fix: insist for the pydantic parser

Author: thiswillbeyourgithub

WDoc/utils/prompts.py

@@ -267,6 +267,9 @@ def remove_thoughts(cls, values) -> List[str]:
 presentation of breast cancers", "classification of breast cancer", etc.
 You can also anticipate the answer like "the most used chemotherapies
 for breast cancers are anthracyclines, taxanes and cyclophosphamide".
+
+DO NOT reply anything that is not either a thought or the queries. You
+have to do your thinking INSIDE the thought json.
 """.strip()+ "\n" + multiquery_parser.get_format_instructions()),
         HumanMessagePromptTemplate.from_template(
             template="Here's the query to expand: '''{question}'''"""),