v4.4.50

What's Changed

• v4.4.50 - Indicators hardening + ThetaData acceptance stabilization by @grzesir in #967

Full Changelog: v4.4.49...v4.4.50

v4.4.49

What's Changed

• v4.4.49 - Parquet backtest artifacts + contract logs by @grzesir in #966

Full Changelog: v4.4.48...v4.4.49

v4.4.48

What's Changed

• v4.4.48: Emit Parquet backtest artifacts (BotSpot perf) by @grzesir in #965

Full Changelog: v4.4.47...v4.4.48

v4.4.47

What's Changed

• v4.4.47 - IBKR seconds timesteps + crypto futures expirations + backtest budget env + downloader queue rename by @grzesir in #964

Full Changelog: v4.4.46...v4.4.47

v4.4.46

What's Changed

• v4.4.43 - Hotfix: add_ohlc actually shipped by @grzesir in #958
• fix get trading days by @brettelliot in #961
• Version 4.4.46 by @grzesir in #963
• v4.4.44 - Hotfix: ship add_ohlc in PyPI wheel by @grzesir in #959

Full Changelog: v4.4.43...v4.4.46

v4.4.45

Full Changelog: v4.4.44...v4.4.45

4.4.39

Highlights

• Router backtesting (IBKR futures/cont_future/crypto): prefetch full backtest window once per series and slice from memory to avoid per-iteration history fetches (major warm-cache speedup).
• Indicators: fix plot_indicators hovertext generation when detail_text is missing/NaN/NA.

Docs

• Added router IBKR speed investigation ledger and backtesting docs updates.

Full diff: PR #953.

4.4.33

4.4.33 - 2026-01-12

Fixed

• SMART_LIMIT (live): avoid scanning full tracked order history in the background loop by using the broker’s active-order fast path, preventing high RSS growth in accounts with large historical order lists.
• Backtesting (router): make dataset lookup timestep-aware so minute requests don’t accidentally resolve to daily Data objects, and routed crypto assets passed as (base, quote) work reliably.
• Backtesting (router): refactor multi-provider routing to a provider registry + adapters (no hard-coded branching), add alpaca/ccxt support, and allow CCXT exchange-id aliases like coinbase/kraken (case/sep-insensitive).
• IBKR (crypto): normalize daily timestep handling (day/1d/1day) so crypto daily bars consistently use the derived-daily path.
• ThetaData: prevent acceptance backtests from hitting the downloader queue by enforcing CI-only warm-cache guardrails consistently (local runs behave like GitHub CI).
• ThetaData: treat session close as “complete coverage” for index minute OHLC to avoid perpetual STALE→REFRESH loops when backtest end dates are represented as midnight.
• Backtest cache (S3): speed up warm-cache hydration by streaming small objects via get_object instead of download_file transfer manager overhead.

v4.4.31

4.4.31 - 2026-01-09

Deploy marker: d5c6b730 ("deploy 4.4.31")

Changed

• Acceptance backtests now run in CI (no longer marked apitest) and CI caps were raised for the long full-year strategies to account for runner variability.
• Backtests running under pytest no longer auto-open HTML artifacts (plots/tearsheets) in a browser.

Fixed

• ThetaData: reduced option chain fanout (chain cache reuse under expiry hints; strike-list prefetch limited to head+tail expirations when unconstrained).
• Backtesting progress: improved per-asset download progress tracking in download_status so UI diagnostics are clearer.

Removed

• ⚠️ Removed ThetaData chain default-horizon env vars (THETADATA_CHAIN_DEFAULT_MAX_DAYS_OUT*). Chain default horizons are now fixed and covered by tests.

v4.4.29

Deploy marker: b8c6a83 (deploy 4.4.29)\n\nFixed\n- Prevent production backtests from OOM-like hard exits (ERROR_CODE_CRASH) when refreshing multi-year intraday ThetaData caches by avoiding deep copies during cache load/write and trimming non-option intraday frames in-memory.