Skip to content

"Vibe-Trading: Your Personal Trading Agent"

31.2 KB
Raw Blame

Data sources and capabilities

πŸ“‘ Data Sources & Smart Fallback

The get_market_data tool supports 28 market-data sources. With source: "auto", it infers the market from the symbol and tries that market’s configured chain. Some sources require credentials, a local terminal or your own files.

Source Markets Auth Role
tencent A-share + HK none first public source in the default A-share and HK chains
mootdx A-share (explicit-only) none retired from the default chain 2026-10: TDX servers stopped answering the client protocol it speaks (#1729)
eastmoney A / US / HK none OHLCV + deep fundamentals & flow tools (throttled)
baostock Β· akshare A (+ US/HK/futures/macro/fx) none free fallbacks
tushare A / HK equities, mainland indices and listed funds token daily bars
gildata A / HK / CN index GILDATA_TOKEN + market_quotes Market quotes (explicit Gildata requests only) Β· source="gildata"
yahoo US / HK / Canada / UK none direct chart/quotes/options; TSX .TO / TSXV .V; LSE .L with declared-currency normalization
sina Β· stooq US none K-line to 1984 Β· EOD CSV
yfinance US / HK / Canada / UK none wrapper; TSX .TO / TSXV .V; LSE .L under the same GBP/GBp contract
longbridge US / HK App Key + App Secret + Access Token optional historical OHLCV source; install the optional SDK
finnhub Β· alphavantage Β· tiingo Β· fmp US key optional providers
nobitex Β· wallex crypto (Iranian Toman pairs) none public UDF endpoints; explicit-only β€” the only Toman-quoted sources, so they never join the crypto chain where a USD-quoted series could stand in for one
okx Β· ccxt Β· binance crypto none OKX + 100+ exchanges + Binance historical / USD-M perps
futu HK / A OpenD optional local FutuOpenD
mt5 forex / metals MT5 terminal MetaTrader 5 (Exness-style) forex / metal bars, 1m–1D
tickerall forex / metals key + account (read-only) same broker MT5 feed, hosted β€” no local terminal, any OS (explicit-only, never in auto fallback)
pykrx Korea (KRX: KOSPI/KOSDAQ) none daily KOSPI / KOSDAQ bars for .KS / .KQ (optional krx extra)
india_broker India (NSE/BSE) broker login read-only Zerodha / Shoonya / Dhan bars for .NS / .BO (fallback-chain tail)
b3_cotahist Brazil (B3) none official daily B3 cash/odd-lot bars (raw, BRL) from the local store populated by python -m src.br_data sync; offline fallback for .SA
local any none your own CSV / Parquet / DuckDB via local: prefix

Local sources can read bar dates from a DatetimeIndex, with or without a configured date_format; an explicitly mapped date column takes precedence.

Fallback chains (by IP-ban risk):

  • A-share β†’ tencent Β· eastmoney Β· baostock Β· akshare Β· tushare Β· local
  • US β†’ yahoo Β· stooq Β· sina Β· eastmoney Β· yfinance Β· tiingo Β· fmp Β· finnhub Β· alphavantage Β· longbridge Β· akshare Β· local
  • HK β†’ tencent Β· eastmoney Β· yahoo Β· futu Β· akshare Β· yfinance Β· tushare Β· longbridge Β· local
  • India (NSE/BSE) β†’ yahoo Β· yfinance Β· india_broker Β· local
  • Brazil (B3 .SA) β†’ yahoo Β· yfinance Β· b3_cotahist Β· local
  • Korea (KOSPI/KOSDAQ) β†’ pykrx Β· yahoo Β· yfinance Β· local
  • Indices β†’ yahoo Β· yfinance Β· local
  • UK (LSE) β†’ yahoo Β· yfinance Β· local (declared GBP/GBp quotes only)
  • Crypto β†’ okx Β· binance Β· ccxt Β· yfinance Β· local
  • Forex / metals β†’ mt5 Β· akshare Β· yfinance Β· local
  • Canada / Vietnam / Argentina β†’ yahoo Β· yfinance Β· local
  • China futures β†’ akshare Β· local  Β·  funds β†’ tushare Β· akshare Β· local  Β·  macro β†’ akshare Β· tushare Β· local

A-share installation and availability

The base install includes Tencent's loader (standard-library HTTP) and AKShare. BaoStock is an optional, token-free TCP source for daily A-share bars; install the ashare extra to make it available:

pip install --upgrade "vibe-trading-ai[ashare]"
# For a uv tool installation that previously used only the base package:
uv tool install --force --python 3.12 "vibe-trading-ai[ashare]"

For uv tool users, --force rebuilds the existing tool environment. Keep any extras you already use when adding ashare (for example, "vibe-trading-ai[ashare,deepseek]") and repeat any existing --with arguments in that command.

mootdx is not included in the base install or the ashare extra. Its legacy loader is explicit-only and retired from automatic fallback; installing its SDK does not guarantee compatibility with the TDX server protocol. A missing mootdx import is expected with the base install.

Automatic fallback skips unavailable loaders. If an Eastmoney-backed request is throttled, select source: "auto" or an available alternative such as tencent or baostock. The startup preflight covers selected services, not every optional loader; an SDK reported as installed does not establish that its remote data service is reachable.

Using Longbridge explicitly

Longbridge is an optional US/HK historical OHLCV loader. Install its SDK with:

pip install "vibe-trading-ai[longbridge]"

Configure the three credentials in .env:

LONGBRIDGE_APP_KEY=...
LONGBRIDGE_APP_SECRET=...
LONGBRIDGE_ACCESS_TOKEN=...

For a backtest, set source in config.json:

{
  "codes": ["QQQ.US"],
  "start_date": "2025-01-01",
  "end_date": "2025-01-10",
  "interval": "1D",
  "source": "longbridge"
}

In an Agent conversation, ask explicitly: "Use Longbridge to fetch QQQ.US historical data." The explicit source request is separate from source: "auto"; auto keeps the normal per-market fallback chain.

Beyond OHLCV, 29 read-only data tools reach into fundamentals & flow β€” fund flow, dragon-tiger, northbound, southbound, margin, block trades, shareholder count, lockup, sector, research reports, broker research search, news, SEC filings, Chinese-market announcements, financial statements, PIT fundamentals, options chains, crypto order-book depth, stock profile, market screening, symbol search, macro, China macro (EDB), iwencai, institutional holdings (13F), ETF look-through, off-exchange fund NAV, prediction markets, and research papers β€” all exposed over MCP. An explicit local: symbol never silently falls back to a network source.

RSSHub event dates and ticker formatting

RSSHub daily event enrichment uses calendar-date keys: timezone-aware as_of values and daily bar labels retain their displayed date, matching the provider’s knowable-date policy. This does not convert intraday publication instants between markets.

RSSHub feed code_style transformations separate the final venue suffix from the ticker, preserving share-class dots: BRK.B.US becomes BRK.B for bare routes and USBRK.B for exchange_prefix; raw remains unchanged.

Data partner Β· Gildata

Each Gildata feature requires a token and a separate opt-in under Settings β†’ Data sources. This includes fund NAVs, macro indicators, announcements, broker reports, combined queries, screening, symbol search, portfolio diagnosis and explicitly requested quotes. All nine features are off by default. A saved token alone enables none of them, including after an upgrade. Even with market quotes enabled, Gildata is used only when explicitly selected. Free sources never fall back to Gildata.


πŸ”© Detailed Capabilities

Detailed inventories are folded below to keep the main README scannable. Open them when you want to inspect the available building blocks.

Finance Skill Library 89 skills across 9 categories
  • πŸ“Š 89 specialized finance skills organized into 9 categories
  • 🌐 Complete coverage from traditional markets to crypto & DeFi
  • πŸ”¬ Comprehensive capabilities spanning data sourcing to quantitative research
Category Skills Examples
Data Source 9 data-routing, tushare, yfinance, okx-market, akshare, mootdx, ccxt, eastmoney, sec-edgar
Strategy 19 strategy-generate, cross-market-strategy, technical-basic, candlestick, ichimoku, elliott-wave, smc, multi-factor, ml-strategy
Analysis 23 factor-research, correlation-regime, macro-analysis, global-macro, valuation-model, investor-lenses, credit-analysis, dividend-analysis
Asset Class 9 options-strategy, options-advanced, convertible-bond, etf-analysis, asset-allocation, sector-rotation
Crypto 7 perp-funding-basis, liquidation-heatmap, stablecoin-flow, defi-yield, onchain-analysis
Flow 8 hk-connect-flow, us-etf-flow, edgar-sec-filings, financial-statement, adr-hshare
Tool 10 backtest-diagnose, report-generate, pine-script, doc-reader, web-reader, vnpy-export, trade-journal
Research 3 alpha-zoo, strategy-dev-manager, strategy-discovery
Risk Analysis 1 ashare-pre-st-filter

The read-only report_audit tool preserves signed financial values in Markdown tables and label: value lines, so losses are checked as losses rather than profits. Backtest reports ground their figures in the engine's own output for that call. For list results, equivalent references such as data.positions[0].contribution_pct and data.positions.0.contribution_pct select the same element; an explicit index cannot borrow another element's value. Correction hints suggest exact references without treating the suggestion itself as evidence.

Custom Data Source register your own historical OHLCV loader

Need a market or vendor we don't ship a loader for? Add your own historical-bar loader and select it with source="<name>". The steps edit package source, so run from a clone (pip install -e .).

  1. Write the loader β€” create agent/backtest/loaders/<name>_loader.py with a class that satisfies DataLoaderProtocol (duck-typed, no base class needed) and is tagged with @register:

    import pandas as pd
    from backtest.loaders.registry import register
    
    @register
    class DataLoader:
        name = "mysource"            # the value you pass as source=
        markets = {"us_equity"}      # a_share/us_equity/hk_equity/crypto/futures/fund/macro/forex
        requires_auth = False
    
        def is_available(self) -> bool:
            return True              # token present? network reachable?
    
        def fetch(self, codes, start_date, end_date, *, interval="1D", fields=None):
            # return {symbol: DataFrame indexed by trade_date,
            #         columns: open, high, low, close, volume}
            ...
    
  2. Register the module so @register fires β€” add "backtest.loaders.<name>_loader" to _loader_modules in agent/backtest/loaders/registry.py.

  3. Allow the name through config validation β€” add "mysource" to _VALID_SOURCES in agent/backtest/runner.py.

  4. (Optional) slot it into a market's FALLBACK_CHAINS in registry.py so source="auto" can reach it.

  5. Use it β€” source="mysource" in a backtest config, or via the CLI / agent.

Real-time ticks / order-book depth are out of scope for loaders β€” the loader layer is point-in-time historical bars only. Live market data flows through the broker connectors instead: okx / binance / ccxt for crypto, futu / tiger for equities.

Broker Connectors 19 brokers β€” read + paper, bounded-live where supported

Connector-first profiles. Most do read + paper-account order placement β€” IBKR is read-only, Robinhood is live-only (no paper account), Scalable Capital is read-only (no paper account exists at all), mStock (India) is read-only live over its Type B REST API (no paper account), and Trading 212 and Toss Securities refuse order placement entirely, paper included; live order placement is bounded by a user-defined mandate (symbol allowlist, order-size / exposure caps, daily trade cap, instant kill switch) and never holds funds β€” the broker executes. Order-placing tools stay off MCP (agent + CLI only). Research / backtest paths are structurally barred from any live endpoint.

Broker Markets Capabilities
IBKR global local TWS / Gateway, read-only
Robinhood US Agentic MCP (desktop OAuth) β€” equity read + bounded live; options order placement is not supported
Scalable Capital DE / EU Agentic MCP (desktop OAuth) β€” fully read-only; no paper account exists
mStock India (NSE/BSE) Separate holdings/net-position reads; no portfolio aggregation or orders
Tiger US / HK / A read + paper + bounded live
Alpaca US read + paper + bounded live (+ TAP credential-isolation mode)
OKX Β· Binance crypto read + paper + bounded live
Futu HK / US / A read + paper + bounded live
eToro global read + paper + bounded live (Public API; demo keys reach only /demo paths, plus copy-trading workflows)
MetaTrader 5 forex / CFD read + paper + bounded live (Exness-style; demo ⇔ paper identity guard)
KIS (ν•œκ΅­νˆ¬μžμ¦κΆŒ) Korea (KOSPI/KOSDAQ) read + genuine paper (λͺ¨μ˜νˆ¬μž, a real broker-side sandbox on a separate host) β€” live order placement not yet wired for this market
Longbridge Β· Dhan Β· Shoonya Β· Zerodha Β· Upbit US / HK Β· India (NSE/BSE) Β· Korea (crypto) read + paper only β€” no runtime paper/live discriminator, so live order placement is hard-refused
Trading 212 Β· Toss Securities UK / EU Β· Korea/US fully read-only β€” place_order / cancel_order hard-refuse even paper (no verified sandbox)

Paper-vs-live is a structural per-broker runtime guard (account-id format, host separation, demo flag, or trade environment), never a config flag the agent can flip. A broker exposing no such discriminator is capped at paper + read-only.

Preset Trading Teams 30 swarm presets
  • 🏒 30 ready-to-use agent teams
  • ⚑ Pre-configured finance workflows
  • 🎯 Investment, trading & risk management presets
Preset Workflow
investment_committee Bull/bear debate β†’ risk review β†’ PM final call
global_equities_desk A-share + HK/US + crypto researcher β†’ global strategist
crypto_trading_desk Funding/basis + liquidation + flow β†’ risk manager
earnings_research_desk Fundamental + revision + options β†’ earnings strategist
macro_rates_fx_desk Rates + FX + commodity β†’ macro PM
quant_strategy_desk Screening + factor research β†’ backtest β†’ risk audit
technical_analysis_panel Classic TA + Ichimoku + harmonic + Elliott + SMC β†’ consensus
risk_committee Drawdown + tail risk + regime review β†’ sign-off
global_allocation_committee A-shares + crypto + HK/US β†’ cross-market allocation

Plus 20+ additional specialist presets β€” run vibe-trading --swarm-presets to explore all. Bring your own: drop preset YAMLs into ~/.vibe-trading/swarm/presets/ β€” they are listed alongside the bundled roster (same-name files override it, like user skills) and survive upgrades.

Alpha Zoo 478 pre-built quant alphas across 5 families
  • 🧬 478 cross-sectional alphas, lookahead-banned at the operator layer
  • πŸ“ˆ IC + IR + alive/reversed/dead categorisation in one CLI command
  • πŸ”¬ AST purity gate + 300-row lookahead sentinel test + pytest-socket network kill-switch
  • πŸ“¦ Apache-2 attribution for Qlib; per-zoo LICENSE.md declaring formulas as mathematical content
  • 🀝 Developer Certificate of Origin (DCO) sign-off workflow for community PRs
Zoo Count Source License
qlib158 154 Microsoft Qlib Alpha158 (Apache-2.0, commit-pinned) Apache-2.0
alpha101 101 Kakushadze (2015), "101 Formulaic Alphas", arXiv:1601.00991 Formulas are mathematical content
gtja191 191 Guotai Junan (2014), "191 Short-period Trading Alpha Factors" Formulas are mathematical content
academic 12 Fama-French 5 + Carhart momentum + Jegadeesh reversal + George-Hwang 52-week-high + Amihud illiquidity + Harvey-Siddique skew + Frazzini-Pedersen betting-against-beta + correlation-rewiring stability (price-based proxies) Public academic literature
fundamental 20 PIT-safe SEC company facts (14, filed-date anchored: earnings yield, book-to-market, ROE, ROA, gross profitability, growth, accruals, leverage, cash-flow yields …) + A-share P/B, P/S TTM and dividend yield (3) + their Shenwan industry-neutral variants (3) Public financial data

Run vibe-trading alpha list to browse, vibe-trading alpha show <id> for formulas + source, vibe-trading alpha bench --zoo X --universe Y --period Z to score a whole zoo, and vibe-trading alpha compare --all to rank zoos side by side.

Backtest Engines 12 engines + options portfolio, cross-market composite
Engine Market Notes
ChinaA A-share T+1, price limits, pre-ST filter
GlobalEquity US / HK / Canada / UK same-session trading; market-specific lots, ticks, settlement currencies, and costs
IndiaEquity India (NSE/BSE) T+1, circuit bands, config-driven STT / stamp / SEBI / GST cost stack
KoreaEquity Korea (KRX: KOSPI/KOSDAQ) long-only, Β±30% band judged at execution time on the unified tick grid, 2026 0.20% transaction tax
VietnamEquity Vietnam (HOSE) long-only, T+2 settlement hold, Β±7% band on the 10/50/100-VND tick grid, 100-share lots, 0.1% sell-side tax
BrazilEquity Brazil (B3) BRL, B3 trading + settlement fees, configurable broker fee, 1- or 100-share lots, R$0.01 tick, shorting allowed
ArgentinaEquity Argentina (BYMA) ARS, long-only; T+2 β†’ T+1 (2024-05-27), settlement approximated from observed sessions; same-session sale netting; declared CEDEAR ratios and credit dates; explicit tick table/panel (ar_tick_table) or per-instrument tick
Crypto crypto spot / USD-M perps funding settlements, execution/mark split
ChinaFutures Β· GlobalFutures futures margin, contract multipliers
Forex FX / metals via the mt5 loader (local terminal) or the hosted tickerall loader (no terminal, any OS)
Composite cross-market one shared capital pool across markets (source="auto")
options_portfolio options multi-leg, Greeks, payoff/scenario

Benchmark selection honors .SH, .SZ, and .BJ A-share suffixes, plus Yahoo's .SS spelling for Shanghai, independently of the chosen loader, including offline local sources.

Bars: 1m / 5m / 15m / 30m / 1H / 4H / 1D, plus weekly / monthly (1W / 1M, built from daily bars). 15 metrics + benchmark comparison, 5 portfolio optimizers (equal-volatility / risk-parity / mean-variance / max-diversification / turnover-aware), and 3 validation tools (Monte Carlo / Bootstrap / Walk-Forward).

The turnover-aware optimizer validates per-name and group caps for single-asset allocations too. It preserves feasible supplied allocations (including cash), and raises an error when the singleton allocation exceeds a cap.

When the opt-in market-data cache is enabled, local-source cache entries are scoped to the configured file, query, and column mapping. Changing a symbol's source declaration does not reuse bars from its previous source. The settled-data cache policy for an unchanged declaration remains the same. Unusable cached metadata is treated as a cache miss: the declared local source is read again, and the replacement entry can serve subsequent cache hits. The turnover-aware optimizer measures allocation changes in signed position space: reversing a 50% long to a 50% short contributes 0.5 turnover, including both the close and the reopen. Its penalty and recorded allocation turnover include these reversals; execution-derived turnover remains a separate metric. Local data date ranges include the entire UTC end day, including subsecond timestamps; bars at the following midnight are excluded before aggregation.

The shared backtest market classifier recognizes BTC/USD, ETH/USD, BNB/USD, SOL/USD, ADA/USD, and DOGE/USD as crypto. Fiat pairs such as GBP/USD, EUR/USD, and USD/JPY retain their forex classification.

Quant Library 310 tested functions across 25 modules, callable from every transport

src/quantlib holds one tested implementation of each piece of finance math the agent needs. Skills import these rather than carrying formulas inside markdown code blocks β€” if you find a pricing formula living in a SKILL.md, that is a bug, not a pattern.

Module What it covers
options Black-Scholes price + greeks, implied-volatility inversion
fixedincome Bond math, Nelson-Siegel / Svensson curve fitting; schedules, accrual, key-rate tenors and yield inversion reject non-finite inputs by name.
credit Altman Z-score, Merton / KMV distance-to-default; CDS premium dates retain their configured frequency, with a final stub when maturity falls between coupons.
timeseries Stationarity, cointegration, GARCH, bootstrap; OU fits use adjacent observed lag pairs; missing observations never become artificial one-step transitions.
risk Β· var_backtest VaR / CVaR / EVT and their backtests
attribution Brinson-Fachler decomposition
performance Β· fundmath TWR / MWR / Modified Dietz; XIRR / MOIC / DPI / TVPI
factormodel Β· eventstudy Factor regressions, event studies; portfolio style exposure reports incomplete factor rows as unmatched weight.
neutralization Optional per-observation OLS/WLS factor residualization, audited control quality selection, and cross-sectional PCA at any panel frequency. Workflow and examples.
combination Optional walk-forward IC/ICIR-weighted composite of several factor panels, with per-factor and composite diagnostics; a caller-supplied weight table handles any other method. Workflow and checklist.
multipletesting Β· crossvalidation Deflated significance, purged CV; purged walk-forward splits omit folds with no training observations left after purging; Sharpe significance summaries reject non-finite inputs by name.
impact Market-impact models
volatility Heston (1993) stochastic-volatility pricing; Heston pricing requires finite inputs and a positive integration limit; invalid inputs return a domain error.
portfolio Hierarchical Risk Parity allocation
copula Gaussian and Archimedean copulas
microstructure VPIN, Roll spread, Amihud illiquidity, Kyle's lambda

The read-only quantlib_call tool reaches all of it through one contract, so the finance math works on the CLI, the Web UI, the REST API and MCP β€” including deployments where bash is gated off. It is structurally not a shell β€” module allowlist, __all__-only dispatch, export_* refused. Econometrics needs the stats extra (pip install "vibe-trading-ai[stats]"); those functions lazy-import and tell you which one is missing.

Valuation & Institutional Research DCF, comps, three-statement, and six research commands

A valuation engine that refuses to invent its own inputs. The one rule in contracts.py: a missing input makes a model NOT RUNNABLE and is never silently defaulted β€” every default in a valuation model is an opinion wearing a constant's clothes.

Model Behaviour worth knowing
run_dcf FCFF bridge, WACC build, mid-year discounting, net-debt bridge, WACCΓ—g sensitivity grid. Dual terminal value: each method is cross-checked against the other's implied multiple and implied g
run_comps EV bridge, LTM + calendar-year calendarisation, multiple matrix. A peer with a non-positive denominator is excluded and reported, never averaged in as a negative multiple
threestatement Linked projection with a hard balance assertion, an explicit revolver plug, and an iterated interest↔debt circularity that must converge or raise

Artifacts are input-hashed and versioned, with xlsx / pptx export.

Six slash commands drive the workflows β€” /comps /dcf /attrib /memo /earnings /screen β€” each carrying a step skeleton and an arithmetic-consistent worked example (the Brinson decomposition sums exactly to active return; the earnings bridge sums exactly to the EPS delta). The investor-lenses skill stacks named-investor reasoning frameworks on top as analysis overlays: each lens is an operating procedure β€” priority signals, disqualifying conditions, typical misuse β€” not a biography, and names no tool.

Beyond bars, src/entities ingests irregular dated cash flows (NAVs, capital calls, coupons) and cashflow_performance reports XIRR / MOIC / DPI / TVPI / TWR / Modified Dietz / MWR over them. This path is deliberately parallel to the bar engines so a nav column can never reach one and get priced as a close. Flows use the holder's sign convention: a transfer into the portfolio is negative, and a transfer out is positive.

Governance & Audit Trail answer "what methodology produced that number?"

Every run writes a manifest hashing the prompt, the skill contents, the tool registry and the package versions, so a number produced last month can be traced to the exact methodology that produced it.

The audit ledger chains each record to its predecessor's hash and fsyncs, so editing or deleting a record is detectable β€” and an edit that recomputes its own hash is still caught one record later via prev_hash_mismatch. Timestamps are always caller-supplied; no module here calls datetime.now().

Trace redaction is sink-aware: tool-call arguments and the live audit ledger use a fail-closed sink where content stays redacted, while the tool-result sink releases it and pattern-scrubs its string leaves. env is never released in either.

πŸ“ Project Structure

Click to expand
Vibe-Trading/
β”œβ”€β”€ agent/                          # Backend (Python)
β”‚   β”œβ”€β”€ cli/                        # CLI package β€” interactive TUI + subcommands
β”‚   β”œβ”€β”€ api_server.py               # FastAPI server β€” runs, sessions, upload, swarm, SSE
β”‚   β”œβ”€β”€ mcp_server.py               # MCP server β€” 81 tools for OpenClaw / Claude Desktop
β”‚   β”‚
β”‚   β”œβ”€β”€ src/
β”‚   β”‚   β”œβ”€β”€ agent/                  # ReAct agent core
β”‚   β”‚   β”‚   β”œβ”€β”€ loop.py             #   5-layer compression + read/write tool batching
β”‚   β”‚   β”‚   β”œβ”€β”€ context.py          #   system prompt + auto-recall from persistent memory
β”‚   β”‚   β”‚   β”œβ”€β”€ skills.py           #   skill loader (89 bundled + user-created via CRUD)
β”‚   β”‚   β”‚   β”œβ”€β”€ tools.py            #   tool base class + registry
β”‚   β”‚   β”‚   β”œβ”€β”€ memory.py           #   lightweight workspace state per run
β”‚   β”‚   β”‚   β”œβ”€β”€ frontmatter.py      #   shared YAML frontmatter parser
β”‚   β”‚   β”‚   └── trace.py            #   execution trace writer
β”‚   β”‚   β”‚
β”‚   β”‚   β”œβ”€β”€ memory/                 # Cross-session persistent memory
β”‚   β”‚   β”‚   └── persistent.py       #   file-based memory (~/.vibe-trading/memory/)
β”‚   β”‚   β”‚
β”‚   β”‚   β”œβ”€β”€ tools/                  # 118 auto-discovered agent tools
β”‚   β”‚   β”‚   β”œβ”€β”€ backtest_tool.py    #   run backtests
β”‚   β”‚   β”‚   β”œβ”€β”€ remember_tool.py    #   cross-session memory (save/recall/forget)
β”‚   β”‚   β”‚   β”œβ”€β”€ skill_writer_tool.py #  skill CRUD (save/patch/delete/file)
β”‚   β”‚   β”‚   β”œβ”€β”€ session_search_tool.py # FTS5 cross-session search
β”‚   β”‚   β”‚   β”œβ”€β”€ swarm_tool.py       #   launch swarm teams
β”‚   β”‚   β”‚   β”œβ”€β”€ web_search_tool.py  #   DuckDuckGo web search
β”‚   β”‚   β”‚   └── ...                 #   bash, file I/O, factor analysis, options, alpha browser + bench, etc.
β”‚   β”‚   β”‚
β”‚   β”‚   β”œβ”€β”€ factors/                # Alpha Zoo β€” 478 alphas across 5 families
β”‚   β”‚   β”‚   β”œβ”€β”€ base.py             #   19 operators (rank/scale/ts_*/delta/decay_linear/safe_div/vwap)
β”‚   β”‚   β”‚   β”œβ”€β”€ registry.py         #   AST-only metadata load + lazy compute + sanity gates
β”‚   β”‚   β”‚   β”œβ”€β”€ bench_runner.py     #   IC + alive/reversed/dead categorisation
β”‚   β”‚   β”‚   └── zoo/                #   qlib158 (154) + alpha101 (101) + gtja191 (191) + academic (12) + fundamental (20)
β”‚   β”‚   β”‚
β”‚   β”‚   β”œβ”€β”€ api/                    # FastAPI route modules
β”‚   β”‚   β”‚   └── alpha_routes.py     #   /alpha/list, /alpha/{id}, /alpha/bench, SSE stream
β”‚   β”‚   β”‚
β”‚   β”‚   β”œβ”€β”€ skills/                 # 89 finance skills in 9 categories (SKILL.md each)
β”‚   β”‚   β”œβ”€β”€ swarm/                  # Swarm DAG execution engine
β”‚   β”‚   β”‚   └── presets/            #   30 swarm preset YAML definitions
β”‚   β”‚   β”œβ”€β”€ session/                # Multi-turn chat + FTS5 session search
β”‚   β”‚   └── providers/              # LLM provider abstraction
β”‚   β”‚
β”‚   └── backtest/                   # Backtest engines
β”‚       β”œβ”€β”€ engines/                #   11 engines + composite cross-market engine + options_portfolio
β”‚       β”œβ”€β”€ loaders/                #   28 sources: tushare, okx, nobitex, wallex, binance, yfinance, akshare, baostock, tencent, mootdx, ccxt, futu, pykrx, local, eastmoney, sina, stooq, yahoo, finnhub, alphavantage, tiingo, fmp, longbridge, mt5, india_broker, tickerall, gildata, b3_cotahist
β”‚       β”‚   β”œβ”€β”€ base.py             #   DataLoader Protocol
β”‚       β”‚   └── registry.py         #   Registry + auto-fallback chains
β”‚       └── optimizers/             #   MVO, equal vol, max div, risk parity
β”‚
β”œβ”€β”€ frontend/                       # Web UI (React 19 + Vite + TypeScript)
β”‚   └── src/
β”‚       β”œβ”€β”€ pages/                  #   Home, Agent, AlphaZoo, RunDetail, Compare, Correlation, Settings
β”‚       β”œβ”€β”€ components/             #   chat, charts, layout
β”‚       └── stores/                 #   Zustand state management
β”‚
β”œβ”€β”€ Dockerfile                      # Multi-stage build
β”œβ”€β”€ docker-compose.yml              # One-command deploy
β”œβ”€β”€ pyproject.toml                  # Package config + CLI entrypoint
β”œβ”€β”€ tools/                          # Repo-level CI helpers
β”‚   └── ci_grep_gates.sh            # rejects yaml.load / trademark / per-stock-data leaks
└── LICENSE                         # MIT