Add observability

Production issues surface in three signals: logs, traces, and metrics. A baseline setup captures all three without much ceremony.

1. Logs

import { logger } from "@workspace/observability"

logger.info("user.signed_up", { userId: user.id })

Structured JSON to stdout; your platform's log shipper picks them up.

2. Traces

The honoLogger middleware in packages/api instruments every request. Custom spans:

import { trace } from "@workspace/observability"

await trace("embed-document", async () => embed(text))

3. Metrics

Counter and gauge helpers wrap the platform's metrics client. The request_total{route=...,status=...} counter ships automatically from the Hono middleware; custom counters live alongside.

What's next

• Run background jobs — instrument workers with the same three signals.
• Production readiness checklist — verify dashboard alerts fire on the right thresholds.