Data Flow

How data moves through the Intentra CLI system.

Event Lifecycle

1. Hook Trigger

AI tool detects action completion and fires hook:

{
  "event": "ai_action_complete",
  "tool": "cursor",
  "action": "chat",
  "tokens": 1250
}

2. Event Normalization

CLI receives hook event and normalizes it:

• Validates event format
• Maps tool-specific fields to unified schema
• Adds timestamp and metadata

Normalized event:

{
  "id": "evt_abc123",
  "source": "cursor",
  "action": "chat",
  "tokens": {
    "input": 500,
    "output": 750
  },
  "timestamp": "2024-01-15T10:30:00Z"
}

3. Local Storage

Events stored in SQLite buffer:

INSERT INTO events (id, source, action, tokens_in, tokens_out, timestamp)
VALUES ('evt_abc123', 'cursor', 'chat', 500, 750, '2024-01-15T10:30:00Z');

4. Scan Aggregation

Periodic aggregation groups events into scans:

• Events within 30 seconds = one scan
• Detects patterns (retry loops, etc.)
• Calculates totals and estimates

5. Optional Sync

If sync is enabled, scans are queued and sent:

POST /v1/scans
Authorization: Bearer <token>

Latency Budget

The hook never blocks the IDE - events are captured asynchronously.

Offline Operation

All data is stored locally first:

• Works without internet connection
• Sync queue persists during offline periods
• Automatic retry when connection restored