factor_agent.ledger.TrialRecord#

class TrialRecord(round, candidate, status, stage, reasons, fingerprint=None, metrics=None, stats=None, attempts=<factory>, direction=None, graph=None, confirmation=None)[source]#

Bases: object

One candidate’s outcome.

status is "accepted", "rejected" (refused by the reviewer, unevaluated, or evaluated and refused by admission) or "failed" (never evaluated: refused by the parser or the Gate after every retry or, for the rewrite of a candidate the reviewer sent back, once, or its graph did not compile). Only an evaluated trial, one with metrics, counts toward the evaluated-trial count. stage is where it ended: "parse", "gate", "review", "evaluate" or "admission". candidate is the last version the proposer wrote, attempts every refused version before it (item, stage, reasons), the one the reviewer sent back to be rewritten with stage "review" and the reviewer’s reason. direction is the planned direction the candidate was written for (mechanism_type, action, parent, direction). stats holds the Gate’s statistics, graph the graph it built (Graph.to_json) and metrics the evaluation, once there are any; confirmation the evaluation on the confirmation segment, which no prompt shows, when the campaign has one.

Attributes#

TrialRecord.confirmation = None#

TrialRecord.direction = None#

TrialRecord.evaluated#: Whether the Evaluator scored the candidate: it counts toward the evaluated-trial count.

TrialRecord.fingerprint = None#

TrialRecord.graph = None#

TrialRecord.metrics = None#

TrialRecord.stats = None#

TrialRecord.round#

TrialRecord.candidate#

TrialRecord.status#

TrialRecord.stage#

TrialRecord.reasons#

TrialRecord.attempts#