Added#
An outcome on the usage record — the counterpart every figure here has been
missing. Chapter two of docs/plan-1.51.md. Everything Trazum reports is a
cost: it can say a workload got forty per cent cheaper and cannot say whether it
stopped working. The missing field is not something this tool can compute; it is
something only the caller knows.
Recorded beside label and session, read from outcome or trazum_outcome,
and declared in the config:
{ "outcomes": { "values": ["resolved", "escalated", "abandoned"], "success": ["resolved"] } }The rate is by spend, never by call. In the worked example in the README, forty of fifty-five calls succeeded — 73% by call and 48.2% by spend, because the expensive half of the traffic was the half that failed. The two figures diverge exactly when it matters, and a call-weighted rate would have read as a healthy product.
outcomes.success is required, and may be empty. Which of your words counts
as success is a judgement about your product rather than your bill, and this tool
has no standing to make it — one that decided escalated was a failure would be
wrong at every company where escalation is the correct, designed outcome for a
class of request. Leaving the field optional would send that question straight
back here to be answered by a default nobody chose. [] is allowed: a product
that records only failures has declared something real, and the report then says
it cannot state a rate rather than inventing one.
Never inferred. No absence of complaint counts as success, no short conversation counts as resolution, no retry counts as failure. Each is a plausible heuristic that would become a metric somebody optimises against — which is how a tool ends up rewarding conversations that ended early because the user gave up.
Nothing recorded is not a rate of zero. A rate of zero is a real and terrible
measurement; "nobody told us" is a different sentence and noRate spells it
differently. A tool that reported 0% success for an uninstrumented product would
get somebody fired over a number that measured nothing.
An undeclared value is named, not bucketed. A misspelled resolvd is a
broken exporter, and folding it into the failure side would report a product
regression that never happened — the direction that gets somebody paged at four
in the morning. It is excluded from both halves of the rate and listed by name.
Every printed rate carries what it does not cover. The share of the bill that carried no outcome is stated beside it, because a rate over an eighth of the spend is a rate about an eighth of the spend.
outcomeTally joins the profile document (measurement only, no judgement — an
aggregate, never a list of calls) and fieldCoverage gains outcome.
Security#
Three guards that an outcome is recorded and never inferred, each proven by planting the violation:
- the outcome module may not mention
session,stop_reason,truncated, a timestamp, a retry or a repeat; - the parser's assignment is compared exactly, so no
?? 'resolved'fallback can be appended; - the rate's type must stay
number | nullwith anoRatebeside it.
The privacy line does not move: an outcome is a small enumerated value, tallied as an aggregate, never alongside content.
Fixed#
A profile assertion was bounded by a phrase rather than by its subject. The
cache-TTL test asserted the whole report contained no cannot say whether, and
the new outcome-coverage line says "cannot say whether it stopped working" for
entirely unrelated and correct reasons — so a true assertion started failing on a
sentence about a different subject. It matches the TTL hedge's own sentences now.
Sixth occurrence of this shape in the repository, and the second this week.