LiteLLM v1.109.0-dev.2: New Providers, Pricing Updates, and Cosign Docker Image Signing

What changed?

LiteLLM v1.109.0-dev.2 adds several new capabilities, fixes, and breaking changes: - All Docker images are now signed with cosign, verifiable with a published public key for cryptographic assurance. - New providers: ai& (aiand), ZeroGPU (OpenAI-compatible), Ollama decision models. - Enforced per-provider and per-model request/token per minute quotas for project model quotas. - Enhanced auditing (SDK calls, token counting, model and provider tests). - Large set of bug fixes spanning price updates, API translations, UI logic, streaming, fallback/control logic, and provider-specific integration improvements. - Breaking changes affecting proxy pagination, fallback rule shims, router logic, context handling, file scoping, quota enforcement, and more (see the full list in the release notes).

Why does it matter to an everyday developer?

This release impacts developers using LiteLLM in several ways: - Verified Docker images: Developers can now cryptographically verify Docker images using cosign, improving supply chain security. - New models/providers: Access to new providers and improved decision model routing expands deployment and integration options. - Pricing and quota enforcement updates may change cost calculations and resource allocation logic for custom integrations or internal tooling. - Numerous bug fixes address token counting, model pricing, fallback behaviors, file management, OpenAI/Anthropic/Azure integration, and edge-case handling, impacting reliability and API compatibility. - Several breaking changes require attention if your code or usage patterns depend on legacy fallback rules, router settings, pagination, or file management.

What can the developer do now?

• If running LiteLLM via Docker, use cosign to verify image signatures before deployment. Prefer verifying via the pinned commit hash for the highest security.