feat: Cost Control Center

.github/scripts/ci-watch.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+# ci-watch.sh: Re-check GitHub Actions CI build until it completes and reports success.
+# Usage: ./scripts/ci-watch.sh [branch]
+
+BRANCH=${1:-$(git rev-parse --abbrev-ref HEAD)}
+
+echo "👀 Watching recent CI run for branch: $BRANCH..."
+
+while true; do
+  # Fetch latest run ID and status using GitHub CLI for lint-and-test workflow
+  OUTPUT=$(gh run list --branch "$BRANCH" --limit 1 --json status,conclusion --jq '.[0]')
+
+  if [ -z "$OUTPUT" ] || [ "$OUTPUT" == "null" ]; then
+    echo "⚠️  No workflows found for branch $BRANCH."
+    exit 1
+  fi
+
+  STATUS=$(echo "$OUTPUT" | jq -r '.status')
+  CONCLUSION=$(echo "$OUTPUT" | jq -r '.conclusion')
+
+  if [ "$STATUS" == "completed" ] || [ "$STATUS" == "" ]; then
+    if [ "$CONCLUSION" == "success" ]; then
+      echo "✅ CI Build Succeeded for branch $BRANCH."
+      exit 0
+    else
+      echo "❌ CI Build Failed (Conclusion: $CONCLUSION) for branch $BRANCH."
+      echo "Please fix the failing tests or lint errors and push again."
+      exit 1
+    fi
+  else
+    echo "⏳ CI Build Status: $STATUS... retrying in 10s"
+    sleep 10
+  fi
+done

.github/workflows/release.yml

@@ -40,4 +40,4 @@ jobs:
       - name: Run Release Script
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: node scripts/release.mjs ${{ inputs.type }}
+        run: node .github/scripts/release.mjs ${{ inputs.type }}

.gitignore

@@ -14,8 +14,10 @@ cloudflared
 
 # Data & Logs
 data/logs/
+data/backups/
 data/token_stats/
 data/last_id.txt
+data/optimizer.skip.json
 dashboard.log
 *.log
 
@@ -31,3 +33,8 @@ test_*.js
 # Test coverage reports
 coverage/
 .nyc_output/
+
+# Local cost-control acceptance artifacts
+tests/cost-control/local/.backup/
+tests/cost-control/local/.reports/
+tests/cost-control/local/.sandbox/

CHANGELOG.md

@@ -7,6 +7,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+- Docker image support: added `Dockerfile`, `.dockerignore`, and CI workflow for automatic image publishing to GitHub Container Registry (#24)
+- Docker usage instructions in README
+
 ## [1.1.4] - 2026-03-13
 
 ### Fixed
@@ -25,9 +29,6 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 - propagate --token and other CLI args from install.sh to setup.sh (PR #26)
-
-
-
 ## [1.1.2] - 2026-03-01
 
 ### Fixed

README.md

@@ -26,6 +26,7 @@
 *   **📱 Mobile-First Dashboard**: Optimized for phone screens. View logs, stats, and controls without squinting.
 *   **🧠 Live Activity Feed**: Watch your agent "think" and execute tools in real-time. Features intelligent parallel logging (no missed background tasks) and deduplication.
 *   **💰 Token Economy**: Track daily/monthly LLM costs and usage trends. Know exactly where your money goes.
+*   **🔍 [Cost Control Center](docs/cost-control-center.md)**: 10 automated diagnostics that scan your usage, calculate savings, and let you apply optimizations with one tap. Cut API costs by 30–90%.
 *   **📜 Memory Timeline**: Browse your agent's daily journals and long-term memory evolution.
 *   **🚀 Mission Control**: Trigger cron jobs, restart services, or kill runaway scripts directly from the UI.
 *   **⚡ Zero-Config Remote**: 

README_CN.md

@@ -26,6 +26,7 @@
 *   **📱 移动端优先**: 专为手机屏幕优化的 UI，随时随地查看日志和状态，无需缩放。
 *   **🧠 实时动态 (Live Activity)**: 像看朋友圈一样查看 Agent 的“思考”和“工具调用”。支持并行日志捕获（不会漏掉后台脚本）和智能去重。
 *   **💰 Token 经济学**: 精确追踪每日/每月的 LLM 成本和 Token 用量趋势。拒绝账单刺客。
+*   **🔍 [成本控制中心](docs/cost-control-center_CN.md)**: 10 项自动化诊断，扫描使用数据、计算节省空间，一键应用优化。可降低 API 成本 30–90%。
 *   **📜 记忆时间轴**: 浏览 Agent 的每日日志和长期记忆演变。
 *   **🚀 任务控制台**: 查看 Cron 定时任务状态，并支持从手机端手动触发脚本、重启服务或终止进程。
 *   **⚡ 零门槛远程**: 

data/diagnostics.config.json.example

@@ -0,0 +1,11 @@
+{
+  "D01_modelCostRatio": 0.5,
+  "D06_cacheHitRateMin": 0.10,
+  "D06_cacheableRatio": 0.80,
+  "D09_outputRatioThreshold": 0.10,
+  "D09_minOutputTokens": 1000,
+  "D09_reductionFactor": 0.30,
+  "D05_thinkingProportion": 0.40,
+  "D05_reductionRatio": 0.75,
+  "D04_idleDaysThreshold": 7
+}

docs/cost-control-center.md

@@ -0,0 +1,69 @@
+# Cost Control Center
+
+The **Cost Control Center** is a built-in diagnostic engine that analyzes your OpenClaw agent's token usage and provides actionable recommendations to reduce API costs.
+
+## Features
+
+### Token Monitoring Dashboard
+- **Today's Cost** — live running total
+- **Grand Total** — cumulative cost since installation
+- **30-Day Forecast** — projected monthly spend based on recent patterns
+- **Top Models** — cost breakdown by model
+- **7-Day Histogram** — daily cost chart with drill-down
+
+### Diagnostic Scanner (10 Rules)
+
+| ID | Diagnostic | Description |
+|----|-----------|-------------|
+| A01 | Model Downgrade | Detects premium models used for routine tasks; suggests cheaper alternatives |
+| A02 | Heartbeat Optimization | Calculates hidden heartbeat polling costs; offers tiered intervals |
+| A03 | Session Reset Waste | Flags excessive session resets that waste context tokens |
+| A04 | Idle Skill Detection | Finds unused skills that bloat system prompts |
+| A05 | Thinking Overhead | Measures extended reasoning costs; recommends minimal mode |
+| A06 | Prompt Caching | Detects missing cache config; enables 90% input savings |
+| A07 | Compaction Safeguard | Prevents runaway costs from unbounded context growth |
+| A08 | Local Ollama Routing | Routes simple requests to local models *(planned)* |
+| A09 | Output Verbosity | Detects verbose output; adds concise directive |
+| A10 | Multi-Model Routing | Cost-aware routing by task complexity *(planned)* |
+
+### One-Tap Optimizer
+- **Apply** button on every recommendation
+- Automatic configuration backup before changes
+- Instant **Undo/Rollback** to restore previous settings
+- Full optimization history in `data/logs/optimizations.jsonl`
+
+## API Endpoints
+
+```
+GET  /api/diagnostics   — Run all diagnostic checks, returns actions with savings
+POST /api/optimize/:id  — Apply a specific optimization (A01–A10)
+```
+
+## How It Works
+
+1. The **DiagnosticsEngine** reads your JSONL usage logs and current configuration.
+2. Each diagnostic rule calculates potential savings based on your *actual* data.
+3. Results are displayed in the dashboard with dollar amounts and configuration diffs.
+4. Tapping **Apply** triggers the **OptimizerService**, which:
+   - Backs up the current config to `data/backups/`
+   - Writes the optimized setting
+   - Logs the action to `data/logs/optimizations.jsonl`
+
+## Files
+
+```
+src/services/diagnostics.js  — 10 diagnostic rules (A01–A10)
+src/services/optimizer.js    — Apply/backup/rollback engine
+src/services/pricing.js      — 340+ model pricing database
+src/routes/diagnostics.js    — GET /api/diagnostics
+src/routes/optimize.js       — POST /api/optimize/:id
+public/js/dashboard.js       — Frontend rendering
+```
+
+## Credits
+
+The 10 diagnostic rules were inspired by a community analysis of common OpenClaw cost pitfalls by [@li9292](https://x.com/li9292/status/2025081922410443243).
+
+## Learn More
+
+See the full series on [clawbridge.app/solutions](https://clawbridge.app/solutions) — 12 articles covering each diagnostic in depth.

docs/cost-control-center_CN.md

@@ -0,0 +1,58 @@
+# 成本控制中心 (Cost Control Center)
+
+**成本控制中心**是一个内置的诊断引擎，它可以分析您的 OpenClaw Agent 的 Token 使用情况，并提供可操作的建议以降低 API 成本。
+
+## 核心功能
+
+### Token 经济学仪表盘
+- **今日成本** — 实时运行总额
+- **累计总额** — 自安装以来的累计成本
+- **30 天预测** — 基于近期消耗模式预测的月度支出
+- **热门模型** — 按模型进行成本细分
+- **7 日历史柱状图** — 每日成本图表及详情点击查看
+
+### 自动化诊断 (10 项规则)
+
+| ID | 诊断项 | 说明 |
+|----|-----------|-------------|
+| A01 | 模型降级 (Model Downgrade) | 检测是否可以将简单的任务分配给更便宜的模型，并一键切换 |
+| A02 | 心跳优化 (Heartbeat Optimization) | 计算无用“心跳”产生的费用，提供分级心跳间隔建议 |
+| A03 | 丢弃上下文浪费 (Session Reset Waste) | 识别由于过于频繁重置 Session 导致的系统提示词重复加载浪费 |
+| A04 | 冗余技能清理 (Idle Skill Detection) | 找到那些不常用但一直在消耗 Input Token 的技能指令 |
+| A05 | 思考 Token 过载 (Thinking Overhead) | 评估扩展思考模式的成本，建议在简单任务中使用精简模式 |
+| A06 | 提示词缓存 (Prompt Caching) | 检测是否未开启缓存，开启后可节省高达 90% 的 Input 成本 |
+| A07 | 上下文紧缩防护 (Compaction Safeguard) | 防止由于对话过长导致的上下文费用超出预期 |
+| A08 | 本地模型路由 (Local Ollama Routing) | 将简单的请求/心跳路由到本地免费模型 *(开发中)* |
+| A09 | 回复精简 (Output Verbosity) | 检测由于回复过长导致的费用浪费，通过指令减少废话 |
+| A10 | 多模型路由 (Multi-Model Routing) | 根据任务复杂度自动分发到不同价格的模型 *(开发中)* |
+
+### 一键优化引擎
+- 每项诊断建议旁都有 **【Apply】** 按钮。
+- 修改前自动备份当前配置到 `data/backups/`。
+- 支持一键 **【Undo/Rollback】**，秒级恢复之前的设置。
+- 完整的优化日志记录在 `data/logs/optimizations.jsonl` 中。
+
+## API 接口
+
+```
+GET  /api/diagnostics   — 运行所有诊断检查，返回建议的操作及预计节省的金额
+POST /api/optimize/:id  — 应用特定的优化操作 (A01–A10)
+```
+
+## 工作原理
+
+1. **诊断引擎 (DiagnosticsEngine)** 读取您的 JSONL 使用日志和当前配置。
+2. 每个诊断规则会根据您的**实际数据**计算潜在的节省金额。
+3. 结果显示在前端仪表盘，包含具体的金额和配置差异 (diff)。
+4. 点击“应用”触发 **优化服务 (OptimizerService)**：
+   - 备份当前配置。
+   - 写入优化后的设置。
+   - 记录操作记录。
+
+## 鸣谢
+
+这 10 项诊断规则的灵感来源于 [@li9292](https://x.com/li9292/status/2025081922410443243) 对 OpenClaw 常见成本痛点的总结。
+
+## 深入了解
+
+查看官网上的详细系列文章：[clawbridge.app/solutions](https://clawbridge.app/solutions) —— 深入讲解每项诊断的原理与实测数据。

index.js

@@ -15,23 +15,27 @@ const { PORT, TUNNEL_TOKEN, LOG_DIR } = require('./src/config');
 const app = require('./src/app');
 const { setupWebSocket } = require('./src/websocket');
 const { checkSystemStatus } = require('./src/services/monitor');
-const { runAnalyzer } = require('./src/services/analyzer');
+const { runAnalyzer, setWss } = require('./src/services/analyzer');
 const { WORKSPACE_DIR } = require('./src/services/openclaw');
 const tunnel = require('./tunnel');
+const disableAnalyzer = process.env.COST_CONTROL_SKIP_ANALYZER === 'true';
 
 // --- Create Server ---
 const server = createServer(app);
 const wss = new WebSocketServer({ server });
 
 // --- WebSocket ---
 setupWebSocket(wss);
+setWss(wss);
 
 // --- Background Tasks ---
 setInterval(() => {
-    checkSystemStatus(() => {});
+    checkSystemStatus(() => { });
 }, 3000);
-runAnalyzer();
-setInterval(runAnalyzer, 60 * 60 * 1000);
+if (!disableAnalyzer) {
+    runAnalyzer();
+    setInterval(runAnalyzer, 60 * 60 * 1000);
+}
 
 // --- Startup ---
 async function main() {

package.json

@@ -22,7 +22,7 @@
     "@octokit/rest": "^22.0.1",
     "eslint": "^8.57.1",
     "google-translate-api-x": "^10.7.2",
-    "jest": "^29.0.0",
+    "jest": "^29.7.0",
     "prettier": "^3.0.0",
     "semver": "^7.7.4",
     "supertest": "^7.0.0"