Iterate in the research REPL (qkt research)

Interactive playback of a strategy over a historical window. Load once, then step / run / seek through a tick replay and watch trades, positions, and equity update — without re-running a full backtest on every edit.

Start a session

qkt research strategy.qkt --from 2024-01-01 --to 2024-02-01 --data-root ./data

Optional: --starting-balance 10000. The data window is read from the local store under --data-root (the same store qkt fetch / qkt backtest use).

Commands

Command	Effect
run	Advance to the end of the window.
step N	Advance N bars on the strategy's primary timeframe.
step 1d / step 30m / step 2h	Advance a wall-clock duration.
run-to 2024-01-15	Advance to a timestamp (a past time resets and runs forward).
run-to next-trade	Advance to the next fill.
reset	Restart from the first tick, same strategy.
reload	Re-read + recompile the file (after you edit it), then reset.
show	Print the current footer without advancing.
quit	Exit.

The tweak loop

Edit the .qkt in your editor, then type reload. A parse error keeps the previous strategy loaded and prints the diagnostics, so a bad edit never drops your session.

Determinism

A full run is bit-identical to qkt backtest over the same window — the REPL and the batch backtest share one replay engine; stepping only changes when the replay pauses.

Not yet supported

Per-bar rule introspection, in-session parameter overrides (set), portfolio files, and backward seek are tracked as follow-ups under the #81 epic.