Athena Event Types and Data Sources

User Stories

“I want all my conversations with Claude Code to be searchable in future conversations and I want to be able to extract knowledge from it and then visualize and query that later”

This drives the need to capture:

• Chat logs (JSONL files from Claude conversations)
• System prompts and responses
• Code changes and git commits from those conversations
• File modifications made during conversations

“I want to understand patterns in how I work - when I’m most productive, what triggers deep work sessions, how my environment affects my focus”

This drives the need to capture:

• Browser bookmarks changes
• Browsing history (if enabled)
• Application usage patterns
• File access patterns
• Home Assistant device states (lighting, temperature, etc.)

“I want the system to learn from everything happening in my physical space and make intelligent decisions about displays, alerts, and automation”

This drives the need to capture:

• Motion detection events
• Person/animal/vehicle detection
• PTZ camera position changes
• Screenshot captures and video recordings
• Sound level changes and audio classification
• Microphone and speaker activity

“I want a complete audit trail of how my home automation responds to events, so I can improve the system and troubleshoot issues”

This drives the need to capture:

• MQTT messages from IoT devices
• Device state changes (lights, sensors, etc.)
• Automation triggers and responses
• Device heartbeats and connectivity changes
• System resource usage and error events

Event Sources

Based on these user stories, we capture data from:

• Chat logs and voice interactions
• Browser and application activity
• File system modifications
• Home automation devices and sensors
• Cameras and audio equipment
• Network and infrastructure systems
• Media files (images, video, audio, documents)

Event Schema Structure

{
  "timestamp": "2025-06-08T16:30:00Z",
  "source_type": "logged-conversation",
  "source_id": "claude-code-session-2025-06-08",
  "data": {
    "file_path": "/path/to/chat.jsonl",
    "size_bytes": 1024,
    "message_count": 15
  },
  "metadata": {
    "confidence": 1.0,
    "processing_time_ms": 5,
    "tags": ["ai", "collaboration", "physics-of-work"]
  }
}

Example source_types:

• captured-still, detected-motion, recorded-video
• logged-conversation, sent-message, updated-prompt
• modified-file, changed-bookmark, accessed-document
• triggered-automation, changed-device-state, sent-mqtt

Storage Targets

Primary Storage

• SQLite database (data/events.db) - structured queries
• File queue (data/queue/) - processing backlog

Future Storage Options

• S3 bucket - long-term archive
• Neo4j graph - relationship analysis
• Time series DB - temporal patterns

Ingestion Pipeline

1. File watchers monitor directories
2. Event extractors parse different formats
3. Event validator ensures schema compliance
4. Storage engine persists to database/queue
5. AI processors analyze and respond

---

Comprehensive event taxonomy for Athena distributed AI system