Provider-agnostic LLM API
Normalizes auth, model naming, streaming, and error semantics across providers so app code stays focused on messages, tools, and outputs; switching vendors becomes a config change.
Pi Monorepo turns agent building into composable building blocks: a provider-agnostic LLM API for model switching, a tool-calling runtime for stateful workflows, and a set of CLIs/UIs plus deployment helpers. Teams can ship internal assistants, automation, and chat interfaces from one TypeScript workspace with consistent scripts, configs, and packaging.
| ✕Traditional Pain Points | ✓Innovative Solutions |
|---|---|
| Provider APIs differ in auth, streaming, and payloads, making switching expensive | A unified multi-provider LLM API decouples model choice from app logic |
| Agent runtime, tool calling, and UI layers often live in separate repos and release cycles | One monorepo packages runtime, CLI, TUI/Web UI, and ops tooling with consistent build/check scripts |
1git clone https://github.com/badlogic/pi-mono.git && cd pi-mono && npm install1npm run build && npm run check1./.test.sh # or ./.pi-test.sh| Core Scene | Target Audience | Solution | Outcome |
|---|---|---|---|
| Terminal dev assistant for engineering teams | engineering teams | ship code Q&A, changes, and task breakdown via a coding-agent CLI/TUI | less context switching and faster review-to-fix cycles |
| Multi-model abstraction layer for product platforms | AI platform teams | use one LLM API to hide provider differences | switch models by cost/compliance/quality without rewriting app logic |
| Inference delivery tooling for infra teams | infrastructure teams | manage vLLM deployments with pods tooling | deliver internal inference endpoints faster with more standardized ops |
