Parameter Extraction

Model extracts, code validates

The parameter extractor node uses an LLM to pull defined fields (order ID, issue, amount) out of free text and returns them as variables. Define each parameter with a type and a clear description. Extraction can still fail (missing fields, wrong types, text instead of JSON), so validate the result before use, for example with a code node, and route invalid results to a clarification question or a human. Dify's node names, menus and options change between versions; check the current Dify documentation.

Validating extracted fields, run

I ran this with plain Python 3 (scikit-learn 1.9.1 where imported). It models or tests one piece of a Dify app locally; Dify itself was not running. Four example extraction outputs are checked against a schema: a complete record passes, a missing amount is caught, a numeric order_id and a text amount are caught, and plain prose is rejected as not JSON.

import json
schema = {"order_id": str, "issue": str, "amount": (int, float)}
def validate(raw):
    try:
        data = json.loads(raw)
    except json.JSONDecodeError:
        return "not JSON"
    problems = [k for k, t in schema.items() if k not in data or not isinstance(data[k], t)]
    return "ok" if not problems else f"invalid fields: {problems}"
for raw in ['{"order_id": "A1042", "issue": "damaged", "amount": 1299}',
            '{"order_id": "A1042", "issue": "damaged"}',
            '{"order_id": 1042, "issue": "late", "amount": "1299"}',
            'The order id is A1042']:
    print(f"{validate(raw):<34} <- {raw}")

Output:

ok                                 <- {"order_id": "A1042", "issue": "damaged", "amount": 1299}
invalid fields: ['amount']         <- {"order_id": "A1042", "issue": "damaged"}
invalid fields: ['order_id', 'amount'] <- {"order_id": 1042, "issue": "late", "amount": "1299"}
not JSON                           <- The order id is A1042

Ask for missing fields

In chatflows, when a required parameter is missing, ask the user for it instead of guessing.

त्वरित जाँच: Why validate parameter extractor output?

• The model can omit fields or return wrong types
• Extraction is always perfect
• Validation trains the model
• It reduces the number of nodes