Task_Reporter/openspec/changes/archive/2025-11-17-add-realtime-messaging/specs/realtime-messaging/spec.md

realtime-messaging Specification

Purpose

Enable real-time bidirectional communication within incident rooms using WebSocket protocol, allowing production teams to collaborate instantly during incidents with message persistence for audit trail and report generation.

ADDED Requirements

Requirement: WebSocket Connection Management

The system SHALL provide WebSocket endpoints for establishing persistent bidirectional connections between clients and server, with automatic reconnection handling and connection state management.

Scenario: Establish WebSocket connection

• WHEN an authenticated user connects to ws://localhost:8000/ws/{room_id}
• THEN the system SHALL validate user's room membership
• AND establish a WebSocket connection
• AND add the connection to the room's active connections pool
• AND broadcast a "user joined" event to other room members

Scenario: Handle connection authentication

• WHEN a WebSocket connection request is made without valid authentication token
• THEN the system SHALL reject the connection with status 401
• AND close the WebSocket immediately

Scenario: Automatic reconnection

• WHEN a WebSocket connection is dropped unexpectedly
• THEN the client SHALL attempt to reconnect automatically with exponential backoff
• AND resume from the last received message sequence number
• AND request any missed messages during disconnection

Requirement: Real-time Message Broadcasting

The system SHALL broadcast messages to all active room members in real-time, ensuring message ordering and delivery acknowledgment.

Scenario: Send text message

• WHEN a room member sends a message via WebSocket:

  {
    "type": "message",
    "content": "Equipment temperature rising to 85°C",
    "message_type": "text"
  }
  

• THEN the system SHALL:
  • Validate user has write permission (OWNER or EDITOR role)
  • Assign a unique message_id and timestamp
  • Store the message in database
  • Broadcast to all active WebSocket connections in the room
  • Return acknowledgment to sender with message_id

Scenario: Send system notification

• WHEN a system event occurs (user joined, room status changed, etc.)
• THEN the system SHALL broadcast a system message:

  {
    "type": "system",
    "event": "user_joined",
    "user_id": "john.doe@panjit.com.tw",
    "timestamp": "2025-11-17T10:00:00Z"
  }
  

• AND all connected clients SHALL display the notification

Scenario: Handle message ordering

• WHEN multiple messages are sent simultaneously
• THEN the system SHALL ensure FIFO ordering using message sequence numbers
• AND clients SHALL display messages in the correct order

Requirement: Message Persistence and History

The system SHALL persist all messages to database for audit trail, report generation, and history retrieval.

Scenario: Store message in database

• WHEN a message is sent through WebSocket
• THEN the system SHALL create a database record with:
  • message_id (UUID)
  • room_id (FK to incident_rooms)
  • sender_id (user email)
  • content (text or JSON for structured messages)
  • message_type (text, image_ref, file_ref, system)
  • created_at timestamp
  • edited_at (nullable for message edits)
  • deleted_at (nullable for soft delete)
  • sequence_number (for ordering)

Scenario: Retrieve message history

• WHEN a user joins a room or reconnects
• THEN the system SHALL load recent messages via GET /api/rooms/{room_id}/messages?limit=50&before={timestamp}
• AND return messages in reverse chronological order
• AND include pagination metadata for loading more history

Scenario: Search messages

• WHEN a user searches for messages containing specific keywords
• THEN the system SHALL query the database with full-text search
• AND return matching messages with highlighted search terms
• AND maintain user's access control (only rooms they're members of)

Requirement: Message Types and Formatting

The system SHALL support various message types including text, image references, file references, and structured data for production incidents.

Scenario: Text message with mentions

• WHEN a user sends a message with @mentions

  {
    "content": "@maintenance_team Please check Line 3 immediately",
    "mentions": ["maintenance_team@panjit.com.tw"]
  }
  

• THEN the system SHALL parse and store mentions
• AND potentially trigger notifications to mentioned users

Scenario: Image reference message

• WHEN a user uploads an image and sends reference

  {
    "type": "message",
    "message_type": "image_ref",
    "content": "Defect found on product",
    "file_id": "550e8400-e29b-41d4-a716-446655440000",
    "file_url": "http://localhost:9000/bucket/room-123/image.jpg"
  }
  

• THEN the system SHALL store the file reference
• AND clients SHALL display image preview inline

Scenario: Structured incident data

• WHEN reporting specific incident metrics

  {
    "type": "message",
    "message_type": "incident_data",
    "content": {
      "temperature": 85,
      "pressure": 120,
      "production_rate": 450,
      "timestamp": "2025-11-17T10:15:00Z"
    }
  }
  

• THEN the system SHALL store structured data as JSON
• AND enable querying/filtering by specific fields later

Requirement: Connection State Management

The system SHALL track online presence and typing indicators for better collaboration experience.

Scenario: Track online users

• WHEN users connect/disconnect from a room
• THEN the system SHALL maintain a list of online users
• AND broadcast presence updates to all room members
• AND display online status indicators in UI

Scenario: Typing indicators

• WHEN a user starts typing a message
• THEN the client SHALL send a "typing" event via WebSocket
• AND the system SHALL broadcast to other room members
• AND automatically clear typing status after 3 seconds of inactivity

Scenario: Connection health monitoring

• WHEN a WebSocket connection is established
• THEN the system SHALL send ping frames every 30 seconds
• AND expect pong responses within 10 seconds
• AND terminate connection if no response received

Requirement: Message Operations

The system SHALL support message editing and deletion with proper audit trail and permissions.

Scenario: Edit own message

• WHEN a user edits their own message within 15 minutes

  {
    "type": "edit_message",
    "message_id": "msg-123",
    "content": "Updated: Equipment temperature stabilized at 75°C"
  }
  

• THEN the system SHALL update the message content
• AND set edited_at timestamp
• AND broadcast the edit to all connected clients
• AND preserve original message in audit log

Scenario: Delete message

• WHEN a user deletes their own message or admin deletes any message
• THEN the system SHALL perform soft delete (set deleted_at)
• AND broadcast deletion event to all clients
• AND clients SHALL show "message deleted" placeholder
• AND preserve message in database for audit

Scenario: React to message

• WHEN a user adds a reaction emoji to a message

  {
    "type": "add_reaction",
    "message_id": "msg-123",
    "emoji": "👍"
  }
  

• THEN the system SHALL store the reaction
• AND broadcast to all connected clients
• AND aggregate reaction counts for display