PhoenixClaw Ledger: Zero-Effort Financial Tracking
PhoenixClaw Ledger automatically extracts financial transactions from your daily conversations and payment screenshots, requiring zero manual input.
Core Capabilities
| Feature | Description |
|---|
| Semantic Expense Detection | AI identifies spending mentions in natural conversation |
| Screenshot Recognition |
Extracts transaction data from payment app screenshots |
|
Smart Categorization | Auto-categorizes based on merchant and context |
|
Budget Tracking | Monthly budget alerts and progress visualization |
|
Financial Insights | Pattern analysis integrated into journal Growth Notes |
|
Goal Management | Savings, budget control, habit, and wishlist goals |
|
Weekly Reports | Automated Sunday 9 PM spending recap |
|
Query Support | Real-time natural language financial inquiries |
|
Spending Trends | Multi-month analytical spending visualization |
|
Transaction Browser | Interactive complete transaction history view |
Workflow
As a PhoenixClaw plugin, Ledger hooks into the post-moment-analysis phase:
- 1. Receive Moments: Get identified moments from PhoenixClaw Core
- Detect Finances: Scan for expense/income signals in text and media
- Text: Semantic patterns (see
references/expense-detection.md)
- Media: Payment screenshots (see
references/payment-screenshot.md)
- 3. Extract Data: Parse amount, merchant, category, timestamp
- Categorize: Apply rules from INLINECODE3
- Deduplicate: Prevent double-counting same transaction
- Store: Write to INLINECODE4
- Export: Generate journal section using INLINECODE5
Explicit Triggers
While passive by design, users can interact directly:
- - "How much did I spend today/this week/this month?"
- "Show my expense breakdown"
- "Set my monthly budget to [amount]"
- "What are my top spending categories?"
- "Generate financial report for [period]"
- "Set a savings goal for [amount] by [date]"
- "Show my spending trends"
- "Browse all my transactions"
- "How am I doing on my goals?"
Output Structure
CODEBLOCK0
Configuration
Ledger-specific settings in ~/.phoenixclaw/config.yaml:
CODEBLOCK1
Cron & Scheduled Reports
Ledger uses PhoenixClaw Core's cron infrastructure plus additional scheduled tasks:
| Task | Schedule | Description |
|---|
| Daily Processing | 10 PM (via Core) | Extracts transactions, generates daily section |
| Monthly Report |
1st of month, 8 AM | Comprehensive monthly financial summary |
|
Weekly Summary | Sunday 9 PM (optional) | Weekly spending recap |
Daily Processing (Automatic)
No separate setup required. Ledger hooks into Core's nightly cron:
- - Core runs at 10 PM → triggers INLINECODE7
- Ledger activates, extracts finances, exports to journal
Monthly Report Setup
CODEBLOCK2
See references/cron-setup.md for full configuration details.
Documentation Reference
References (references/)
- -
expense-detection.md: Semantic patterns for conversation parsing - INLINECODE11 : Screenshot recognition and OCR extraction
- INLINECODE12 : Merchant to category mapping rules
- INLINECODE13 : Category definitions and hierarchy
- INLINECODE14 : Budget alerts and progress calculation
- INLINECODE15 : Pattern analysis for Growth Notes
- INLINECODE16 : Scheduled tasks and report automation
- INLINECODE17 : Financial goals and progress tracking
- INLINECODE18 : Natural language query templates and logic
Assets (assets/)
- -
expense-callout.md: Template for conversation-detected expenses - INLINECODE21 : Template for screenshot-detected expenses
- INLINECODE22 : Journal integration template
- INLINECODE23 : Monthly summary template
- INLINECODE24 : Annual summary template
技能名称: phoenixclaw-ledger
详细描述:
PhoenixClaw Ledger：零操作财务追踪
PhoenixClaw Ledger 能从你的日常对话和支付截图中自动提取财务交易记录，无需任何手动输入。
核心功能
| 功能 | 描述 |
|---|
| 语义化支出检测 | AI 识别自然对话中的消费提及 |
| 截图识别 |
从支付应用截图中提取交易数据 |
|
智能分类 | 根据商户和上下文自动分类 |
|
预算追踪 | 月度预算提醒与进度可视化 |
|
财务洞察 | 模式分析，整合到日记成长笔记中 |
|
目标管理 | 储蓄、预算控制、习惯和心愿单目标 |
|
周报 | 每周日晚上9点自动发送消费总结 |
|
查询支持 | 实时自然语言财务查询 |
|
消费趋势 | 多月份分析性消费可视化 |
|
交易浏览器 | 交互式完整交易历史查看 |
工作流程
作为 PhoenixClaw 的一个插件，Ledger 接入到 post-moment-analysis 阶段：
- 1. 接收时刻：从 PhoenixClaw 核心获取已识别的时刻
- 检测财务信息：扫描文本和媒体中的支出/收入信号
- 文本：语义模式（参见 references/expense-detection.md）
- 媒体：支付截图（参见 references/payment-screenshot.md）
- 3. 提取数据：解析金额、商户、类别、时间戳
- 分类：应用 references/merchant-category-map.md 中的规则
- 去重：防止同一笔交易重复计数
- 存储：写入 ~/PhoenixClaw/Finance/ledger.yaml
- 导出：使用 assets/daily-finance-section.md 生成日记章节
显式触发
虽然设计为被动模式，但用户可以直接交互：
- - 我今天/这周/这个月花了多少钱？
- 显示我的支出明细
- 将我的月度预算设置为[金额]
- 我的主要消费类别是什么？
- 生成[时间段]的财务报告
- 设定一个在[日期]前达到[金额]的储蓄目标
- 显示我的消费趋势
- 浏览我所有的交易
- 我的目标完成得怎么样了？
输出结构
~/PhoenixClaw/
├── Journal/
│ ├── daily/2026-02-02.md # 包含 💰 财务章节
│ └── weekly/2026-W05.md # 每周财务总结
│
└── Finance/ # 账本专用目录
├── ledger.yaml # 结构化交易数据
├── budget.yaml # 预算配置
├── goals.yaml # 财务目标追踪
├── transactions.md # 交易浏览器视图
├── monthly/
│ └── 2026-02.md # 月度财务报告
└── yearly/
└── 2026.md # 年度总结
配置
~/.phoenixclaw/config.yaml 中的账本专用设置：
yaml
plugins:
phoenixclaw-ledger:
enabled: true
default_currency: CNY # 或 USD, EUR 等
budget_monthly: 5000 # 月度预算金额
categories_custom: [] # 用户自定义类别
screenshot_confidence: 0.7 # 自动记录的最低置信度
定时任务与计划报告
Ledger 使用 PhoenixClaw 核心的 cron 基础设施以及额外的计划任务：
| 任务 | 计划 | 描述 |
|---|
| 每日处理 | 晚上10点（通过核心） | 提取交易，生成每日章节 |
| 月度报告 |
每月1日上午8点 | 全面的月度财务总结 |
|
每周总结 | 周日晚9点（可选） | 每周消费回顾 |
每日处理（自动）
无需单独设置。Ledger 接入核心的夜间 cron：
- - 核心在晚上10点运行 → 触发 post-moment-analysis
- Ledger 激活，提取财务信息，导出到日记
月度报告设置
bash
openclaw cron add \
--name PhoenixClaw Ledger 月度报告 \
--cron 0 8 1 \
--tz auto \
--session isolated \
--message 生成上个月的月度财务报告。
完整配置详情请参见 references/cron-setup.md。
文档参考
参考文献（references/）
- - expense-detection.md：用于对话解析的语义模式
- payment-screenshot.md：截图识别与 OCR 提取
- merchant-category-map.md：商户到类别的映射规则
- category-rules.md：类别定义与层级结构
- budget-tracking.md：预算提醒与进度计算
- financial-insights.md：用于成长笔记的模式分析
- cron-setup.md：计划任务与报告自动化
- goal-management.md：财务目标与进度追踪
- query-patterns.md：自然语言查询模板与逻辑
资产（assets/）
- - expense-callout.md：对话检测到的支出模板
- receipt-callout.md：截图检测到的支出模板
- daily-finance-section.md：日记集成模板
- monthly-report.md：月度总结模板
- yearly-report.md：年度总结模板
