Read staging and live
- Read staging with
GET /api/v2/agentEventRules/staging?targetAccountId=.... - Read live with
GET /api/v2/agentEventRules/default?targetAccountId=....
id as the agentConfigId where a configuration-specific operation requires one.
Update staging
Send writable Agent settings toPOST /api/v2/agent?targetAccountId=.... Omitted settings retain their current values.
Arrays such as actions, apps, skills, and reportingFields represent the complete authored array when supplied. Read before replacing them. Do not send calculated read-only values such as function eligibility, active slots, or effective reporting fields back as authored configuration.
After an update, read staging again and compare the settings you intended to change. Test the staged Agent before publishing.
Publish
Publish withPOST /api/v2/agent/publish?targetAccountId=.... A successful response means the publish request completed; follow it with an independent live read and verify the expected values.
Versions
Saved versions are stored under the liveagentConfigId. A version can snapshot either staging or live. Restoring defaults to staging, so a restored staging version still needs to be tested and published.
Use the Agent version endpoints to list, create, replace, rename, restore, and delete saved versions. Use numeric versionId values returned by those endpoints.
Experiments
Experiments distribute new conversations deterministically across two or more saved versions. Use the liveagentConfigId and saved numeric versionId values. Only one experiment for an Agent configuration can be active at a time, and variants can be edited only while the experiment is a draft.
Do not use an account ID in place of agentConfigId.