OptionaldescriptionWhat you last asked for. state can lag this briefly after DELETE.
OptionalgateThe sandbox trial's promotion verdict. Absent for the whole 24-hour trial and present once it ends, so an absent gate means the trial has not finished. A run that started in LIVE without a trial has it from the start: the verdict of the earlier run it relies on, with inheritedFrom naming that run. passed is the verdict; the rest is diagnostic detail whose shape is not yet stabilized as public API: treat it as opaque.
OptionalnameOptionalpaperThe run's paper trading configuration as accepted at start, normalised. Absent when the run has no paper trading.
Increments on every accepted PUT .../params call, including one that resends the current values.
OptionalreasonWhy the run stopped or failed, when there is something to say; absent otherwise. It is never a
stack trace or an internal message. Either resource: ... (the platform stopped the run for
exceeding its resource allowance; the text says which limit) or one of a fixed set of sentences for
a FAILED run: the strategy cannot consume the source type the run was started with, the run's
definition was refused, the run could not start after several attempts, the strategy failed while
processing data, the run lost its data feed, or the generic The run failed.. The set may grow:
read an unrecognised sentence as a failure and do not parse it. A FAILED run usually stays
desired: RUNNING until you stop it, and counts as active (409 on a new start, and toward your
live-run limit) until then; one that can never run because its strategy cannot consume its source
type is stopped by the platform itself (desired: STOPPED), so it holds no place.
Whether this run's signals are relayed over the WebSocket channel described in the "Live execution" guide. It is the value requested at start, in either stage.
This run's own id — its canonical identity for PATCH/PUT .../params and for GET /live/public.
A new run starts SANDBOX, a 24-hour trial in which it is compared against a second execution and checked for resource use and stability. A run that passes moves to LIVE automatically when the 24 hours are up. A run of a strategy that an earlier run of yours already took through the trial starts LIVE, unless you started it with sandbox: true.
Epoch milliseconds.
The run's health right now: STARTING (no runner has reported on it yet), RUNNING, LAGGING (behind the market data, usually while catching up; clears by itself), HUNG (stuck inside one strategy call for longer than allowed; clears when it returns), DEGRADED (its independent executions produced different signals), FAILED (refused, could not start or failed while running; reason says why) or STOPPED. LAGGING, HUNG and DEGRADED come and go on a running run. The set may grow: read an unknown value as a running run with something to look at. See the Live execution guide.
OptionalstatsThe run's latest counters. Returned by the two reads, GET /strategy/{strategyId}/live and
GET /live/{runId}; never by starting or stopping a run. Absent until the first snapshot exists
(a run that has just started), and absent is not zero.
A live run's full state, as returned by stopping it through its strategy, and the base of the representations returned by starting and by reading it.
warmFromis declared by those two (LiveRunWithStreamandLiveRunDetail) and not here, because the response to stopping a run does not carry it.