LiteLLM audit logs
Recording who created, changed, deleted or regenerated keys, teams, users and models on a self-hosted LiteLLM gateway, exporting the trail to S3, and why SSO must come first for the log to identify people.
LiteLLM features
Gateway governance and operations features of a self-hosted LiteLLM proxy: audit trails, log retention, key management, multi-region topology, PII masking, metrics and per-team logging.
LiteLLM spend log retention and prompt storage
Deciding whether a self-hosted LiteLLM stores prompts and responses in spend logs, and setting a retention period and cleanup schedule so the spend log table, and its privacy exposure, does not grow without bound.
LiteLLM master key and salt key management
Separating the LiteLLM admin credential from the key that encrypts stored provider credentials, rotating the master key safely, and why the salt key must never change once models are added.
LiteLLM multi-region deployment
Running LiteLLM proxies in several regions of one cloud with a shared PostgreSQL database and regional Redis, either active-active for latency or active-passive for disaster recovery, and how licensing and keys carry across.
LiteLLM PII and PHI masking with Presidio
Masking or blocking personal and health data in prompts and responses at the gateway with the LiteLLM Presidio guardrail, which Presidio containers you must run yourself, and which modes fit which risk.
LiteLLM Prometheus metrics and monitoring
Monitoring a self-hosted LiteLLM gateway with its Prometheus /metrics endpoint: enabling the callback, the spend, request and budget metrics available, and the labels to keep under control.
LiteLLM team and key based logging
Sending each team's or key's LLM traffic to its own logging destination, or turning logging off for a team, on a shared LiteLLM gateway, and when that separation matters for compliance.
