m) => m.name).join(', '); const refs = moaConfig.referenceModels
.map((m) => `${m.name} \`${m.model}\``)
.join(', ');
lines.push(` **MoA** — ${refs}`); lines.push(` **MoA** — ${refs}`);
} }