ProvenanceTracker - ProvenanceKit

Polls pk.sessionProvenance() and displays a live feed of recorded actions with a timeline UI. Ideal for chat apps and multi-step AI workflows where you want users to see provenance building in real time.

Preview
Code

import { ProvenanceTracker } from "@provenancekit/ui";

// Auto-polls by sessionId — shows new actions as they arrive
<ProvenanceTracker
  sessionId="research-session-a1b2c3"
  pollInterval={3000}
  onNewAction={(action) => {
    console.log("New action:", action.type, action.id);
  }}
/>

// Headless mode — pass session directly (no polling)
<ProvenanceTracker session={mySession} maxActions={10} />

What the preview shows

The tracker is loaded with a 6-step AI research workflow session:

Step	Type	Performer	Description
Upload sources	`create`	Sarah Kim (human)	Source documents uploaded
Synthesise	`transform`	Research Assistant (AI · Claude Opus 4.6)	Extracts key facts from sources
Draft summary	`create`	Research Assistant (AI · Claude Opus 4.6)	Writes the initial report draft
Generate chart	`create`	Research Assistant (AI · GPT-4o)	Produces a supporting data chart
Revise	`transform`	Sarah Kim (human)	Human edits and refinements
Finalise	`verify`	Research Assistant (AI · Claude Opus 4.6)	Combines and verifies the final report

Each action item shows:

Action type chip (create / transform / verify)
Latest badge on the most recent action
AI tool badge (violet) when ext:ai@1.0.0 is present
Verified badge (green) when ext:verification@1.0.0 status is "verified"
Relative timestamp

The session header shows live stats: total actions, resources, and entities recorded.

Props

Prop	Type	Default	Description
`sessionId`	`string`	—	Session ID to poll
`pollInterval`	`number`	`3000`	Polling interval in milliseconds
`session`	`SessionProvenance`	—	Headless mode — pass session directly
`maxActions`	`number`	`20`	Maximum number of actions to display
`onNewAction`	`(action: Action) => void`	—	Called when a new action is detected
`className`	`string`	—	Additional CSS class

Common pattern — chat apps

// Start a session when the conversation begins
const session = await pk.startSession({ label: "user-chat-123" });

// After each AI response, record an action
await pk.record({
  action: {
    type: "create",
    performedBy: "your-ai-entity-id",
    extensions: {
      "ext:ai@1.0.0": {
        provider: "anthropic",
        model: "claude-opus-4-6",
        autonomyLevel: "assistive",
        tokensUsed: response.usage.totalTokens,
      },
    },
  },
  sessionId: session.sessionId,
});

// Show the live feed in the UI
<ProvenanceTracker
  sessionId={session.sessionId}
  pollInterval={2000}
/>