Skip to content

evidence.ingest

Append one or more evidence records to the tamper-evident evidence log for the caller’s tenant. Use this to write back what an agent did and observed (the trace that later derives lessons). Server computes tenant_id/hashes/id; do not supply them. actor_type/actor_id are optional — they default to your token’s principal; supply them only to attest a different actor (e.g. a CI bot). When the record reports an action preflight.action already judged, set outcome: {decision_id from that verdict, heeded (did you do what it advised — required, never guessed), subject_ref naming what you acted on}. That link is what makes governance measurable; describing it only in summary prose leaves it unreadable.

FieldTypeRequiredDescription
evidencearray (nullable)yes(items 1–500)
optionsobject
FieldTypeRequiredDescription
action_typeActionTypefrom the closed high-risk action taxonomy (schemas.md §1)
actor_idstring
actor_typeActorType
artifact_refsarray of string (nullable)
entity_mentionsarray of string (nullable)
event_tsstring (date-time)yes
labelsarray of string (nullable)
metadataobject
outcomeobject (nullable)
payload_refstring
risk_classRiskClass
scope_keystringyeslevel:id, e.g. repo:payments-service
sensitivitySensitivity
source_refstringyes
source_typeSourceTypePublicyes
summarystringyes
workspace_idstringoptional; defaults to your token’s workspace
FieldTypeRequired
decision_idstringyes
heededboolean (nullable)yes
subject_refstring
FieldTypeRequired
deriveboolean (nullable)
idempotency_keystring
FieldTypeRequired
acceptedarray (nullable)yes
chain_headobject (nullable)
rejectedarray (nullable)yes
request_idstring
FieldTypeRequired
chainstringyes
chain_statestringyes
dedupedbooleanyes
derivation_task_idstring
event_hashstringyes
idstringyes
prev_hashstring
sequencer_task_idstring
FieldTypeRequired
chainstringyes
head_hashstringyes
lengthintegeryes
FieldTypeRequired
codestringyes
indexintegeryes
messagestringyes