Temporal Action Intelligence

TAI observes operator actions, segments them into sessions, runs PrefixSpan sequence mining, and synthesizes automation rules — alerts, state machines, and control agents — with operator-in-the-loop feedback.

Pipeline

graph LR S["Stream<br/><small>Capture actions</small>"] --> SEG["Segment<br/><small>5-min gap</small>"] SEG --> M["Mine<br/><small>PrefixSpan</small>"] M --> SC["Score<br/><small>Confidence</small>"] SC --> SYN["Synthesize<br/><small>Alert / SM / Agent</small>"] SYN --> FB["Feedback<br/><small>Approve / Reject</small>"] FB -.->|refine| SC

Action Capture

ActionType enum, UiInteractionKind, collector, tai_bridge.

Pattern Mining

PrefixSpan config, Sequence/Reaction/Cascade pattern types.

Automation Synthesis

Reactions → Alerts, Sequences → State Machines, Cascades → Agents.

Feedback & Approval

Confidence scoring, lifecycle, permanent exclusion after 3 rejections.

Claude Tools

6 tool schemas for LLM agents + ClickHouse persistence tables.

Key Types

TypeCratePurpose
ActionEventnerve-taiUnified action capture unit
MinedPatternnerve-taiSequence / Reaction / Cascade
Patternnerve-taiValidated pattern with lifecycle metadata
AutomationDraftnerve-taiGenerated config for operator review
PatternRegistrynerve-taiIn-memory pattern store with dirty tracking

Questions?

Reach out for help with integration, deployment, or custom domain codecs.

Get Started Request a Demo