Auto-optimize your prompts using local LLMs — with scoring, ranking, and feedback
PromptTuner helps you refine your prompts by:
- Generating multiple variations of a given base prompt
- Sending them to a local LLM (e.g., TinyLLaMA or Phi via Ollama or LM Studio)
- Collecting outputs and scoring them against a reference output
- Ranking the most effective prompt variants
- Python 3.10+
- Ollama or LM Studio
- Local LLMs: TinyLLaMA / Phi / Mistral
- Sentence Transformers (for scoring)
- Streamlit (coming soon)
- Clone the repo
- Set up virtualenv
- Install dependencies: pip install -r requirements.txt
- Start: python main.py
MIT - Free to use, improve, and contribute.
PRs welcome! Mode models, scoring metrics, and UI coming soon.