BricksLLM

Why we included this project

LLM-backed features usually grow into a shared-key problem: several teams calling the same provider account, with nobody able to say who spent what. BricksLLM fixes that by sitting between your app and providers like OpenAI, Azure OpenAI, Anthropic, and vLLM, so you can issue API keys with per-key rate and spend limits. That makes it practical to hand keys to internal developers, students, or different pricing tiers and track usage per user, organization, or environment. It also covers the unglamorous parts of running LLMs in production: request caching, retries, failover between providers, model-level access control, and PII detection and masking on outbound requests. If you already self-host Postgres and Redis, the Docker deployment is straightforward, and the analytics and Datadog integration keep spend visible as usage grows.