Quick Start¶

This guide will get you up and running with the Bewerbung Generator in just a few minutes.

Prerequisites¶

Before starting, make sure you have:

Completed the Installation steps
Prepared your profile and job description files
Configured at least one AI provider (or use sample content)

Your First Application¶

Prepare Input Files

Place your files in the appropriate directories:
- Profile: profil/YYYYMMDD_your_name.pdf
- Job Description: Stellenbeschreibung/YYYYMMDD_job_title.txt
Example:
```
# Profile file
profil/20250604_dr_setz.pdf

# Job description file
Stellenbeschreibung/20250624_61383_SeniorDevOpsEngineer.txt
```
Generate Application
```
make generate
```
This will:
- Find the newest profile and job description files
- Create an output directory
- Generate personalized application documents
- Create PDF versions
- Generate documentation and regeneration scripts
Review Output

Your generated application will be in:
```
Ausgabe/YYYYMMDD_jobid_JobTitle-YYYYMMDD_profile_name/
```
The output includes:
- anschreiben.md - Cover letter
- lebenslauf.md - CV/Resume
- anlagen.md - Attachments list
- pdf/ - PDF versions
- README.md - Generation documentation
- regenerate.sh / regenerate.bat - Regeneration scripts

Understanding the Output Structure¶

Depending on your OUTPUT_STRUCTURE setting:

Legacy Structure (legacy):

Ausgabe/20250624_job-20250604_profile/
├── anschreiben.md
├── lebenslauf.md
├── anlagen.md
└── pdf/

By Model Structure (by_model):

Ausgabe/20250624_job-20250604_profile/
├── claude_sonnet-3-5/
│   ├── anschreiben.md
│   ├── lebenslauf.md
│   ├── generation.log
│   └── README.md
└── llama_3-2-latest/
    ├── anschreiben.md
    ├── lebenslauf.md
    └── README.md

Both Structures (both):

Combines both legacy and by-model organization.

Using Different AI Providers¶

Automatic Provider Selection:

make generate

Force Specific Provider:

# Use Claude only
AI_PROVIDER=claude make generate

# Use Llama only
AI_PROVIDER=llama make generate

# Use sample content only
AI_PROVIDER=sample make generate

Advanced Generation Options¶

Generate with Metadata:

INCLUDE_GENERATION_METADATA=true make generate

Generate without Documentation:

GENERATE_DOCUMENTATION=false make generate

Organize by Model:

OUTPUT_STRUCTURE=by_model make generate

Testing Your Setup¶

Test AI Providers:

make test-providers

Generate and Test Regeneration:

make generate-and-test

Analyze Content Variants:

make variants

Understanding Generated Content¶

The AI generates content for these sections:

Einstiegstext: Opening paragraph expressing interest
Fachliche Passung: Technical qualifications and experience
Motivationstext: Motivation and enthusiasm for the role
Mehrwert: Value proposition and achievements
Abschlusstext: Professional closing and call to action

Each section is personalized based on:

Your profile information
The job description requirements
The specific AI provider’s capabilities

Regeneration Scripts¶

Each generated application includes scripts to reproduce the exact same output:

Unix/Linux/macOS:

cd Ausgabe/your_application_folder/
./regenerate.sh

Windows:

cd Ausgabe\your_application_folder\
regenerate.bat

These scripts:

Set the exact AI provider and model used
Reproduce the same generation environment
Generate identical content (for deterministic providers)

Next Steps¶

Now that you have generated your first application:

Configuration - Learn about detailed configuration options
Tutorials - Explore advanced usage scenarios
../development/testing - Learn about testing and validation

Common Issues¶

No AI Provider Available:

If you see “Using sample content”, either:

Configure an AI provider in .env
Install and start Ollama for local generation
Continue with sample content for testing

PDF Generation Failed:

Install system dependencies:

# macOS
brew install pango

# Ubuntu/Debian
sudo apt-get install libpango-1.0-0

Empty Generated Content:

Check your input files:

Profile file exists and is readable
Job description file contains text content
File naming follows the YYYYMMDD pattern