Skip to content

Find what went wrong (Alerts) and what you changed (Changes)

Goal: after something failed or looks off, see what went wrong on any context (Alerts) and which writes you sent and how they ended (Changes).

Prerequisites

  • The TUI has been running: Alerts holds what happened since it started (in memory, the last 500); Changes is kept across restarts.

Alerts

  1. Notice them: the context row under the tab bar says how many alerts are new, on every tab.
  2. Open Alerts: Infra top tab, 6 (Alerts) in the default layout. Showing the pane marks them read.
  3. Read the rows (newest first): Last, Severity, Context, Source, Subject, Text, Count, First. Sources:

    • risk — a P&L worker's limit breach (sc.<S>.risk.limit.breach);
    • worker — a worker failed or failed to load, a degraded or disconnected status (a feed, an EventStore), a gap in a worker's input;
    • halt — a scenario's trading halt set or lifted;
    • context — a context's api going down, and answering again;
    • tui — any TUI action that failed (the same text as its error toast).

    A repeat counts up in its row (Count) and moves back to the top. 4. Narrow: /, e.g. severity:error, source:worker, context:prod (severities: error, warning, info). 5. Follow up: for a worker alert, find the worker on Infra › Workers (/ then its id or name) and press Enter for its last load/start error and recent lifecycle events; for a feed, Infra › Connections, Enter. 6. Tidy: d dismisses the highlighted alert, c clears the list.

Changes

  1. Open Changes: Infra top tab, 7 (Changes).
  2. Read the rows (newest first): Time, Context, Change (a worker created, started, stopped, deleted or tagged; a worker command other than status; a state entry saved or deleted; a trigger scheduled or deleted), Outcome and Detail. Outcome is done, failed (red, with the reason) or declined (you said no on a protected context).
  3. Narrow: /, e.g. outcome:failed, context:prod.

The log is ~/.local/state/virtufin/tui/changes.jsonl (one JSON object per line; at 4 MB it moves to changes.jsonl.1); the last 500 are read back at start. It has no keys: it is a record.

What you should see

  • Alerts: one row per distinct problem, with how often and since when.
  • Changes: every write the TUI itself sent, pane or AI tab alike.

Troubleshooting

Symptom Cause and fix
Alerts is empty after a restart It is in memory only; check ~/.local/state/virtufin/tui/tui.log for earlier failures.
A change made by a script is missing Scripts run as their own processes with their own client: they are not in Changes.
A halt from before the TUI started is not an alert Halts found at connect are shown in the context row (HALTED: <scenarios>), not as alerts.
More detail on a failure ~/.local/state/virtufin/tui/tui.log (rotates at 10 MB, three old files).

Reference: Keybindings › Alerts and › Changes.