We're politely asking the LLM to only change some fields

Please update the following sections of `json_resume`'s fields to better match the `job_description`: "work", "skills", "interests", "projects", "summary", "label"

And to not change others

        Do not change any field names
        Do not change any values for the following fields: "name", "email", "phone", "url", "location", "profiles", "education", "certificates", "publications", "languages"

This should be more of a hard limit.

One possible solution: only pass the change-available sections to the LLM

Keep in mind that the JSON is deeply nested

Mainly for debugging

Given a JSON-like string (original_json_cv)

• add a context
• add a prompt template
• add a job_description_text
• the response (improved_json_cv) should be JSON-like, with the same fields as the original JSON, but with the updated values

Should not be changed from the original

• JSON field names
• Dates
• User data (name, phone, url, location, profiles

Either we check the response, or we don't send these fields to the model.

For

• flake8
• isort

Related to PR #15

To be able to hear "a different bell"

Something like

class ResumeRebuilder:
    """Given an original resume and a filtered_resume, update the original one to include the filtered one"""

    def __init__(self, original_resume: dict, filtered_resume: dict) -> None:
        self.original = original_resume
        self.filtered = filtered_resume

    def rebuild(self) -> dict:
        return

• ImproveMyCV.llm_setup
• cv_fields_filter

Related to #1