T-autolog 2.4.0

Zero-config auto-instrumentation logging for Python

autolog

Zero-config, structured, async-safe logging for Python. Drop one line into your entrypoint and every function call gets instrumented — inputs, outputs, duration, errors, traces. No changes to your business logic required.

---

Features

• Zero-touch — autolog.start() patches your packages at import time
• Structured output — pretty for dev, json for production (Datadog/Loki/ELK)
• Async-native — sync, async def, generators, async generators all handled
• Sensitive data redaction — passwords, tokens, API keys auto-masked
• Trace IDs + bound context — bind(user_id=42) adds fields to every downstream log
• Sampling — sample=0.01 to log 1% of calls in hot paths
• Exclude patterns — skip noisy modules without touching their code
• Per-package levels — DEBUG one package, WARN another
• Production-safe — rotating files, traceback capture, body-size guards, upstream trace propagation
• Single CLI — autolog run myapp.main, no entrypoint changes needed

---

Installation

pip install autolog
pip install "autolog[fastapi]"   # FastAPI middleware
pip install "autolog[flask]"     # Flask middleware

---

Quick start

import autolog
autolog.start()

# everything you import from here on is auto-logged
from myapp.services import process_order
process_order(123)

Output:

[2026-05-07 14:23:01.042] [INFO ] [myapp.services.process_order]
  ▸ trace    : 3f4a1b2c-...
  ▸ inputs   : {"order_id": 123}
  ▸ result   : {"status": "ok"}
  ▸ duration : 0.0031 s

---

The start() API

start() is the single entry point. All options are optional.

autolog.start(
    packages=["myapp", "utils"],         # None → auto-discover
    exclude=["myapp.hot_path.*"],        # glob patterns to skip
    service="my-api",                    # label in logs
    level={"myapp": "DEBUG", "default": "INFO"},
    format="json",                       # or "pretty" (default)
    log_file="app.log",                  # rotating, 10MB × 5 backups
    sample=0.1,                          # log 10% of calls
    config="pyproject.toml",             # explicit config path
)

Resolution order

kwargs > env vars > [tool.autolog] in pyproject.toml > defaults

---

Three ways to use it

1. Code

import autolog
autolog.start(packages=["myapp"])

2. Config in pyproject.toml

[tool.autolog]
packages = ["myapp", "utils"]
exclude  = ["myapp.cache.*"]
service  = "my-api"
level    = { myapp = "DEBUG", default = "INFO" }
format   = "json"
log_file = "app.log"
sample   = 0.5

import autolog
autolog.start()   # reads pyproject.toml automatically

3. CLI runner — zero entrypoint changes

autolog run myapp.main
autolog run myapp.main --format json --level DEBUG
autolog run myapp.main --exclude "myapp.hot.*" --sample 0.1
autolog show                              # show resolved config

---

Bound context fields (structlog-style)

Attach fields once, get them on every log inside the same async context:

from autolog import bind, unbind

async def request_middleware(request, call_next):
    bind(user_id=request.user.id, org_id=request.org.id)
    try:
        return await call_next(request)
    finally:
        unbind()

Now every log inside this request includes user_id and org_id automatically — no need to pass them around.

---

Decorator mode (opt-in)

from autolog import log, no_log

@log
def calculate(x, y): ...

@log
class OrderService:
    def create(self, data): ...

@log(sample=0.01)         # log 1% of calls
def hot_path(): ...

@no_log                    # never log this one, even if package is auto-instrumented
def secret_op(): ...

@log works on both functions and classes — handles @staticmethod, @classmethod, @property, sync/async/generators automatically.

---

Trace IDs

Async-safe per-request correlation:

from autolog import new_trace_id, get_trace_id, set_trace_id

new_trace_id()                        # generate UUID4 for this context
set_trace_id("custom-id")             # use an externally-supplied one
get_trace_id()                        # read current

---

Framework middleware

FastAPI

from fastapi import FastAPI
from autolog.middleware.fastapi import AutoLogMiddleware

app = FastAPI()
app.add_middleware(AutoLogMiddleware, service="my-api")

• Honors upstream X-Request-ID, X-Trace-ID, traceparent headers
• Skips body capture for multipart/, octet-stream, text/event-stream, image/*, video/*
• Truncates large bodies at max_body_bytes (default 64KB) — won't OOM on uploads
• Async-safe: logging runs in BackgroundTask, never blocks the response

Flask

from flask import Flask
from autolog.middleware.flask import init_autolog

app = Flask(__name__)
init_autolog(app, service="my-api")

---

Configuration via environment variables

Var	Effect
AUTOLOG_LEVEL	Override level
AUTOLOG_FILE	Mirror logs to file
AUTOLOG_FORMAT	pretty or json
AUTOLOG_SERVICE	Service name
AUTOLOG_DISABLE=1	Disable all instrumentation
AUTOLOG_FILE_MAX_BYTES	Rotation size, default 10MB
AUTOLOG_FILE_BACKUPS	Rotated backups, default 5
NO_COLOR=1	Disable ANSI colors

---

Sensitive data redaction

These keys (and tokens within them) are auto-redacted:

password, passwd, token, secret, auth, credential, apikey, jwt, bearer, api_key, access_token, refresh_token, private_key

Token-aware: apiKey, API_KEY, client.secret all match. keyword, monkey, author do not match (no false positives).

---

License

MIT