Configuration

This guide covers all configuration options for the Bewerbung Generator.

Environment Variables

The application is configured through environment variables in the .env file.

AI Provider Configuration

AI_PROVIDER

Selects which AI provider to use.

  • auto (default): Try Llama → Claude → Sample content

  • claude: Use Claude API only

  • llama: Use Ollama/Llama only

  • sample: Use built-in sample content only

ANTHROPIC_API_KEY

Your Claude API key from Anthropic Console.

ANTHROPIC_API_KEY=sk-ant-api03-...
LLAMA_MODEL

Specific Llama model to use with Ollama.

LLAMA_MODEL=llama3.2:latest
LLAMA_MODEL=llama3.2:3b
OLLAMA_BASE_URL

Base URL for Ollama API server.

OLLAMA_BASE_URL=http://localhost:11434

Output Configuration

OUTPUT_STRUCTURE

How to organize generated files.

  • legacy: Single directory structure (backward compatible)

  • by_model: Separate subdirectories for each AI client/model

  • both: Generate both legacy and by-model structures

INCLUDE_GENERATION_METADATA

Whether to generate generation_info.json with technical details.

  • true: Include detailed metadata

  • false: Skip metadata generation

GENERATE_DOCUMENTATION

Whether to create README files and regeneration scripts.

  • true (default): Generate comprehensive documentation

  • false: Skip documentation generation

Testing Configuration

REGENERATION_TEST_MODE

How strictly to compare regenerated content.

  • content: Compare actual content (handles AI variations)

  • structure: Compare only file structure

  • strict: Exact byte-for-byte comparison

REGENERATION_TEST_TIMEOUT

Maximum time (seconds) for regeneration tests.

REGENERATION_TEST_TIMEOUT=120

File Organization Patterns

Input File Naming

Profile Files (profil/ directory):

Pattern: YYYYMMDD_identifier.pdf

Examples:

profil/20250604_dr_setz.pdf
profil/20250101_max_mustermann.pdf

Job Description Files (Stellenbeschreibung/ directory):

Pattern: YYYYMMDD_jobid_title.txt

Examples:

Stellenbeschreibung/20250624_61383_SeniorDevOpsEngineer.txt
Stellenbeschreibung/20250615_12345_PythonDeveloper.txt

Output Directory Structure

Legacy Structure:

Ausgabe/YYYYMMDD_jobid_title-YYYYMMDD_profile/
├── anschreiben.md
├── lebenslauf.md
├── anlagen.md
├── pdf/
│   ├── anschreiben.pdf
│   ├── lebenslauf.pdf
│   └── anlagen.pdf
├── README.md
├── regenerate.sh
└── regenerate.bat

By-Model Structure:

Ausgabe/YYYYMMDD_jobid_title-YYYYMMDD_profile/
├── claude_sonnet-3-5/
│   ├── anschreiben.md
│   ├── lebenslauf.md
│   ├── anlagen.md
│   ├── generation.log
│   ├── generation_info.json
│   ├── README.md
│   └── regenerate.sh
└── llama_3-2-latest/
    ├── anschreiben.md
    ├── lebenslauf.md
    ├── anlagen.md
    └── README.md

Template Configuration

The application uses Jinja2 templates located in the templates/ directory.

Template Variables

Environment Variables (UPPERCASE):

Available in all templates from .env file or environment.

{{ BEWERBUNG_ADRESSAT_FIRMA }}
{{ BEWERBUNG_STELLE }}
Dynamic Variables (lowercase):

Generated at runtime by AI or extracted from input.

{{ einstiegstext }}
{{ fachliche_passung }}
{{ motivationstext }}
Profile Variables (mixed case):

Extracted from profile files.

{{ vollstaendiger_name }}
{{ telefon }}
{{ email }}

Custom Templates

You can customize templates by editing files in templates/:

  • anschreiben.md.j2 - Cover letter template

  • lebenslauf.md.j2 - CV/Resume template

  • anlagen.md.j2 - Attachments template

Example template customization:

# Custom cover letter section
{% if custom_section %}
## {{ custom_section_title }}

{{ custom_section_content }}
{% endif %}

AI Provider Settings

Claude API Configuration

Model Selection:

The application automatically uses the latest Claude model. You can specify a different model by modifying the client code.

API Limits:

Respects Claude API rate limits and includes retry logic.

Caching:

Generated content is cached to avoid redundant API calls.

Ollama/Llama Configuration

Model Management:

# List available models
ollama list

# Pull a specific model
ollama pull llama3.2:3b

# Remove a model
ollama rm llama3.2:3b

Performance Tuning:

# Set custom Ollama options
OLLAMA_NUM_PARALLEL=4
OLLAMA_MAX_LOADED_MODELS=2

Custom Model Settings:

You can modify the Llama client to use custom parameters:

# In llama_api_client.py
generation_params = {
    "temperature": 0.7,
    "max_tokens": 2000,
    "top_p": 0.9
}

Logging Configuration

Log Levels:

  • INFO: General operation information

  • WARNING: Non-critical issues

  • ERROR: Critical problems

Log Files:

  • generation.log: Detailed generation process log

  • Console output: Important messages and errors

Log Format:

2025-06-25 02:00:18 | INFO     | === Bewerbung Generation Started ===
2025-06-25 02:00:18 | INFO     | AI client: ClaudeAPIClient
2025-06-25 02:00:18 | INFO     | AI model: sonnet-3-5

Advanced Configuration

Fallback Chain Configuration

The default AI provider fallback chain is:

  1. Llama (if available and configured)

  2. Claude (if API key provided)

  3. Sample content (always available)

You can modify this in ai_client_factory.py.

Content Caching

Cache Location: .cache/ai_content_cache.json

Cache Behavior:

  • Caches AI-generated content by input hash

  • Survives between runs

  • Can be cleared manually

Disable Caching:

# In AI client initialization
use_cache = False

PDF Generation Settings

WeasyPrint Configuration:

Modify CSS styles in templates or add custom CSS files.

PDF Output Quality:

# Custom PDF settings
pdf_options = {
    'page-size': 'A4',
    'margin-top': '1in',
    'margin-bottom': '1in',
    'encoding': 'UTF-8'
}

Configuration Validation

Test Configuration:

# Validate all settings
make test-providers

# Test specific provider
AI_PROVIDER=claude make test-providers

Debug Configuration:

# Show current configuration
python -c "
import os
from src.ai_client_factory import AIClientFactory
f = AIClientFactory()
print('Current config:', f.get_config_summary())
"

Environment File Examples

Minimal Configuration (.env):

AI_PROVIDER=sample
OUTPUT_STRUCTURE=legacy

Full Configuration (.env):

# AI Provider Configuration
AI_PROVIDER=auto
ANTHROPIC_API_KEY=sk-ant-api03-your-key-here
LLAMA_MODEL=llama3.2:latest
OLLAMA_BASE_URL=http://localhost:11434

# Output Configuration
OUTPUT_STRUCTURE=by_model
INCLUDE_GENERATION_METADATA=true
GENERATE_DOCUMENTATION=true

# Testing Configuration
REGENERATION_TEST_MODE=content
REGENERATION_TEST_TIMEOUT=120

Development Configuration (.env.development):

AI_PROVIDER=sample  # Fast for development
OUTPUT_STRUCTURE=both  # Test both structures
INCLUDE_GENERATION_METADATA=true
GENERATE_DOCUMENTATION=true