Important

This repository is RETIRED (2026-07). All content merged into the stock-analysis skill in ReByteAI/rebyte-skills — the single financial skill (analysis + data-lake access + backtesting). The realtime-api-data pillar was retired entirely: the platform serves historical (T+1) lake data only, with no direct provider access. Do not clone or update this repo; use rebyte-skills/stock-analysis instead.

AnyFinancial

A top-level router skill for financial work on Rebyte. The umbrella SKILL.md runs a task as three phases:

1. Prepare the Data — hosted historical datasets (data/) or real-time/API market data (realtime-api-data/, US & CN).
2. Analyze the Data — write code, compute indicators, run event-driven backtests (backtesting/).
3. Generate the Report — finance templates (financial-templates/), always in the Kami style (report-style/), with price charts rendered via TradingView Lightweight Charts (financial-charts/).

Skills

Phase	Directory	What it does
1 · Prepare	data/	Read-only historical SQL over the Rebyte Financial Data Service (catalog → schema → query). Feeds backtests and research.
1 · Prepare	realtime-api-data/	Recent/realtime provider APIs. Self-contained US point-query behavior and CN Stocks provider access.
2 · Analyze	backtesting/	Realistic event-driven backtests with NautilusTrader. Phase-gated: setup → parameters → data → execute → report.
3 · Report	financial-templates/	Financial statement analysis, valuation, company/industry research, and finance output patterns. If no template fits, bring your own structure.
3 · Report	report-style/	The Kami design system — mandatory style for every HTML report.
3 · Report	financial-charts/	TradingView Lightweight Charts for price/OHLCV/candlestick/indicator/backtest-marker charts. ECharts stays for generic charts.

Layout

anyfinancial/                     ← umbrella skill (this repo)
├── SKILL.md                      ← top-level router (three phases)
├── data/                         ← Phase 1: hosted historical data access
│   ├── SKILL.md
│   ├── scripts/                  (anyfinancial_cli.py, shared/constants.json)
│   ├── data_builder/us_news/     (local mirror + semantic search)
│   ├── README.md · TEST_PLAN.md · .env.example
├── realtime-api-data/            ← Phase 1: recent/realtime provider APIs
│   ├── SKILL.md
│   ├── scripts/                  (anyfinancial_api_data.py)
│   ├── us/SKILL.md               (US point-query behavior)
│   └── cn/SKILL.md               (CN Stocks provider)
├── backtesting/                  ← Phase 2: analysis / backtest execution
│   ├── SKILL.md
│   ├── scripts/                  (setup.sh, fetch_data.py, run_backtest.py, run_workflow.sh)
│   ├── strategies/ · references/ · evals/
│   └── config.example.json
├── financial-templates/          ← Phase 3: finance report templates
│   ├── SKILL.md
│   ├── statement-analysis.md
│   ├── valuation-analysis.md
│   ├── company-research.md
│   ├── industry-research.md
│   ├── output-patterns.md
│   └── daily-market-review.html  (daily 复盘 + 交易指南, Kami HTML)
├── report-style/                 ← Phase 3: shared Kami style (mandatory form)
│   ├── README.md
│   ├── styles.css                (Kami tokens + components — link first)
│   ├── report.css                (report layer: doc flow, tables, --chart-* — link after)
│   ├── CHEATSHEET.md
│   └── design.md
├── financial-charts/             ← Phase 3: TradingView price/OHLCV charts
│   ├── SKILL.md
│   ├── templates/price-chart.html
│   └── examples/report-price-chart.json
├── evals/                        ← umbrella routing evals
└── README.md · .gitignore        ← ripple-level

How the skills compose

backtesting retrieves prices through data: its fetch_data.py calls the data skill's CLI (data/scripts/anyfinancial_cli.py) to pull OHLCV bars from us.eod / us.bars_1m, then runs them through NautilusTrader. Use data for historical exploration that supports backtesting.

realtime-api-data/us recreates the US point-query behavior inside this repo: latest price, latest bars, latest news, point fundamentals, and today's OHLC.

realtime-api-data/cn documents the CN Stocks provider: POST /api/data/cn-stocks/<operation> with operations bars, adj_factor, bars_1min, valuation, financials, universe, and news.

financial-templates contains template-only analysis structures. Use those files to shape outputs, but fill them only with verified provider data, filings, spreadsheets, or user-supplied inputs. If no template fits, use your own structure — the report form is still Kami (report-style/).

report-style is the shared Kami style, and every HTML report must use it (warm parchment canvas, ink-blue accent, serif-led hierarchy). Reports follow a two-part architecture: the shared style (report-style/styles.css + report-style/report.css) and a content template that holds only HTML + placeholders and links those two stylesheets — no inline or duplicated CSS. Restyle every report by editing the shared CSS once. financial-charts renders any price/OHLCV/candlestick/indicator/backtest-marker chart with TradingView Lightweight Charts — fill a declarative JSON spec plus a sibling data JSON and deliver an artifact (chart canvas colors come from the shared --chart-* tokens); ECharts stays for generic charts.

Quickstart

# Backtesting data support — discover and query historical data
python3 data/scripts/anyfinancial_cli.py catalog
python3 data/scripts/anyfinancial_cli.py schema us.eod
python3 data/scripts/anyfinancial_cli.py query "SELECT t, c FROM us.eod WHERE ticker='AAPL' ORDER BY t DESC LIMIT 5"

# Backtesting execution — install, then follow backtesting/SKILL.md
bash backtesting/scripts/setup.sh
cp backtesting/config.example.json backtesting/smoke.config.json
# Edit smoke.config.json to a short completed range before running.
bash backtesting/scripts/run_workflow.sh --config backtesting/smoke.config.json --splits full

# Real-time/API Data / CN — provider lookup
AUTH_TOKEN=$(/home/user/.local/bin/rebyte-auth)
API_URL=$(python3 -c "import json; print(json.load(open('/home/user/.rebyte.ai/auth.json'))['sandbox']['relay_url'])")
curl -fsS -X POST "$API_URL/api/data/cn-stocks/universe" \
  -H "Authorization: Bearer $AUTH_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{}'

# Real-time/API Data / US — local CLI
python3 realtime-api-data/scripts/anyfinancial_api_data.py us latest-price AAPL
python3 realtime-api-data/scripts/anyfinancial_api_data.py us latest-news TSLA --limit 5

# Financial Templates — template-only, no API call
sed -n '1,120p' financial-templates/valuation-analysis.md

Inside a Rebyte VM/workspace the backtesting data CLI and direct provider APIs use the sandbox token and relay URL from /home/user/.rebyte.ai/auth.json. See each pillar's SKILL.md for the full workflow.