feat: add LOG_FORMAT=json for structured JSON logging (#21747)

* feat: add LOG_FORMAT env var with JSON formatter for early logging

Introduce LOG_FORMAT environment variable (set to "json" to enable).
When active, logging.basicConfig() uses a JSONFormatter that outputs
single-line JSON objects with fields: ts, level, msg, caller, error,
stacktrace. This covers all log messages emitted during module imports
before Loguru's start_logger() takes over.

* feat: add JSON sink for Loguru when LOG_FORMAT=json

Add _json_sink() as a Loguru sink function that writes single-line JSON
to stdout. In start_logger(), conditionally use the JSON sink instead of
the plain-text stdout_format when LOG_FORMAT is set to "json".

* feat: suppress ASCII banner and fix alembic logging in JSON mode

- Wrap the ASCII art banner print in main.py with a LOG_FORMAT != "json"
  guard so JSON output stays machine-parseable.
- Skip alembic's fileConfig() call in migrations/env.py when
  LOG_FORMAT=json to prevent it from replacing the JSON log handlers
  installed during early startup.
This commit is contained in:
Andrei Efanov
2026-02-22 17:40:17 -06:00
committed by GitHub
parent 3ad2ea6f28
commit 9e81e1dda1
4 changed files with 101 additions and 11 deletions
+9 -1
View File
@@ -1,8 +1,9 @@
import logging
from logging.config import fileConfig
from alembic import context
from open_webui.models.auths import Auth
from open_webui.env import DATABASE_URL, DATABASE_PASSWORD
from open_webui.env import DATABASE_URL, DATABASE_PASSWORD, LOG_FORMAT
from sqlalchemy import engine_from_config, pool, create_engine
# this is the Alembic Config object, which provides
@@ -14,6 +15,13 @@ config = context.config
if config.config_file_name is not None:
fileConfig(config.config_file_name, disable_existing_loggers=False)
# Re-apply JSON formatter after fileConfig replaces handlers.
if LOG_FORMAT == "json":
from open_webui.env import JSONFormatter
for handler in logging.root.handlers:
handler.setFormatter(JSONFormatter())
# add your model's MetaData object here
# for 'autogenerate' support
# from myapp import mymodel