Skip to main content

CLI API

TradeJS command-line interface is exposed as the @tradejs/cli package. Use it after package install via:

npx @tradejs/cli <command> [flags]

Run command-level help whenever you need the exact current syntax:

npx @tradejs/cli backtest --help

Command Map​

Setup and diagnostics:

  • doctor - validate required services and optional ML connectivity.
  • infra-init - initialize local infrastructure state.
  • infra-up / infra-down - start or stop local infrastructure.
  • server-health - check service health.
  • user-add - create a user config.
  • migration - run migration tasks.

Research and backtesting:

  • backtest - run strategy backtests or refresh market cache.
  • results - inspect, merge, update, or clear selected strategy results in Redis.
  • runtime-parity - replay runtime entries against deterministic backtest logic.
  • replay - replay stored runtime data.
  • runtime-evidence - collect runtime evidence.
  • replay-runtime-evidence - replay collected runtime evidence.
  • execution-calibration - inspect runtime execution assumptions.

Runtime and signals:

  • signals - evaluate runtime strategies on the latest closed candle.
  • signals-summary - summarize recent runtime signal/order state.
  • bot - run the Telegram bot.

research:auto and agent-run are source-repository maintainer workflows, not supported external package commands. They are intentionally outside this public command map even though the package currently contains their launchers.

Market data and maintenance:

  • continuity - check or repair candle continuity.
  • binance:market-ingest - ingest Binance market data.
  • candles:migrate-provider - migrate candle provider naming.
  • derivatives:ingest - ingest derivatives context.
  • derivatives:ingest:coinalyze:all - ingest Coinalyze derivatives context for all configured targets.
  • spread:ingest - ingest Binance/Coinbase spread context.
  • maintenance:cleanup-market-context - clean old market-context records.
  • clean-dir, clean-redis, clean-tests - maintenance cleanup helpers.

ML and AI:

  • ml-export - export ML dataset rows.
  • ml-inspect - inspect ML datasets.
  • ml-train:latest - train/select the latest ML dataset.
  • ai-export - export AI prompt/evaluation datasets.
  • ai-train - replay AI datasets and measure gate behavior.
  • ai-pocket-search - search deterministic AI-gate feature pockets.
  • test-ml, test-script - diagnostic test commands.

Common Examples​

# Check local dependencies
npx @tradejs/cli doctor

# Refresh candles without running tests
npx @tradejs/cli backtest --updateOnly --tickers BTCUSDT,ETHUSDT

# Run a cache-only backtest over one symbol
npx @tradejs/cli backtest --config MaStrategy:base --tickers BTCUSDT --cacheOnly

# Evaluate runtime signals without placing orders
npx @tradejs/cli signals --tickers BTCUSDT --cacheOnly --skipScreenshots

# Inspect selected backtest results for one strategy
npx @tradejs/cli results --strategy TrendLine --coverage

Backtest​

backtest is the main research command.

Ticker and execution scope:

  • -t, --tickers - comma-separated symbols.
  • -e, --exclude - comma-separated symbols to skip.
  • -l, --tickersLimit - cap the number of symbols.
  • -n, --tests - cap the number of config tests.
  • -s, --skip - skip the first N tests.
  • -p, --parallel - worker count.
  • -f, --timeframe - interval in minutes.
  • -d, --days - run only the recent N-day window.
  • --startTime, --endTime - explicit backtest window, epoch seconds or milliseconds.

Data and cache:

  • -u, --updateOnly - refresh market cache only.
  • -C, --cacheOnly - use cached market data only.
  • -L, --showTickersList - print resolved ticker list and exit.
  • -o, --connector - connector provider/name, for example bybit, binance, coinbase, or custom.

Config and output:

  • -c, --config - backtest config key, for example MaStrategy:base.
  • -T, --top - number of best result/config buckets to print for grid runs.
  • -g, --progressStep - progress logging interval.
  • -U, --user - Redis user config, default root.
  • --fast - skip per-test artifact persistence and keep summary/dataset output.
  • -K, --continue - continue the latest compatible interrupted run.
  • -R, --runId - explicit backtest run id to continue.

Execution assumptions:

  • --backtestPriceMode - delayed entry execution price mode: open, close, or mid.
  • --backtestEntryDelayBars - delay entry execution by N closed bars after the signal.

Dataset output:

  • -m, --ml - write ML dataset rows to per-worker JSONL chunks.
  • -A, --ai - write AI prompt rows to per-worker JSONL chunks.

Backtest config keys are stored in Redis under users:<user>:backtests:configs:<StrategyName:configName>. See Create a backtest config.

Signals​

signals evaluates configured runtime strategies on the latest closed candle.

  • -t, --tickers - comma-separated symbols.
  • -e, --exclude - symbols to skip.
  • -l, --tickersLimit - cap the number of symbols.
  • -f, --timeframe - interval in minutes.
  • -m, --makeOrders - allow order placement when strategies and connector config permit it.
  • -N, --notify - send Telegram notifications.
  • -S, --skipScreenshots - skip screenshot generation.
  • -u, --updateOnly - refresh market cache only.
  • -C, --cacheOnly - use cached market data only.
  • -L, --showTickersList - print resolved ticker list and exit.
  • -p, --parallel - signal evaluation worker count.
  • -R, --showSkipStats - show aggregated skip stats by strategy.
  • -c, --chunk - split the universe, for example 1/3.
  • -U, --user - Redis user config.
  • -o, --connector - connector provider/name.

Signals Summary​

signals-summary summarizes recent runtime signal/order state.

  • -u, --user - Redis user config.
  • --connector - connector provider/name for reconciliation.
  • -H, --hours - summary window in hours.
  • -P, --printOnly - print summary instead of sending it to Telegram.
  • --debugAttachment - attach runtime Redis debug JSON for replay diagnostics.

Results​

results inspects and manages selected strategy results saved in Redis.

  • -s, --strategy - strategy name. Required.
  • -C, --coverage - show coverage table.
  • -u, --update - update results config in Redis.
  • -m, --merge - merge results config in Redis.
  • -c, --clear - clear results config in Redis.
  • -V, --verbose - verbose output.
  • -U, --user - Redis user config.

Use this command to inspect candidate configs before promoting them into runtime settings.

Runtime Parity​

runtime-parity compares runtime entry records with deterministic backtest replay.

  • -u, --user - Redis user config and runtime journal.
  • -o, --connector - connector provider/name.
  • -d, --days - replay window in days.
  • -b, --startTime, -e, --endTime - explicit replay window, epoch seconds or milliseconds.
  • -s, --strategy - compare one strategy only.
  • -t, --tickers - replay comma-separated symbols for configured strategies.
  • -C, --cacheOnly - do not refresh market history before replay.
  • -a, --toleranceBars - allowed entry timestamp drift in bars.
  • --fullUniverse - replay every configured strategy across the connector universe.
  • --runtimeGates - force runtime AI/ML gates for all replay targets.
  • -N, --notify - send parity summary to Telegram.
  • -D, --details - print unmatched entry details.

AI Train​

ai-train replays exported AI rows and summarizes approval behavior.

  • -o, --outDir - dataset directory, default data/ai/export.
  • -s, --strategy - strategy filter for the merged file.
  • -f, --file - explicit merged dataset file path.
  • -n, --recent - recent rows to evaluate from the end, 0 means all rows.
  • -k, --skip - recent rows to skip before selecting replay rows.
  • -p, --parallel - concurrent AI requests.
  • -m, --model - model id override for replay.
  • -M, --minQuality - minimum quality required to approve entry.
  • -l, --localOnly - replay deterministic adapter gate without provider calls.
  • -U, --user - Redis user config.
  • -c, --chart - save compact chart data for the strategies UI.
  • -j, --json - print structured JSON summary.
  • -S, --since, -u, --until - restrict evaluated rows by timestamp.
  • -P, --period - trailing selected-row period such as last365d, last90d, or last30d.
  • -q, --qualityThresholds - comma-separated qN+ thresholds.
  • -d, --dumpEvaluations - write evaluated rows as JSONL.
  • -G, --dumpFeatures - feature snapshot for dumped rows: none, gateFeatures, or baseContext.
  • -Q, --symbolQuarantine - apply per-strategy/per-symbol quarantine overlay to approved rows.

ai-pocket-search searches deterministic gate feature pockets over AI exports.

Dataset selection:

  • -o, --outDir, -s, --strategy, -f, --file
  • -n, --recent, -k, --skip
  • -S, --since, -u, --until, -P, --period

Gate and search shape:

  • -M, --minQuality
  • -q, --qualityThresholds
  • -g, --scope - all, approved, rejected, or candidates.
  • -d, --maxDepth
  • -m, --minSupport
  • -F, --minProfitFactor
  • -W, --minWinRate
  • -R, --minTotalProfit
  • -a, --maxAtomicPredicates
  • -C, --maxCombinations
  • -V, --validationSplit
  • -N, --minValidationSupport
  • -D, --dedupeEquivalentSelections

Output and feature scope:

  • -t, --top
  • -Y, --includeSymbol
  • -E, --includeGateContext
  • -p, --featureProfile - compact or all.
  • -r, --reportDir
  • -B, --reportFile
  • -j, --json
  • -O, --output

Doctor​

doctor validates infrastructure assumptions before running workflows.

  • --require-ml - treat ML gRPC connectivity as required.
  • --skip-ml - skip ML gRPC checks.

Deep Dives​