Thanks. We use a minimal schema[1] to prompt the LLM under the hood. We were inspired by BAML [2]. Then the output is wrangled with a converter and ajv [3]. [

Thanks. We use a minimal schema[1] to prompt the LLM under the hood. We were inspired by BAML [2]. Then the output is wrangled with a converter and ajv [3]. [1] https://github.com/inferablehq/l1m/blob/main/api/src/schema.ts#L31 https://github.com/inferablehq/l1m/blob/main/api/src/schema.... [2] https://github.com/boundaryml/baml https://github.com/boundaryml/baml [3] https://ajv.js.org https://ajv.js.org