pymcp_logging

Zero-dependency Python logging handler that ships log events to a Supabase table in near-realtime. Stdlib only — no requests, no httpx.

Features

• Drop-in logging.Handler for Supabase (PostgREST-compatible).
• Bounded in-memory queue with a 1.5s / 50-row batch flush.
• Gzipped NDJSON disk spool when the network is unreachable; auto-replays on recovery.
• Backpressure: drops debug/info first when the queue fills, keeps warning+.
• Redaction of common secret keys (password, token, secret, api_key, Authorization, Cookie) and JWT / sb_* / AWS-key value patterns.
• Stable machine ID (SHA-256 of Windows MachineGuid) + per-process session UUID.
• Runtime kill switch via a flag file — toggle live streaming without restarting the app.

Install

pip install pymcp_logging

Quick start

import os
import pymcp_logging as plg

os.environ["SUPABASE_URL"] = "https://<project>.supabase.co"
os.environ["SUPABASE_ANON_KEY"] = "<anon-jwt>"

plg.install()  # adds handler to root logger, registers sys.excepthook

import logging
logging.getLogger(__name__).info("hello supabase", extra={"order_id": 42})

Supabase table

Expects a public.logs table with columns roughly: id, ts, level, logger, message, machine_id, session_id, hostname, app_version, extra (jsonb), traceback.

Kill switch

Create or delete the flag file to toggle live streaming at runtime: %LOCALAPPDATA%\PyMCP\livestream_debug.flag (Windows) — checked each flush (~1.5s).

License

MIT