Why you can trust the output
Every claim below is backed by code in the public repo. No black box.
What the system does not prove
Public performance evidence is still small
The live track-record page is an activity record, not proof of market outperformance. Unverified backtests and sample curves are not shown.
Upstream data can disappear
Free public endpoints change, rate-limit and geo-block requests. Missing data is surfaced as unavailable and may stop a decision.
LLMs can still be wrong
Role separation and validators reduce some failures but do not eliminate hallucination or bad judgment. Source claims still need human verification.
Optional capabilities are deployment-specific
Macro, crypto, billing, scheduled briefs and other integrations work only when their packages, keys and protected jobs are configured and tested.
Real data sources
Production decisions do not use synthetic market-data fallbacks. Missing upstream data stays missing or fails explicitly; optional capabilities participate only when configured in the deployment.
- Optional OpenBB SDK + FRED REST → Macro analyst when configured
- SEC EDGAR XBRL → US fundamentals, point-in-time by filing date
- akshare → A-share OHLCV + fundamentals
- CCXT (Binance default) → Crypto OHLCV + technicals
- Reddit JSON → US/crypto news + sentiment, no API key
- 东方财富股吧 → A-share retail discussion mining
Strict no-lookahead
Every adapter enforces strict no-lookahead at the boundary. yfinance.info (current snapshot only) and akshare realtime endpoints return empty stubs for asof > 7 days; analyst prompt explicitly tells the LLM not to fabricate numbers. SEC EDGAR is filtered by filing date — zero leak. Reddit + Guba posts filtered by created_utc.
Cross-validated against Backtrader
Backtests can be replayed through Backtrader as a second engine. A result is labelled verified only after that cross-check actually completes.
Honest cost model
Backtest cost defaults are intentionally pessimistic — 5bp commission + 5bp slippage = 10bp per side, A-share sells add 5bp stamp tax. Higher than the industry-typical 3bp because under-charging gives misleading 'good' backtest results that disappoint live.
Auditable per-decision trace
The source repository is linked so adapters, prompts, tests, and deployment gates can be inspected. Paid subscriptions are not currently open.
Tests + CI
Automated regression tests run on every push via GitHub Actions, covering lookahead enforcement, costs, annualisation, EDGAR, Reddit, router selection and critical API regressions.
Run a decision and inspect its full trace, including data and reasoning stages.