API Changelog
Track changes to the NotAI Public REST API across versions.
Minor fixes and improvements
Minor fixes and improvements
Documentation and site updates (no API changes).
Minor fixes and improvements
Changes
- Security scheme
bearer_authbearer format updated:Not AI API key (aik_v1_*)→NotAI API key (aik_v1_*) - API description text updated
- The spec names the product NotAI:
info.titleis "NotAI Public REST API", and the description, the contact name and the API key's bearer format ("NotAI API key (aik_v1_*)") follow. No route, field, status code or behavior changed. - A test delivery (
POST /v1/webhooks/{subscriptionId}/test) now carriesdata.message"This is a test webhook delivery from the NotAI Public REST API." Recognise a test delivery bydata.testbeingtrue, never by the text of its message.
Changes
- Added new optional field:
remoteMcp
Minor fixes and improvements
Breaking
- Changed
botAtminimum: 0 → 50
Other Changes
- Added response code
Minor fixes and improvements
Changes
- New response on every API-key route:
503with the error codeAUTH_UNAVAILABLEand aRetry-Afterheader. It means the key check itself could not complete, for example during a brief storage outage. It says nothing about whether the key is valid, so do not rotate or discard the key. Wait the number of seconds inRetry-Afterand send the same request again: the request was refused before it ran, so the retry is safe on every route. See Error Handling and Authentication. - Every
401(MISSING_API_KEY,INVALID_API_KEY) is now the answer of a completed key check. A failed check used to answer401too, which looked like a revoked key.
Documentation and site updates (no API changes).
Changes
- Parameter
cursordescription updated on GET /v1/users
Minor fixes and improvements
Minor fixes and improvements
Changes
- Parameter
cursordescription updated on GET /v1/users/{userId}/sessions and 2 other endpoints
Changes
- Added new optional field:
emptySessions - Added new optional field:
emptySessionsToday GET /v1/stats:totalSessionsandsessionsTodayleave out sessions that closed without recording any activity, and two new fields count them:emptySessionsandemptySessionsToday.totalSessions+emptySessionsis every session, the valuetotalSessionshad before.GET /v1/analyticsalready left these sessions out.GET /v1/stats:botsDetectednow counts sessions at or above the platform's fixed bot flag (botScore0.70 or higher), the rule behind the session statusflagged. This is a correction to match the documented 0.70 rule: it counted from 0.50, so the number can be lower. It still does not follow your risk thresholds.GET /v1/usage?metric=...now returns the metered rows for that metric. A filtered request returned an empty list; unfiltered requests are unchanged.GET /v1/statsaverageBotScoreandGET /v1/usersaverageBotScoreandmaxBotScorenever round up onto 0.40 or 0.70 when the score is below it: 0.6999 reads 0.69.averageBotScoreinGET /v1/statsaverages numeric scores only. When that average sits on 0.40 or 0.70, the scores are read again (up to 200,000) so the exact mean decides: 0.4, 0.71 and 0.99 still read 0.70. If that second read fails or there are more scores, the first average stands, and an exact 0.70 can read 0.69.GET /v1/users: a stored score ofNaNno longer fails the page. It is skipped: the user's average and peak come from the other scores, or read 0 when there are none.
Changes
- Added response code
- Parameter
cursordescription updated on GET /v1/users/{userId}/sessions and 2 other endpoints - Parameter
statusdescription updated on GET /v1/sessions - Parameter
botOnlydescription updated on GET /v1/sessions GET /v1/analyticsandGET /v1/usersnow follow your integration's risk thresholds, set withPUT /v1/settings/risk-thresholds(defaults:mediumAt40,botAt70). An integration that never set them sees no change. Each response uses the thresholds read for it: normally the saved values; if the primary read fails, a regional copy can briefly still return the previous values, and if no read succeeds in time the response uses the defaults and still succeeds.GET /v1/analytics:summary.uniqueBots,timeSeries[].botCountandtopOrigins[].botPercentagecount sessions whosebotScoreis at or abovebotAt. ThebotDetectionbands follow the same thresholds:highConfidenceBotsstarts at the higher ofbotAtand 0.90,mediumConfidenceBotsruns frombotAtup to that,lowConfidenceBotsfrommediumAtup tobotAt, andlikelyHumansis belowmediumAtor unscored.GET /v1/users:riskLevel(high at or abovebotAt, medium frommediumAtup tobotAt, low belowmediumAtor unscored),botSessionCountandmediumSessionCountuse the same thresholds on each user's raw sessionbotScore.- Unchanged, still on the platform's fixed bot flag (
botScore0.70 or higher): thebotOnlyfilter and the session statusflaggedofGET /v1/sessions, andBotDetectedwebhooks. - Unchanged:
botsDetectedinGET /v1/statskeeps its own fixed platform rule and does not follow the risk thresholds.
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Documentation and site updates (no API changes).
Documentation and site updates (no API changes).
Changes
- Added new optional field:
subsystems - Added new optional field:
edgeAuthSnapshotAgeSeconds - New schema
healthSubsystemsadded
Documentation and site updates (no API changes).
Documentation and site updates (no API changes).
Changes
- Fixed: webhook threshold subscriptions now work end to end. Previously, threshold conditions (
pass/blockscore cutpoints) could be stored corrupted, causing threshold-filtered deliveries to fail and threshold subscriptions to error on read. - Spec correction: nullability is now annotated accurately across request and response schemas (response fields that can return
null, and optional request fields such asnotesand session label properties, are markednullable). - Schema documentation refactor: existing schema references (webhook threshold, subscription tier info) are wrapped in allOf so nullability can be attached. Wire shapes are unchanged; the automated differential showed the wrapping as field removals and additions.
Changes
- Webhook event payload schemas are now fully documented:
botDetectedPayload,sessionAnomalyPayload,thresholdExceededPayload,alertTriggeredPayload,reportReadyPayload,sessionCorrelatedPayload, andwritingSessionScoredPayload. - API key header authentication (
x-api-key) is now documented as an accepted alternative toAuthorization: Beareron every endpoint. Both methods have always been accepted; earlier specs documented only Bearer. - Spec correction: webhook
eventTypeandlastDeliveryStatusare documented as their string names (for examplebotDetected,delivered). The API has always sent strings; earlier specs misdocumented these fields as integers. - Spec correction: required/optional flags and validation constraints now match server behavior exactly across request and response schemas (webhook
urlandeventType, errorcode/message/details, session labelisBot, riskthresholds,noteslength cap, timestamp formats). No server-side validation changed. - Spec cleanup: removed the phantom
jsonElementschema and itsvalueKind/itemfields, a serializer artifact that never appeared on the wire and previously left webhook schemas unresolvable. - The allowed values of the
statusfilter onGET /v1/sessionsand the enum/default semantics of the session labelconfidencefield are now documented (both existed before; their constraints were unspecified). - API description text updated (the spec no longer claims the API is read-only; it declares its write operations).
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Minor fixes and improvements
Initial release of the public API documentation for this environment.