validate-command-parameters

Validating extracted parameters

Pydantic answers one field at a time: right type, matches the pattern, was supplied. It cannot answer the questions that only exist once the whole call is on the table — is this end date before its own start date, is anything actually selected for this bulk command to operate on, does this decision value exist in the backend's vocabulary. validate_extracted_parameters is where those go.

The contract

A static method on the command's Signature class:

class Signature:
    class Input(BaseModel):
        ...

    @staticmethod
    def validate_extracted_parameters(
        workflow: fastworkflow.Workflow,
        command: str,
        cmd_parameters: "Signature.Input",
    ) -> tuple[bool, str]:
        ...
        return (True, '')

Return (True, '') to proceed, or (False, message) to send message back to the caller. The hook is optional — a command without it simply skips this stage.

Where it runs in the order

The framework validates in this sequence, and stops at the first failure:

1. type coercion per field   str/int/float/bool/Enum/list, plus any `pattern`
2. missing required fields   sentinel values count as missing; empty list counts as missing
3. db_lookup per field       fuzzy resolution against a live key set
4. validate_extracted_parameters      <- only reached if 1-3 all passed

So this hook may assume every field is present and of the declared type. Do not re-check types, required-ness or regex here — declare those on the field and let stage 1 do it. What is left is exactly what the field declarations cannot express.

It is a correction hook, not just a check

The cmd_parameters object it receives is the one the command body will read, so writing to it is the intended way to normalise:

    @staticmethod
    def validate_extracted_parameters(workflow, command, cmd_parameters):
        if not cmd_parameters.order_id.startswith('#'):
            cmd_parameters.order_id = f'#{cmd_parameters.order_id}'
        return (True, '')