Job Application LaTeX GUI Generator (Gemini Version)

A modern, desktop-based GUI application built in PySide6 (Qt) designed to streamline the creation of highly tailored job applications (motivation letters and resumes). It wraps your existing modular LaTeX + Python compiling pipeline and harnesses the Google Gemini API (supporting fast, high-quality models) to write compelling content that matches your own writing voice.

---

Technical Features

1. AI Generation with Google Gemini: Utilizes the official google-generativeai SDK to connect to Google's advanced model suites (e.g. gemini-2.5-flash, gemini-2.5-pro).
2. Deep Voice Profiling ("Re-learn"): Automatically reads cover letters and resume bullets from your previously generated applications (generated/Application_*.pdf) using PyMuPDF to extract text. It feeds these examples to Gemini to build a personalized style profile stored in data/style_profile.txt so new content matches your phrasing, rhythm, and DACH register perfectly.
3. Responsive Async Workers: Generation and document compilation are executed in background QThread processes, keeping the UI silky-smooth and fully responsive.
4. Rich Multi-Tab LaTeX Editors: Allows full manual tweaks for each section (ML Subject, ML Recipient, ML Body, Resume Experience, Resume Competencies) with built-in Undo/Redo and a "Reset to Generated" capability.
5. Integrated PDF Preview: Displays your compiled output instantly side-by-side using high-fidelity rendering from PyMuPDF. Supports multi-document switching (Letter, Resume, Full Bundle), interactive zooming, page navigation, and enforces target page limits with clear warnings.
6. Flexible Settings Panel: Exposes adjustable sliders for generation temperature, spinboxes for token limits, editable dropdowns for Gemini model selection, and secure API credential storage to a local .env file.
7. Preset Manager: Stores form parameters under custom names to let you instantly apply similar profiles for Metrology, Semiconductor, R&D, or Machine Vision roles.
8. Asynchronous Recipient Extractor: Input a job URL or description, click "Fetch & Parse Info", and watch the company, hiring manager, address, and target title get instantly filled in!

---

Installation & Setup

To bypass macOS Homebrew python restrictions and ensure system stability, follow these steps to run the application in a local virtual environment:

1. Create a Virtual Environment

Run the following commands in the repository root directory:

# Create the virtual environment
python3 -m venv .venv

# Activate the virtual environment
source .venv/bin/activate

2. Install Dependencies

Install all package requirements directly inside the active virtual environment:

pip install -r requirements.txt

3. Set up your Gemini API Key

Create a .env file in the root directory (or let the app read your existing one) and add your key:

GEMINI_API_KEY=your_actual_gemini_api_key_here

The application will read this on startup. It never hardcodes your credentials.

---

How to Run

Ensure the virtual environment is active (or call its python binary directly) and execute:

# Run with active venv
python app.py

# Or run directly via path
.venv/bin/python app.py

---

How the Learning Corpus Works

The "Learn from my past applications" engine establishes a powerful feedback loop to automate a personalized writing style:

1. Startup Scan: On launch, the app scans your generated/ directory for any completed application bundles matching Application_*.pdf.
2. Text Extraction: Using pypdf, it isolates the Cover Letter (Page 1) and Resume (Pages 2-3) texts from each PDF and caches them inside data/style_cache.json with timestamp tracking.
3. Style Profile Generation ("Re-learn"): When you click the "Re-learn Voice Profile" button in the Tools menu, the app selects the Cover Letter texts and asks Gemini to conduct a professional analysis of your phrasing rhythm, DACH-region register, greeting hooks, closing structures, and technical framings.
4. Integration: The result is saved as a persistent style document at data/style_profile.txt. When creating new letters, this profile is fed into the Gemini generation system prompt to force it to replicate your exact tone.
5. Feedback Loop (Manual Corrections): If you manually edit the generated LaTeX text in the revision pane and click "Recompile", src/main.py builds the new PDF directly into your generated/ folder. The next time you click "Re-learn", this updated PDF is automatically ingested, folding your manual corrections back into your persistent style profile!

---

File Structure

• app.py: Application loader and UI QSS stylesheet.
• requirements.txt: Project dependency listing.
• src/gui/: Core GUI implementation package:
  • main_window.py: Window controller, menus, status indicators, and threads.
  • form_tab.py: Dynamic parameter form, asset lists, custom skills, and async URL fetcher.
  • editor_tab.py: Split screen LaTeX editors and high-fidelity fitz-based PDF viewer.
  • settings_dialog.py: Model tuner, token limits, and secure API key writer.
  • generator.py: System prompts, XML response parser, and Gemini caller.
  • preset_manager.py: JSON loader/saver for saved forms.
  • style_learner.py: PDF text extractor and voice analysis engine.