ALP Specification — Event Sourcing & Deterministic Replay

Version: 80.0.0
Status: Stable

---

1. Event Sourcing Overview

flowchart TD
    EventLog[events.jsonl] --> Event1[Event 1]
    EventLog --> Event2[Event 2]
    EventLog --> Event3[Event 3]
    Event1 --> Type1[status_changed]
    Event2 --> Type2[object_created]
    Event3 --> Type3[task_claimed]
    Engine[EventEngine] --> Replay[replay]
    Engine --> Filter[filter by type/id/time]
    Replay --> State[Reconstructed State]
    Filter --> Events[Filtered Events]

2. Overview

ALP v10.1.0 introduces Event Sourcing: an append-only, schema-versioned event log that records every workspace mutation. Instead of overwriting state, ALP records the sequence of events that led to the current state. This enables deterministic replay, incident forensics, and auditability without external dependencies.

The event log lives at .alp/.events/events.jsonl. Each line is a self-contained JSON object with a unique id, ISO timestamp, event type, payload, ALP version, and schema version. Events are never mutated or deleted.

---

2. The Event Log

2.1 Location

.alp/.events/events.jsonl

2.2 Event Schema

{
  "id": "abc123-...",
  "timestamp": "2026-07-20T12:00:00Z",
  "type": "status_changed",
  "payload": {
    "object_id": "task-login-ui",
    "old_value": "[ ]",
    "new_value": "[x]"
  },
  "version": "10.1.0",
  "schemaVersion": 1
}

Field	Type	Description
id	String	Unique event identifier
timestamp	ISO 8601	When the event occurred
type	EventType	The kind of mutation
payload	Object	Arbitrary structured data describing the event
version	String	ALP version that produced the event
schemaVersion	Number	Schema version for forward-compatibility

---

3. Event Store API

3.1 TypeScript

import { EventStore } from '@autonomous-lifecycle-protocol-alp/parser';

const store = new EventStore('.alp');

// Append a new event
store.append('status_changed', { object_id: 'task-1', old: '[ ]', new: '[x]' });

// Read all events (in order)
const events = store.readAll();

// Filter and replay
const result = store.replay({
  from: '2026-07-20T00:00:00Z',
  to: '2026-07-20T23:59:59Z',
  types: ['status_changed', 'object_created'],
  objectId: 'task-1'
});

3.2 Python

from alp_sdk import EventStore

store = EventStore('.alp')
store.append('status_changed', {'object_id': 'task-1', 'old': '[ ]', 'new': '[x]'})
events = store.read_all()