v0.91.0 Release Recap

v0.91.0 is a non-breaking dashboard release that makes the admin debugger feel more like an agent observability console by surfacing metrics first.

What's New

Area	New in v0.91.0	Why it matters
Metrics API	/admin/api/metrics	Provides a compact dashboard payload for health, ingest, retrieval, and latency
Metrics dashboard	Health signals, rate bars, latency cards, and raw JSON on /admin/	Gives operators a quick local view before drilling into node-level debugger evidence
Debugger continuity	Search and belief inspection remain in the same dashboard	Keeps the workflow from metric signal to candidate node to audit evidence in one place
Feature metadata	admin-metrics-dashboard	Lets live deployments advertise the richer admin dashboard surface

Operator Flow

Open the dashboard:

open http://localhost:7702/admin/

Fetch the metrics payload directly:

curl http://localhost:7702/admin/api/metrics

Search and inspect a node from the same page:

curl "http://localhost:7702/admin/api/search?ns=my-app&q=auth&limit=10"
curl "http://localhost:7702/admin/api/belief?ns=my-app&id=NODE_UUID"

Compatibility

This release is intentionally non-breaking. Existing admin stats, debugger search, belief audits, REST, GraphQL, SDK, and manifest verification behavior are unchanged.

Verification

Release verification included:

go test -count=1 ./internal/admin
go test -count=1 ./...
npm run docs:build
git diff --check

The next dashboard slice should introduce a Svelte admin app shell so metrics cards, debugger search, and ranking comparisons can evolve as typed components.