This guide explains how to set up the project and run a test execution.

From the root of the repository (UDA-Bench):

cd systems/lotus

conda create -n lotus_env python=3.10.16
conda activate lotus_env

pip install -r requirements.txt

Use the .env.example file to create your personal .env file in the root of lotus

python main.py \
  --sql "<your_sql_query>" \
  [--limit <num_rows>] \
  [--cascade]

• --sql SQL query (o lista di query) da eseguire.

• --limit (optional) Limita il numero di righe del dataset da processare. Default: -1 (nessun limite).

• --cascade (optional) Abilita la strategia di LM cascade.

Results are available in the folder results

• Primary LLM: gemini/gemini-2.5-flash (Google Gemini 2.5 Flash)
• Cascade Mode: Uses MODEL_MINI (gemini/gemini-2.5-flash) and MODEL_PRO (gemini/gemini-3-flash-preview) for the cascade strategy