[BUG]: <Provide a clear, descriptive title>

[yamseats]

Describe the bug

Easy Apply changed Phone Country to Democratic of Congo +243

Steps to reproduce

Upon running the code, it fills the first module for easy apply to democratic of congo and outputs a chatgpt response of "I do not have access to personal information such as mobile phone numbers. Is there anything else I can help you with?" via the mobile phone number field.

Attempted to use the example code as well after it may be my information via https://github.com/AIHawk-FOSS/Auto_Jobs_Applier_AI_Agent?tab=readme-ov-file#3-plain_text_resumeyaml but still gives same output/error

personal_information:
  name: "Jane"
  surname: "Doe"
  date_of_birth: "01/01/1990"
  country: "USA"
  city: "New York"
  address: "123 Main St"
  zip_code: "520123"
  phone_prefix: "+1"
  phone: "5551234567"
  email: "jane.doe@example.com"
  github: "https://github.com/janedoe"
  linkedin: "https://www.linkedin.com/in/janedoe/"

Expected behavior

Enter my information, not input Democratic Republic of Congo

Actual behavior

After letting it run, thinking it may be a slight error, it still persists on all applications

Branch

main

Branch name

No response

Python version

3.12

LLM Used

ChatGPT/Open AI

Model used

GPT-4o then switched to GPT-4o-mini to test

Additional context

No response

[Tgenz1213]

This happens when the model enters the answer wrong the first time. Go to answers.json and update it.

[surapuramakhil]

@yamseats @Tgenz1213 I believe especially on smaller size models we need to say to model that "phone & mobile number are same" in prompt engineering.

@49Simon can share what you think?

If that's the case, updating prompts would work.

[49Simon]

If it's the first time you're running it and it puts +243, there's an issue and we may need to improve the prompts. But if the answer from previous runs was saved in answers.json as 243, thats why it will override it because it will fetch from that file first.

@surapuramakhil We may need to improve prompts for some cases especially answering textbox areas where it currently fails to answer them and throws an error. And this affects bigger models too gpt4o and gemini so far I've seen.