LLM function-calling as a structured-extraction tool

Draft
  • engineering
  • llm
  • testing

Most of the interesting LLM work I've done isn't chat. It's using a model as a parser: point it at unstructured noise, hand it a schema, get back rows.

Draft — full write-up coming. Outline below.

  • The setup: thousands of Cypress failures, no two phrased alike.
  • Function-calling as a contract: the schema is the prompt.
  • Validation and retries — treating the model like a flaky API, because it is one.
  • Cost and caching at CI scale.
  • Why this generalizes beyond tests.