earendil-works/pi

What you get here is a single repository that tries to solve the "glue code" problem most teams hit when building internal AI tooling. At its core there is a coding agent you can run from the command line, a unified API layer that normalizes calls across different language model providers, ready-made terminal and browser UI components, a Slack bot integration, and utilities for spinning up and managing vLLM inference pods. The idea is that instead of sourcing five separate libraries and writing the connectors yourself, you pull one repo and start from a working skeleton. The case for it is straightforward: small engineering teams building internal developer tools or AI-assisted workflows can skip several weeks of scaffolding. The unified LLM layer alone saves meaningful time if you want to swap providers without rewriting call sites. The honest reservation is that broad-surface repositories like this can go stale unevenly, with some modules kept sharp and others quietly rotting. Vet each component you actually plan to use before committing. -> Best for: small platform or devtools teams who want a headstart on internal AI infrastructure without assembling the stack piece by piece.