How far turns travel through the orchestrator — distinct turns that reached each stage.
Turns received per day.
Current status across all turns.
Each workflow node is an agent step — runs, share of turns, time at that stage, and last activity, from workflow_steps.
| Agent / node | Runs | % of turns | Avg time at stage | Last active |
|---|---|---|---|---|
| loading… | ||||
How often each node executed.
The metric that makes this a real agent dashboard.
Token counts and cost per agent aren't in the current schema (wa_turns / workflow_steps), so this stays empty instead of showing invented numbers. To populate it, the orchestrator needs to log tokens-in / tokens-out and model per LLM call — into workflow_steps.payload or a dedicated llm_calls table — or pull usage from the model provider's API. Then this becomes per-agent cost, tokens/turn, and a daily spend trend.
Your Obsidian notes as a constellation, linked by [[wikilinks]]. Drag to rearrange, hover to focus, click for details. Read-only — contents stay in Obsidian.
Notes written per day — the rhythm of the system's memory.