config.py

# Ollama Configuration
OLLAMA_MODEL = "gemma3:4b"  # Model to use (e.g., "llama2", "mistral", "gemma3:1b")
OLLAMA_BASE_URL = "http://localhost:11434"  # Default Ollama URL
OLLAMA_THINK_MODE = False  # Enable thinking mode if supported by model
OLLAMA_CONTEXT = 4096  # Context window size
OLLAMA_KEEP_ALIVE = 30  # Keep model loaded for N minutes

# File Paths
PROMPT_FILE = "prompt.txt"  # Template prompt with {INPUT} placeholder
INPUT_FILE = "input.txt"  # Input file with one item per line
OUTPUT_FILE = "output.jsonl"  # Output file in JSONL format
ERROR_FILE = "errors.log"  # Error log file

# Performance Settings
# Benchmark results: gemma3:1b achieves 2.65 items/s with 3 workers (100% efficiency)
# Benchmark results: gemma3:4b achieves 0.85 items/s with 3 workers (3x slower but more accurate)
PARALLEL_WORKERS = 3  # Benchmark test
REQUEST_TIMEOUT = 120  # Maximum seconds per request (not currently enforced)