bowline · evidence explorer · synthetic dataset
$ bowline export evidence --run-id fixture-run

Read a decision the way an operator would.

This page renders one exported evidence bundle, unmodified, straight from the JSON Bowline's export evidence command produces. Every identifier below is fictional — a fixture run, not a customer, a route, or a workload that ever touched real traffic.

Synthetic dataset — every run id, supply id, and workload identifier below is fictional. Nothing on this page is a real organization, route, app, model, or endpoint.

$ cat ./generated-from

Traceability.

Every number on this page traces back to a run id and a pair of digests — the policy and registry inputs Bowline hashed at export time. Reproduce the export, recompute the digests, and the two must match.

run id
fixture-run
feed version
fixture-v1
policy digest
sha256:a3c5fff3be2d…
registry digest
sha256:63a333765724…
attribution digest
sha256:d403a0eabbd0…
owned-cost digest
sha256:d1b85cd93569…
accepted
1
recorded
1
dropped
0
truncated
0
unmapped
0
unpriceable
0
segments
1
clean shutdown
yes
$ ls ./decisions

Allowlisted decisions.

Every decision Bowline recorded for this run — the actual supply that served the request, the supply Bowline would have allocated in shadow mode, and the policy exposure of the actual choice.

decision reftask classactual supplyshadow supplyactual costshadow costpolicy
fixture-run:1mechanicalsupply/actualsupply/selected$0.001000$0.000400compliant
$ cat ./disclosure

Complete disclosure.

Bowline discloses what it doesn't know as plainly as what it does. A bundle marked incomplete, or carrying integrity warnings or coverage gaps, says so here — never silently.

complete: true
integrity warnings: none
coverage gaps: none
confidence legend
observed directly observed from the recorded outcome
declared derived from operator-declared registry or owned-cost inputs
canary-verified supported by matching canary evidence
unverified modeled or not independently verified
$ cat ./coverage

Coverage gaps, named.

Coverage says how much of the traffic this run observed Bowline could actually classify — by protocol, by observation source, and per decision.

total records
1
supported
1
unsupported
0
complete
yes
$ cat ./aggregates

Sovereignty and counterfactual evidence.

Sovereignty and counterfactual figures are modeled evidence, not realized savings: what the recorded traffic actually cost, set against what Bowline would have allocated in shadow mode, and against an all-frontier baseline — each figure carrying its own confidence label.

sovereignty ratio (actual): 0.00observed
sovereignty ratio (shadow): 1.00declared
actual cost: $0.001000observed
shadow cost: $0.000060declared
all-frontier cost: $0.002000observed
modeled delta vs. all-frontier: $0.001940unverified
quality parity: not measured — no parity canary evidence for this reportunverified
task classrecordsactual costcheapest clearing rentedmodeled delta
mechanical1$0.001000observed$0.002000observed$-0.001000unverified

This is the whole bundle.

No hidden fields, no dashboard you have to trust — the same JSON is checked into the repository.