- New: Bound instrumentation data
- New: Experimental behavior scoring
- New: Interaction detection. Count clicks and key presses that a script dispatched.
- New: AI agent detection for dozens of agent browsers
- New:
analyzeBehavior,BEHAVIOR_WEIGHTS,DEFAULT_BEHAVIOR_THRESHOLDandWEAK_CAPare exported for servers that score traces themselves. - Fixed: Block malformed data from a client instead of throwing
- Fixed: Block all failure markers by default
Upgrading: Custom instrumentationGenerator functions should return the whole result of generateInstrumentation, including binding. One that only copies expectedVals now makes generateChallenge throw, so it fails when the challenge is issued instead of on every redeem. See the README.