Skip to main content
Alchemy runs typed functions in Firecracker microVMs. A handful of words cover almost everything the API, the web UI and these docs talk about.

How they fit together

Say you create a function enrich-company from a Python handler and a JSON Schema contract. Alchemy stores the handler, builds version 1 and marks it READY. When you POST an input to the function, the call becomes an invocation. A host with a free slot takes it, checks the input against the contract, restores a microVM from the version’s snapshot and runs the handler. The output is checked against the contract on the way back, and the record of the invocation lands in PostgreSQL. If the handler raises, the invocation runs again as a second attempt after a backoff. If you change the handler, you create version 2. Callers that pinned version 1 keep running it, and everyone else moves to version 2 once it is READY.

Choose what to learn next

  • Architecture follows one invocation from the server through the queue to a microVM and back.
  • Functions and versions covers the settings, the spec fields and how to pin a version.
  • Contracts explains the two schema kinds and what happens when an input or output breaks one.
  • Invocations covers sync and async calls, queueing, retries, idempotency keys and statuses.
  • Runtimes and builds explains how an image or a set of source files becomes something a host can boot.
  • Isolation describes what separates one invocation from another.
  • Secrets and Logs cover what a function receives and what it leaves behind.