📘 Technical Documentation

1. Project Overview

Resumate is a local, AI-powered agent designed to tailor resumes for specific job applications. It functions as an intelligent ETL (Extract, Transform, Load) pipeline:

Extract: Ingests raw text from PDF/DOCX resumes and scrapes Job Descriptions (JDs) from URLs.
Transform: Utilizes the DeepSeek-V3 Large Language Model (LLM) to analyze strict ATS compatibility, score the resume, and rewrite content to align with the target role.
Load: Exports the tailored content into a clean, formatted Microsoft Word (.docx) document ready for submission.

2. Technology Stack

Component	Technology	Reasoning
Frontend UI	`Streamlit`	Enables rapid development of a reactive web interface entirely in Python.
AI Engine	`DeepSeek-V3`	Selected for its high reasoning capabilities and extreme cost-efficiency (~$0.14/1M tokens).
AI Client	`OpenAI SDK`	DeepSeek is fully compatible with the OpenAI API standard, allowing us to use the robust `openai` Python library by simply changing the base URL.
Web Scraping	`LangChain`	Specifically `WebBaseLoader` for robustly parsing text from Job Posting URLs.
File Processing	`PyPDF` & `Python-Docx`	Lightweight libraries for reliable text extraction (PDF) and document generation (DOCX).

3. Architecture & File Structure

The project follows a modular architecture to separate UI logic from business logic.

resumate/
├── main.py                 # Controller: UI layout and orchestration logic
├── requirements.txt        # Dependency manifest
├── .gitignore              # Security: Ignores secrets and virtual env
├── .streamlit/
│   └── secrets.toml        # Configuration: Stores the DeepSeek API Key
└── src/
    ├── llm_engine.py       # The "Brain": Manages DeepSeek API interactions
    └── doc_utils.py        # The "Hands": Handles File I/O and formatting

4. Module Deep Dive

🧠 `src/llm_engine.py` (The AI Core)

This module manages all communication with the DeepSeek API.

1. Client Initialization (get_deepseek_client)