Everything is an adapter: building with zero opinions

Most platforms lock you into their choices. Their model. Their identity provider. Their cloud. Their billing. The lock-in is the business model. We decided to do the opposite — and it shaped the entire engine.

The engine talks to traits

In Chimes, every integration point — LLM, voice, identity, secrets, storage, search, billing, CRM — is defined by a trait in the adapter SDK. The core only ever talks to the trait. It has zero imports of any specific vendor.

Swapping a provider is a single line of configuration and a restart. The router calls a trait; the registry resolves the implementation from config. That's the entire dependency — which means no provider is ever load-bearing.

We swapped our LLM provider on a Friday afternoon. One config line. Nothing else moved.

Agnostic, by design

The architecture maps to three layers: a core engine (Layer 1), pluggable adapters (Layer 2), and a managed multi-tenant platform (Layer 3) that runs and scales it for you. You configure providers behind their adapters; we operate everything.

No lock-in isn't a marketing line for us. It's a property of the architecture — your providers are configuration, never assumptions, so you can swap any of them in a line.