Logfire

Warnings

• breaking logfire.configure() must be called before any spans, logs, or instrument_*() calls. Omitting it causes all telemetry to be silently dropped — no error is raised.
  Fix: Call logfire.configure() at application startup, before importing or initializing any instrumented libraries. For FastAPI: call it before app = FastAPI().
• breaking instrument_httpx() parameters changed: capture_request_json_body, capture_request_text_body, capture_request_form_data, and capture_response_json_body were removed and replaced by capture_request_body and capture_response_body.
  Fix: Replace: logfire.instrument_httpx(capture_request_json_body=True) with logfire.instrument_httpx(capture_request_body=True, capture_response_body=True)
• breaking install_auto_tracing() requires explicit modules parameter — there is no default. Calling it without modules raises TypeError.
  Fix: Always specify modules: logfire.install_auto_tracing(modules=['myapp', 'myapp.services'])
• gotcha pydantic-logfire on PyPI is a deprecated stub package that just depends on logfire. Use pip install logfire directly.
  Fix: pip install logfire. Remove pydantic-logfire from requirements. import logfire not from pydantic_logfire import ...
• gotcha Logfire pricing changed January 2026. Free Personal plan now limited to 1 seat and 3 projects. Teams exceeding these limits needed to upgrade by February 1, 2026 or be auto-upgraded.
  Fix: Check current plan at logfire.pydantic.dev. Team plan ($49/mo) for 5 seats/5 projects. Growth plan ($249/mo) for unlimited.
• gotcha Without LOGFIRE_TOKEN, logfire.configure() defaults to local console output only. Cloud export requires the token. Easy to develop locally and forget to set the token in production.
  Fix: Set LOGFIRE_TOKEN env var in production. Verify export is working: logfire.configure() logs the dashboard URL if token is valid.

Install

• pip install logfire Core SDK
• pip install logfire[fastapi] With FastAPI auto-instrumentation
• pip install logfire[httpx,requests,sqlalchemy,asyncpg] With common integrations
• pip install pydantic-logfire DEPRECATED stub — installs logfire automatically

Imports

• logfire.configure

  import logfire
  
  # Must call configure() FIRST — before any spans or instrumentation
  # Reads LOGFIRE_TOKEN env var automatically if set
  logfire.configure()
  
  # Then instrument libraries
  logfire.instrument_fastapi(app)
  logfire.instrument_httpx()
  logfire.instrument_sqlalchemy(engine)
  
  # Structured logging
  logfire.info('User {user_id} logged in', user_id=123)
  
  # Manual spans
  with logfire.span('Processing order {order_id}', order_id=order.id):
      result = process(order)

  logfire.configure() must be called before any logging, spans, or instrumentation. Without it, spans are created but not exported. The LOGFIRE_TOKEN env var is read automatically at configure() time.

Quickstart

Always call logfire.configure() first. LOGFIRE_TOKEN from env. Free tier: 1 seat, 3 projects.

import logfire
import os

# Set env var: export LOGFIRE_TOKEN='your-token'
# Or pass directly (not recommended for production):
logfire.configure(
    token=os.environ.get('LOGFIRE_TOKEN'),
    service_name='my-service',
    environment='production'
)

# Structured log with attributes
logfire.info('Processing request', user_id=42, endpoint='/api/data')
logfire.warning('Rate limit approaching', limit=100, current=95)
logfire.error('Database error', error_type='ConnectionTimeout', db='postgres')

# Span (traces an operation)
with logfire.span('Fetch user data {user_id}', user_id=42):
    # Code here is traced
    data = fetch_user(42)
    logfire.info('Fetched {record_count} records', record_count=len(data))

# FastAPI integration
from fastapi import FastAPI
app = FastAPI()
logfire.instrument_fastapi(app)  # auto-traces all routes