Spin up, query, and tear down a local OTLP backend (Phoenix, Grafana LGTM, OpenObserve, SigNoz, Jaeger, Tempo, Uptrace) for hermes-otel development and demos. Use when you need to SEE traces/metrics/logs in a real UI, stand up a backend for before/after validation, or debug "why don't I see my telemetry". Covers the docker-compose project-name trap, the port-conflict map, which backend supports which signal, the UI URLs + logins, and the query gotchas that bite everyone (Prometheus histogram _sum/_count, the 5-minute staleness window, OpenObserve PromQL, Phoenix GraphQL for spans).
Prove a hermes-otel change actually emits the span / metric / attribute you intended, end-to-end through a real Hermes run into a real backend. Use after implementing any telemetry feature, or to capture before/after evidence for a PR. Covers choosing a model that triggers your hook, running `hermes -z`, reading the plugin debug.log (the fastest signal), force-flush timing, the metric-ingest/staleness lag, and querying the backend to assert the telemetry landed. Pairs with hermes-otel-backends (which stands up the backend).
Turn on and understand OpenTelemetry observability for THIS Hermes agent via the hermes-otel plugin — choose a backend, configure export, and read the traces, metrics, and logs your agent produces. Use when a user asks "can I see what my agent is doing", "set up tracing/metrics for Hermes", "where do my traces go", or wants to debug agent behavior with telemetry.
End-to-end workflow for shipping a feature, bug fix, or new hook in the hermes-otel plugin as a green, well-documented pull request. Use whenever you are implementing a GitHub issue or any code change destined for a PR in this repo — it covers branching, the plugin's hook/span/metric conventions, the test fixtures, running the EXACT CI checks locally (ruff + black + coverage gate), the docs that count as acceptance criteria, optional before/after verification against a real OTLP backend, and opening the PR linked to its issue.