Adapter Execution

Adapters don’t care why they’re called — they just run. Their purpose is to handle a single responsibility (extract, send, log, evaluate) in a way that’s fully observable and context-aware.

They work like stateless microservices, but smarter — because they plug directly into LLM-driven flows and agentic reasoning.

Runtime Protocol (Under the Hood)

All adapters conform to a standard request format:

{
  "input": { ... },
  "context": { ... },
  "adapter": "adapter-name"
}

And they return:

{
  "output": { ... },
  "context": { ... }
}

That means they’re easy to test, monitor, and plug into anything — from cron jobs to agent loops.

Why This Matters

Adapters are how you scale without rewriting logic. You can create once, configure through the UI, and deploy everywhere — whether it’s a contract review flow, a fraud detection agent, or a fully autonomous research loop.