Observability and metrics

Prometheus metrics

federiq serve exposes metrics at /metrics in Prometheus text format.

Metric	Kind	Labels
`federiq_queries_total`	counter
`federiq_query_errors_total`	counter
`federiq_query_duration_seconds`	histogram
`federiq_http_requests_total`	counter	`route`
`federiq_auth_failures_total`	counter

Scrape config:

scrape_configs:
  - job_name: federiq
    static_configs:
      - targets: ["federiq.internal:7878"]

export FEDERIQ_LOG_FORMAT=json
federiq serve

Emits one JSON object per log event (stdout), with event fields flattened so they're indexable by downstream shipper.

Log levels are controlled by RUST_LOG:

RUST_LOG=federiq=debug,federiq_core=debug federiq serve

Every source's ATTACH latency is tracked by the engine:

federiq query "SELECT 1" --timing

per-source attach timing:
  events               parquet  2.14ms
  sales                postgres 84.2ms

And from Python:

import json
print(json.dumps(engine.attach_timings(), indent=2))

DuckDB's EXPLAIN ANALYZE gives operator-level timing that naturally includes federated source scan costs:

federiq query "SELECT COUNT(*) FROM events" --explain-analyze