Skip to content

README.md

Latest commit

 

History

History
148 lines (108 loc) · 4.07 KB

README.md

File metadata and controls

148 lines (108 loc) · 4.07 KB
Raw

PDF CV Generator

AI-powered resume builder that adapts your CV to each job description. Uses any OpenAI-compatible API to rewrite bullet points and summary to match the target role.

Setup

cp .env.example .env
npm install

Configure your AI provider in .env:

Variable Description
AI_API_KEY Your API key
AI_BASE_URL API base URL
AI_MODEL Model name
AI_RESPONSE_PATH Response path in the API output

Default values for DeepSeek (OpenAI-compatible):

AI_API_KEY=sk-your-key
AI_BASE_URL=https://api.deepseek.com
AI_MODEL=deepseek-chat
AI_RESPONSE_PATH=choices.0.message.content

For any other OpenAI-compatible provider, just change the values. The AI_RESPONSE_PATH follows dot notation with numeric indices for arrays (e.g., choices.0.message.content).

Quick Start

Create your data.json with the template:

npm run init
# Edit data.json with your data

Generate a CV from a job description file:

npm run generate -- vaga.txt
# Output: output/cv-en.pdf

With flags:

npm run generate -- vaga.txt --lang pt-BR
# Output: output/cv-pt-BR.pdf

Commands

Command Description
npm run init Create data.json template
npm run generate -- arquivo_vaga.txt Generate CV adapted to the job (output in output/ folder)

Flags for generate command

Flag Description
--lang pt-BR, -l pt-BR Output in Brazilian Portuguese
--extract Extract plain text alongside PDF as .txt
--skip-range YYYY-YYYY Omit experiences falling entirely within a year range

Examples

# Portuguese CV
npm run generate -- vaga.txt --lang pt-BR
# Output: output/cv-pt-BR.pdf

# Extract text for copy-paste
npm run generate -- vaga.txt --extract
# Output: output/cv-en.pdf + output/cv-en.txt

# Skip experiences before 2020
npm run generate -- vaga.txt --skip-range 2017-2019

Output

Generated PDFs and text files are saved to the output/ directory (gitignored).

Data Source

Your resume data lives in data.json (gitignored). Edit it directly to add metrics, adjust bullets, or restructure experience entries.

Custom Prompts

AI behavior for rewriting is controlled by two files in config/:

File Purpose
config/prompts.template.json Default AI rules
config/prompts.custom.json Your custom overrides

Created automatically by npm run init. Custom rules are deep-merged into the template by key — you only need to define what you want to override.

Example — config/prompts.custom.json:

{
  "rewriteBullets": {
    "system": {
      "rules": {
        "animation_boost": "ANIMATION BOOST: Prioritize Lottie and After Effects bullets if the JD mentions animation"
      }
    }
  }
}

Valid keys per section:

  • rewriteBullets.system: rules (merged by key), examples (concatenated), preamble (replaces)
  • rewriteSummary.system: rules (merged by key)

Translation

When --lang is not English, the AI translates the resume. Configured separately in config/:

File Purpose
config/translation.template.json Default translation rules
config/translation.custom.json Your custom overrides

Inside langRules (in config/translation.custom.json), each key is a language code (ISO 639-1 or BCP 47 like pt-BR), and must match the value passed to --lang. The value is an array of additional instructions — the base "Write all output in {language}" is always prepended automatically. Example — config/translation.custom.json:

{
  "langRules": {
    "pt-BR": [
      "Use an impersonal tone. Do not refer to self.",
      "Use masculine gender consistently when referring to the developer"
    ],
    "fr": [
      "Use formal tone (vous)",
      "Use feminine gender"
    ]
  }
}

This is the right place to adjust control tone, gender, formality for each language. The template handles the core translation rules; the custom file is only for language-specific writing instructions.