Skip to content

Scope

In scope (current implementation)

  • Operations-focused run lab around NautilusTrader
  • One engine: NautilusTrader
  • One example instrument: BTCUSDT
  • RunSpec strategy metadata (name/version) as scenario identity for traceability (e.g. ops_smoke_demo) with one built-in local scenario path; no custom strategy execution surface yet
  • Two run modes: backtest and paper
  • Local-first execution model
  • Reproducible runs from run specs
  • Run metadata with config/data/code hashes
  • Predictable artifact layout, operational journal, and basic reports
  • Basic observability hooks
  • Reconciliation checks and kill switch behavior
  • Fixture-backed 1-minute OHLCV candle data prepare/fingerprint workflow
  • data.fingerprint and observability.* are currently metadata/reserved fields (not runtime gating toggles)
  • RunSpec connectivity_readiness contract with env placeholder validation and local-only readiness evaluation command
  • Readiness artifacts (connectivity_readiness.json), metadata summary patch, journal event, and optional report section update
  • Artifact-backed readiness Prometheus metrics derived from connectivity_readiness.json
  • Local loopback connectivity probe command (tc connectivity probe) with read-only HTTP GET
  • Probe artifacts (connectivity_probe.json), metadata summary patch, journal event, and optional report section update
  • Artifact-backed probe Prometheus metrics and Grafana visibility from connectivity_probe.json
  • Backtest-vs-paper operational evidence artifact workflow via tc evidence compare
  • Evidence JSON + markdown report under artifacts/evidence/<backtest_run_id>__<paper_run_id>/
  • Aggregate artifact-backed evidence metrics and Grafana evidence panels
  • Three failure drills:
  • stale market data
  • forced reconciliation mismatch
  • restart/recovery

Implemented in current repository scope:

  • Local data prepare/fingerprint workflow
  • Spec-driven Nautilus smoke backtest and bounded synthetic paper lifecycle skeleton
  • Artifact-driven metrics export
  • File-based kill switch workflow
  • File-based reconciliation checks
  • Deterministic failure drills with runbooks

Out of scope (current implementation)

  • Building a custom trading engine
  • Building a strategy library, quant research platform, or custom strategy plugin loader
  • PnL/performance comparison between backtest and paper
  • Claims of live/testnet equivalence from synthetic paper evidence
  • Orderbook / limit-order-book (LOB) data ingestion or replay
  • Live trading with production-grade reliability requirements
  • Real exchange/testnet/live connectivity checks in readiness or probe paths
  • Provider-side credential validation or secret verification
  • Account/balance/position fetching
  • Real order/fill quality analysis
  • Order submission, cancel/replace, or position flattening
  • External reconciliation against provider APIs
  • Generic all-purpose run diff framework
  • Adapter framework and multi-venue architecture
  • Low-latency gateway engineering
  • Profitability or alpha claims
  • Multi-engine or broad multi-instrument framework design
  • Generic Kubernetes platform work
  • Full production observability stack hardening

Local failure-mode visibility — making existing operational failure states inspectable through artifacts, metrics, and dashboard signals — is the focus of 0.8.0 — Expanded Failure Modes. See Failure modes for the full inventory.

Future work may expand depth, but this repository intentionally keeps the current implementation local, reproducible, and non-production.

Definition of Done (current scope)

The current scope is complete when a user can run one NautilusTrader backtest smoke workflow and one paper lifecycle skeleton workflow from run specs, generate a backtest-vs-paper operational evidence artifact, inspect artifacts/reports, run local readiness/probe checks, serve artifact-backed metrics for dashboards, exercise file-based safety/reconciliation behavior, and run/read deterministic failure drills and runbooks.