Intelligent Request Routing With Load Balancing

via “intelligent-provider-routing-with-load-balancing”

Unified API for 100+ LLM providers — OpenAI format, load balancing, spend tracking, proxy server.

Unique: Implements a pluggable routing strategy system where each strategy (round-robin, least-busy, cost-optimized, latency-optimized) is a separate function that scores deployments based on real-time metrics. Tracks per-deployment latency percentiles and error rates in memory, enabling intelligent decisions without external observability tools. The cooldown management system (cooldown_manager.py) prevents thrashing by temporarily deprioritizing failed deployments.

vs others: More sophisticated than simple round-robin; unlike Anthropic's batching API, supports real-time cost-aware routing across heterogeneous providers; more lightweight than full service mesh solutions like Istio