ari-llm

Simple local scripts to:

• read prompt .txt files from prompts/
• run them with either Hugging Face or Ollama
• append results to a CSV
• print saved results in the terminal

Files

• script.py: main runner (generation + CSV logging)
• results.py: prints prompt_file and output from CSV
• requirements.txt: Python dependencies
• prompts/: input prompt text files

Install

pip install -r requirements.txt

Configure

Edit the variables at the top of script.py:

• PROVIDER: "hf" or "ollama"
• MODEL_ID: model name for the selected provider
• USE_STRUCTURED_OUTPUT: True/False (used by Ollama path)
• PROMPTS_DIR: input directory (default prompts)
• OUTPUT_CSV: output file (default ari_llm_outputs.csv)
• MAX_NEW_TOKENS: generation budget
• TEMPERATURE: sampling temperature (0.0 is forced automatically for structured Ollama)
• OLLAMA_URL: Ollama server URL

Run generation

python script.py

View results

python results.py