Telemetry / Active
Live system metrics
These are real numbers from the running system, refreshed every few seconds. Not vanity metrics; just what's actually in memory, in the skill library, and in the knowledge graph.
Last sync // —
CH 01▍Substrate
MEMOFFLINE
Q&A pairs
SKLOFFLINE
Skills
GRFOFFLINE
Graph triples
NODOFFLINE
Graph nodes
PATOFFLINE
Patterns
CH 02▍Graph density
EDGOFFLINE
Edges
DEGOFFLINE
Avg node degree
(2·edges / nodes — the densification gauge)
CH 03▍Where Q&A pairs come from
No data — waiting on first live sample.
CH 04▍Skill library by kind
No data — waiting on first live sample.
CH 05▍Operational signals
FIXOFFLINE
Errors corrected
(mistakes Finch caught and replayed correctly — monotonic)
UPDATE // substrate counts, graph density, the source mix and the skill kinds all refresh live every few seconds. Failure totals fall back to a stored snapshot when Finch is offline.
SCOPE // What you won't find here: model internals, prompts, evaluator details, scoring logic, or any of the orchestration plumbing. By design.