Langfuse Security Basics

Overview

Security practices for Langfuse LLM observability integrations. Langfuse captures prompts, completions, and metadata from LLM calls -- making data privacy and access control critical since traced data may contain PII.

Prerequisites

• Langfuse instance (cloud or self-hosted)
• API keys provisioned
• Understanding of data privacy requirements

Instructions

Step 1: Credential Separation

Langfuse uses separate public and secret keys. Only the secret key should be protected.

import os

# Public key: safe for client-side (identifies project)
LANGFUSE_PUBLIC_KEY = os.environ["LANGFUSE_PUBLIC_KEY"]

# Secret key: NEVER expose (grants write access to traces)
LANGFUSE_SECRET_KEY = os.environ["LANGFUSE_SECRET_KEY"]

# Host (for self-hosted)
LANGFUSE_HOST = os.environ.get("LANGFUSE_HOST", "https://cloud.langfuse.com")

# Validate on startup
assert LANGFUSE_SECRET_KEY, "LANGFUSE_SECRET_KEY required"
assert not LANGFUSE_SECRET_KEY.startswith("pk-"), "Using public key as secret key!"

Step 2: PII Scrubbing Before Tracing

Langfuse stores everything you send. Scrub PII before tracing.

import re
from langfuse import Langfuse

langfuse = Langfuse()

def scrub_pii(text: str) -> str:
    text = re.sub(r'\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z]{2,}\b', '[EMAIL]', text, flags=re.IGNORECASE)
    text = re.sub(r'\b\d{3}[-.]?\d{3}[-.]?\d{4}\b', '[PHONE]', text)
    text = re.sub(r'\b\d{3}-\d{2}-\d{4}\b', '[SSN]', text)
    return text

def traced_llm_call(prompt: str, user_id: str):
    trace = langfuse.trace(
        name="llm-call",
        user_id=user_id,  # OK: user IDs are fine
        input=scrub_pii(prompt),  # scrub before tracing
    )
    response = call_llm(prompt)  # send original to LLM
    trace.update(output=scrub_pii(response))  # scrub output too
    return response

Step 3: Access Control for Self-Hosted

# docker-compose.yml for self-hosted Langfuse
services:
  langfuse:
    image: langfuse/langfuse:latest
    environment:
      - AUTH_DISABLE_SIGNUP=true           # prevent open registration
      - AUTH_DOMAINS_WITH_SSO_ENFORCEMENT=company.com  # require SSO
      - LANGFUSE_DEFAULT_PROJECT_ROLE=VIEWER  # least privilege default
      - ENCRYPTION_KEY=${ENCRYPTION_KEY}   # encrypt data at rest

Step 4: Trace Data Retention

Configure automatic cleanup of old traces to limit exposure.

# Self-hosted: set retention via environment
# LANGFUSE_RETENTION_DAYS=90

# Cloud: use API to delete old traces
from datetime import datetime, timedelta

def cleanup_old_traces(langfuse: Langfuse, max_age_days: int = 90):
    cutoff = datetime.now() - timedelta(days=max_age_days)
    # Use Langfuse API to list and delete traces older than cutoff
    # Implement based on your Langfuse version's API
    print(f"Cleaning traces older than {cutoff.isoformat()}")

Error Handling

Examples

Secure Initialization

from langfuse import Langfuse
langfuse = Langfuse(
    public_key=os.environ["LANGFUSE_PUBLIC_KEY"],
    secret_key=os.environ["LANGFUSE_SECRET_KEY"],
    host=os.environ.get("LANGFUSE_HOST", "https://cloud.langfuse.com")
)

Resources

• Langfuse Security Docs
• Self-Hosting Guide

Output

• Configuration files or code changes applied to the project
• Validation report confirming correct implementation
• Summary of changes made and their rationale