Why I Use It
I switch between DeepSeek, GPT-4o, and Claude depending on what I'm building—cost, speed, or capability. LiteLLM gives me one API surface for all of them, plus it handles retries, fallbacks, and load balancing. I don't have to rewrite prompts when I change models.
When I Recommend It to a Client
When clients are committed to vendor diversity or want cost optimization through intelligent routing. Also useful when they need A/B testing different models without code changes.
Full Review
LiteLLM is my routing layer for all LLM calls. It lets me switch between providers without changing code, handles retries and fallbacks automatically, and gives me detailed logging for cost tracking.