LiteLLM 1.109.0-dev.2 adds daily limits and gateway controls

LiteLLM 1.109.0-dev.2 is a development build. Its release notes include substantive gateway controls alongside many fixes, but teams should not treat the version label as a stable production recommendation.

More explicit limits

Deployment request-per-day and token-per-day limits join rate-limit enforcement, with a configurable daily reset time. Project quotas gain provider- and model-specific RPM/TPM requirements, while model budgets can address model families and access groups.

Integration and operator changes

The MCP toolset editor gains search and multi-select. The proxy accepts priority through x-litellm-priority, and a Claude Code gateway user can select a session team. Provider additions include OpenAI-compatible endpoints and new speech, image and vector-search integrations; those integrations do not include provider accounts or free inference.

Several entries are marked breaking, including customer-list pagination and fallback-rule changes. The migration precheck changes to a rolled-back dry run of pending migrations. Review the exact release notes and stage an upgrade against existing clients, budgets and fallback rules before relying on this development build.

See the litellm hub.