Skip to content

Limitations

This project is intentionally constrained in its current implementation.

  • Local-only operations lab; no live exchange connectivity
  • NautilusTrader-only engine scope
  • Single example instrument scope (BTCUSDT)
  • Example data support is 1-minute OHLCV candles only (candles_1m.csv); orderbook / LOB data is not supported (deferred to future work)
  • Backtest path is a Nautilus engine smoke run over prepared candles, not a strategy performance report or custom strategy harness
  • RunSpec strategy fields (for example ops_smoke_demo) provide scenario identity and currently select only built-in local scenarios; no custom strategy loading or plugin-style strategy extension yet
  • RunSpec data.fingerprint is metadata/reserved and is not a runtime enforcement gate yet
  • RunSpec observability.journal|metrics|report fields are metadata/reserved; current lifecycle paths still write standard artifacts
  • RunSpec connectivity_readiness is local readiness contract metadata (env var names only), not real provider credential validation
  • In current examples, RunSpec venue values binance and binance_testnet are label-only; they do not indicate active exchange/testnet/live connectivity
  • tc connectivity readiness is local preflight only: env placeholder presence check, no network calls, no exchange/testnet/live connectivity
  • tc connectivity probe is local loopback-only preflight: read-only HTTP GET against loopback URL only, no exchange/testnet/live connectivity
  • Probe results are artifact-backed (connectivity_probe.json) and probe metrics are derived from artifacts, not live external telemetry
  • Backtest-vs-paper evidence is artifact-backed operational comparison, not strategy-performance analytics
  • Evidence status/known-gap panels summarize operational state, not profitability or execution quality
  • Probe does not store response body, response headers, or secret values
  • Readiness does not fetch balances, positions, account state, or external runtime state
  • Readiness does not submit/cancel orders and does not flatten positions
  • Probe does not submit/cancel orders and does not flatten positions
  • Readiness metrics are artifact-backed and do not expose env var names or env var values
  • No secret storage in readiness artifacts, metadata, journal, report, or metrics
  • Paper path is a bounded synthetic lifecycle skeleton with no market data feed and no exchange/testnet connectivity
  • Kill switch behavior is file-based and local
  • Reconciliation is file-based, fixture-driven, and report-only
  • Failure drills are deterministic artifact checks, not real orchestration
  • No profitability/alpha claim
  • No PnL/Sharpe/returns claim
  • No live/testnet paper-trading equivalence claim
  • No real order/fill/account/balance/position state evidence
  • No strategy optimization or plugin framework
  • No generic run-diff framework
  • No production-grade safety, reliability, or low-latency claim

The following failure-mode categories are explicitly deferred and do not appear in the current implementation or in 0.8.0:

  • Rate-limit exhaustion simulation — no local rate-limit fixture or model exists
  • Disconnect drill as a standalone command — covered by probe_unreachable in the connectivity probe workflow
  • Stale orderbook drill — orderbook / LOB data is not supported (see above)
  • External exchange / testnet disconnect events — no external connectivity in any current command

For the local failure modes that are supported, see Failure modes.

These limitations are deliberate to keep the current implementation focused on operational workflow quality rather than breadth or performance claims.