Skip to content
Ódinn ForgeÓdinn Forge Docs

Tasks and activity

The console provides two complementary operational views.

Tasks

Section titled “Tasks”

Tasks focuses on meaningful user, agent, and automation runs. Routine console reads stay hidden unless you enable System activity.

Use Tasks to:

  • search and filter recorded work;
  • inspect the actual timeline and outcome;
  • stop an active supervised job;
  • review evidence and related recovery state;
  • run a task again when its recorded input is declared retry-safe.

The server enforces retry safety. A replay button cannot turn an unsafe or unknown external action into a safe one.

Activity

Section titled “Activity”

Activity combines:

  • an Overview with operational usage and recent model conversations;
  • a History tab with searchable, filterable audit records.

Recorded token counts and model-call totals are operational telemetry, not a provider invoice. Provider billing remains authoritative.

History can filter by type, tool, actor, outcome, and date. It also supports bounded export and audit-chain verification.

CLI inspection

Section titled “CLI inspection”
Terminal window
odinn runs
odinn audit
odinn audit verify

Inspect a run:

Terminal window
odinn run show <run-id>
odinn run events <run-id>
odinn run verify <run-id>

Run-ledger verification and audit verification answer different questions. The run ledger checks one run’s event chain; audit verification checks the signed local audit journal.

Integrity checks detect damage. They do not make local state tamper-proof against an attacker who controls the state directory.

Use odinn doctor for a redacted health snapshot instead of collecting the raw state directory.