Snapshot vs Live Events

FlowLayer observability combines point-in-time state with streaming updates.

Connection sequence

On /ws connection:

1. server sends hello
2. server sends snapshot
3. server streams live events (service_status, log)

Why both exist

• snapshot gives deterministic initial state
• live events keep the view current
• replay with get_logs fills gaps after reconnect or event loss

Reliability model

• live log events are best-effort
• get_logs with after_seq is the continuity mechanism
• effective_limit in responses is the authoritative bound

Practical recovery flow

1. store highest seen seq
2. reconnect
3. request get_logs with after_seq
4. merge by sequence number

Use this model for both TUI behavior and custom clients.

References:

• WebSocket Lifecycle
• Events
• get_logs replay for log continuity