Skip to Content
sedge

Logs and runs

See what sedge did, how long it took, and what went wrong, on your machine only.

sedge keeps a record of what it does, for you. Nothing is sent anywhere: there is no telemetry to us or anyone else, and no exporter to turn on by accident.

Runs

Every run is recorded when it ends (events triage skipped never started one; they show in the inbox as skipped):

  • which agent ran, and what woke it (a Telegram message, a signal, a timer);
  • how long it took, and whether it worked; if not, why;
  • each tool call, with how long it took and whether it failed;
  • the tokens it used, and their cost at the model's listed price, when it lists one (see Limits and costs).

sedge dash runs (or /sedge runs in the TUI) lists them, newest first. Select one for its tool calls on a timeline.

sedge dash: runs, with the tool calls of the selected one

Logs

Everything the daemon logs is kept: its own lines, and whatever pi, MCP servers, channels and signals print. sedge dash logs shows them, newest first; l steps through levels (all, info, warn, error).

sedge dash: the log, at warn and above

Numbers

The dashboard's header counts, since the daemon started:

  • runs, and how many failed;
  • how long runs take on average, and the slowest;

and what the agents used today: tokens and their cost, against your daily limits when you set them.

These are Effect metrics, kept in memory. The daemon also counts tool calls, events by source, and deliveries.

Where it's kept

FileWhat
~/.sedge/logs/runs.jsonlOne line per finished run.
~/.sedge/logs/sedge.jsonlOne line per log line, with its level and source.
~/.sedge/logs/daemon.logThe daemon's plain output, as before.

Each moves aside past 10 MB, keeping one previous file. They're plain JSON lines: read them with jq, or delete them whenever you like.

On this page