Skip to content

A run with its decision receipt

GET
/api/v1/runs/{id}
curl --request GET \
--url https://example.com/api/v1/runs/2489E9AD-2EE2-8E00-8EC9-32D5F69181C0

The run and the receipt that records how it was reached.

id
required
Id
string format: uuid

Successful Response

Media typeapplication/json
AnalysisRun

A run together with its decision receipt.

object
band
Any of:
Band

Backend-assigned sharpness band of the model output.

string
Allowed values: confident unsure cannot_tell
config_version
required
Config Version
string
created_at
required
Created At
string format: date-time
decider_model
Any of:
string
failure_code
Any of:
string
id
required
Id
string format: uuid
lane
required
Lane

Suggested-disposition bucket used to order the queue (nothing is hidden).

string
Allowed values: looks_consistent review_suggested needs_local_knowledge insufficient_evidence rules_only
latency_ms
Latency Ms
integer
0
nature
Any of:
Nature

Q1 option set: what the changeset looks like.

string
Allowed values: consistent_edit likely_good_faith_mistake possible_deliberate_damage needs_local_knowledge
nature_probabilities
Nature Probabilities
object
key
additional properties
number
<= 1
order_sensitivity
Order Sensitivity
number
0
panel
Any of:
PanelView

The fact panel as reviewers see it: every probe reading, the table rows and cross-checks.

object
advisory_nature
required
PanelNature

The four-way reading the model also gives; shown for audit and never used for lanes.

object
advisory
Advisory
boolean
default: true
choice
required
Nature

Q1 option set: what the changeset looks like.

string
Allowed values: consistent_edit likely_good_faith_mistake possible_deliberate_damage needs_local_knowledge
probabilities
required
Probabilities

Every option with its probability; no single figure is shown

object
key
additional properties
number
<= 1
composition
Any of:
PanelCompositionView

Which table rows produced the suggested response and reading.

object
nature_row
required
Nature Row
string
response_row
required
Response Row
string
unknown
Unknown

Probe ids read as unknown

Array<string>
yes
Yes

Probe ids read as yes

Array<string>
cross_checks
Cross Checks

Panel reading compared with the value computed from parsed evidence

object
key
additional properties
string
Allowed values: agree disagree undecided n/a
high
required
High
number
<= 1
low
required
Low
number
<= 1
probes
required
Probes
Array<object>
PanelProbe

One yes/no fact the Decider was asked, with its probability and a three-way reading.

object
id
required
Id
string
label
required
Label

Plain-language statement of the fact that was asked

string
probability
required
Probability
number
<= 1
reading
required
Tri

No below the low threshold, yes above the high, else unknown

string
Allowed values: no unknown yes
use
required
Use

Composed: feeds the suggestion; cross_check_only and advisory are shown only

string
Allowed values: composed cross_check_only advisory
yes_means
required
Yes Means

Whether a yes raises or lowers the level of review suggested

string
Allowed values: concern reassurance
receipt
required
DecisionReceipt

Everything needed to audit or replay one decision. Immutable; corrections are new rows.

object
aggregated
Aggregated
object
key
additional properties
AggregatedChoice

Mean over permutations of one question: probabilities, argmax and sharpness statistics.

object
argmax
required
Argmax
string
confidence_agg
required
Confidence Agg
number
<= 1
order_sensitivity
required
Order Sensitivity
number
probabilities
required
Probabilities
object
key
additional properties
number
<= 1
band
Any of:
Band

Backend-assigned sharpness band of the model output.

string
Allowed values: confident unsure cannot_tell
case_id
Any of:
string
config_version
required
Config Version
string
coverage
required
Coverage
object
key
additional properties
any
decider
required
Any of:
DeciderIdentity

What GET /health said about the model that produced the answers.

object
base_model
Any of:
string
checkpoint
required
Checkpoint
string
device
required
Device
string
max_length
required
Max Length
integer
model
required
Model
string
prefix_cache
Any of:
boolean
server_git_sha
Any of:
string
temperature
Any of:
number
errors
Errors
Array<string>
evidence_version
required
Evidence Version
string
human_outcome_ids
Human Outcome Ids
Array<string>
lane
Any of:
Lane

Suggested-disposition bucket used to order the queue (nothing is hidden).

string
Allowed values: looks_consistent review_suggested needs_local_knowledge insufficient_evidence rules_only
nouls
Nouls
object
key
additional properties
number
<= 1
panel
Any of:
FactPanel

Validated panel answers: readings for every probe plus the advisory nature distribution.

object
advisory_nature
required
AdvisoryNature

The single-order nature choice; stored for audit and evaluation, never used for lanes.

object
advisory
Advisory
boolean
default: true
choice
required
Choice
string
confidence
required
Confidence
number
<= 1
order
required
Order
Array<string>
probabilities
required
Probabilities
object
key
additional properties
number
<= 1
high
required
High
number
<= 1
ids
Ids
Array<string>
low
required
Low
number
<= 1
readings
required
Readings
object
key
additional properties
NoulReading

One panel probe: the model probability and its tri-state reading.

object
id
required
Id
string
probability
required
Probability
number
<= 1
state
required
Tri

A noul probability read against the panel thresholds.

string
Allowed values: no unknown yes
panel_composition
Any of:
object
key
additional properties
any
panel_cross_checks
Any of:
object
key
additional properties
string
panel_questions_hash
Any of:
string
permutations
Permutations
Array<object>
PermutationResult

One Decider choice answer for one option order (e.g. nature_p0).

object
confidence
required
Confidence
number
<= 1
name
required
Name
string
order
required
Order
Array<string>
probabilities
required
Probabilities
object
key
additional properties
number
<= 1
questions_hash
required
Any of:
string
receipt_version
Receipt Version
string
default: 1
reconciliation
Any of:
ReconciliationRecord

Q2 reconciliation against Q1; serialised with the key from.

object
applied
required
Applied
boolean
from
Any of:
string
to
Any of:
string
rules_fired
Rules Fired
Array<object>
RuleHit

A deterministic policy rule that fired, with the parsed evidence that triggered it.

object
code
required
Code
string
label
required
Label
string
severity
required
Severity

Severity of a policy rule hit.

string
Allowed values: info warn block
triggered_by
Triggered By
object
key
additional properties
any
snapshot_id
Any of:
string
state_tokens
State Tokens
integer
0
timing_ms
Timing Ms
object
key
additional properties
integer
reconciled
Reconciled
boolean
response
Any of:
Response

Q2 option set: suggested next step for the reviewer.

string
Allowed values: no_action_suggested welcome_and_tips ask_source_question request_local_check escalate_to_experienced_reviewer
response_probabilities
Response Probabilities
object
key
additional properties
number
<= 1
rules_fired
Rules Fired
Array<object>
RuleHit

A deterministic policy rule that fired, with the parsed evidence that triggered it.

object
code
required
Code
string
label
required
Label
string
severity
required
Severity

Severity of a policy rule hit.

string
Allowed values: info warn block
triggered_by
Triggered By
object
key
additional properties
any
snapshot_id
required
Snapshot Id
string format: uuid
status
required
RunStatus

Outcome of an analysis run.

string
Allowed values: succeeded failed

Example

{
"band": "confident",
"lane": "looks_consistent",
"latency_ms": 0,
"nature": "consistent_edit",
"order_sensitivity": 0,
"panel": {
"advisory_nature": {
"advisory": true,
"choice": "consistent_edit"
},
"cross_checks": {
"additionalProperty": "agree"
},
"probes": [
{
"reading": "no",
"use": "composed",
"yes_means": "concern"
}
]
},
"receipt": {
"band": "confident",
"lane": "looks_consistent",
"panel": {
"advisory_nature": {
"advisory": true
},
"readings": {
"additionalProperty": {
"state": "no"
}
}
},
"receipt_version": "1",
"rules_fired": [
{
"severity": "info"
}
],
"state_tokens": 0
},
"reconciled": false,
"response": "no_action_suggested",
"rules_fired": [
{
"severity": "info"
}
],
"status": "succeeded"
}

Not Found

Problem

Application/problem+json body. Extension members (for example job_id) are allowed.

object
detail
Any of:
string
errors
Any of:
Array<object>
FieldError

One validation failure: where, what, and a stable machine code (never the input value).

object
loc
required
Loc
Array
msg
required
Msg
string
type
required
Type
string
instance
Any of:
string
request_id
required
Request Id
string
status
required
Status
integer
title
required
Title
string
type
Type
string
default: about:blank
key
additional properties
any

Example

{
"type": "about:blank"
}

Validation Error

Media typeapplication/json
HTTPValidationError
object
detail
Detail
Array<object>
ValidationError
object
ctx
Context
object
input
Input
loc
required
Location
Array
msg
required
Message
string
type
required
Error Type
string

Example generated

{
"detail": [
{
"ctx": {},
"input": "example",
"loc": [
"example"
],
"msg": "example",
"type": "example"
}
]
}

Map data © OpenStreetMap contributors, under the Open Database Licence. Code under the Apache-2.0 licence. Made by Yogesh.

Changeset Second Opinion is an independent open source tool for OpenStreetMap reviewers. It is not endorsed by, sponsored by or affiliated with the OpenStreetMap Foundation. OpenStreetMap is a trademark of the OpenStreetMap Foundation, used under itstrademark policy.