nulog

Logging + metrics as a Nu app. Append-only logs and kh57-backed metric series kept in nu.v, written and read in Nu. Every write is a Nu Command; every read is a Nu Query.

Usage

Everything happens inside a nulog.store(...) bracket. Writes are Nu Commands, reads are Nu Queries that yield list[dict] or dict[str, int].

import nu, nulog

tree = nu.With(nulog.store(),
    body=nu.v.Transaction(
        nulog.info("app", "server started", port=8080)
        >> nulog.warn("app", "cache miss", key="user:42")
        >> nulog.error("app", "request failed", path="/checkout", code=500)
        >> nulog.observe("cpu_load", 0.42)
    )
    >> nu.v.Snapshot(nu.print("tail:", nulog.tail("app", 3)))
    >> nu.v.Snapshot(nu.print("errors:", nulog.errors("app")))
    >> nu.v.Snapshot(nu.print("count:", nulog.count_by_level("app"))),
)

nu.run(tree)

Read atoms cover the search space:

• Logs: tail(stream, n), head(stream, n), by_level(stream, level), errors(stream), since(stream, ts_us), between(stream, start_us, end_us), search(stream, text), count_by_level(stream).
• Metrics: range_metric(name, begin, end), sample_metric(name, n, begin, end).

Rows come out as plain dicts. Log rows: {"key": int, "ts_us": int, "level": str, "msg": str, "fields": dict}. Metric rows: {"ts_us": int, "ts": float, "value": float}.

UI

The viewer is a bracket too: nulog.ui(streams, port=) boots a nudle server under nulog.store(...). Same shape as any Nu-nudle app.

import asyncio, nu, nulog

tree = nu.With(
    nulog.store("logs.db"),
    nulog.ui(["app", "scraper"], port=8080),
    body=nulog.info("app", "server started")
         >> nu.ForeverDo(
             nu.v.Transaction(nulog.info("app", "tick")) >> nu.Delay(1.5)
         ),
)

asyncio.run(nu.arun(tree))

Open http://127.0.0.1:8080: live newest-first entry table, stream switcher, level filter, message search, per-level count stats. The table + counts repaint every second off the same read atoms above.