LiteLLM looks like an AI gateway that ships like a product under pressure. ToolVitals shows 1,939 commits in 30 days, 24 release events in 30 days, 30 releases in 90 days, and 26 active contributors. The official site now frames it as a Python SDK and AI Gateway for calling 100+ LLM APIs in OpenAI format, with model access, fallbacks, cost tracking, guardrails, load balancing, and logging (LiteLLM, AI Gateway).
What the release train says
The latest release notes back that up. In 1.84.0-dev.1, the repo highlights cosign-signed Docker images, pins the signing key to commit 0112e53, and lists fixes for GCP IAM token caching, proxy worker health checks, dashboard session cleanup, and pricing. That is ops work, not vanity work. The stable v1.83.10-stable page keeps the same signature-verification framing.
The repo itself is big and active. GitHub shows 45.5k stars, 1.2k issues, 1.6k pull requests, and 38,706 commits. That is not a toy integration layer. It is a large surface area with a lot of churn.
What the data does not tell you
ToolVitals cannot tell you whether LiteLLM routes traffic correctly for your stack, whether the guardrails match your policy, or whether the docs get you to production fast. It does not see code quality, revenue, support quality, or customer satisfaction. The site claims 240M+ docker pulls, 1B+ requests served, and 80% uptime, but that is vendor marketing, not ToolVitals telemetry.
How it compares
Among adjacent developer tools in the dataset, LiteLLM’s 24 release events in 30 days sit above Hermes Agent’s 6 and just below OpenCode’s 28. On stars, LiteLLM’s 45,513 trails OpenCode’s 153,674 and Hermes Agent’s 129,983. The pattern is clear. LiteLLM is spending a lot of energy on maintenance, but it is not the loudest social signal in the group.
Bottom line
If your team needs a shared gateway for multiple LLM providers, budgets, logging, and OpenAI-compatible access, evaluate LiteLLM now. If you only need a thin SDK wrapper, this is probably more machinery than you want. The project looks alive, hardened, and very opinionated about production use.