Promptterfly

Promptterfly is your smart, easy-to-use, local prompt manager that automatically optimizes and versions your prompts.

Quick Start

./setup.sh
./start.sh

Inside the terminal UI, type:

model add

Follow the simple interactive prompts to select a provider (OpenAI, Anthropic, Ollama, etc.) and enter your API key.

prompt create

Promptterfly will guide you. Use {variables} in your templates.

If you have a vague prompt and want an LLM to automatically generate a highly optimized template for you, run:

auto generate "I want a prompt that summarizes text into bullet points"

If you already have a prompt created and a dataset with a few examples (.promptterfly/dataset.jsonl), run:

optimize improve <id>

Promptterfly will use DSPy behind the scenes to magically find the best few-shot examples and create a new version of your prompt.

Test your prompt with variables using a JSON file:

prompt render <id> vars.json

Auto-prompt naming: Lazily create prompts and the LLM will name them for you!
Auto-improve generation: Ask for a prompt and a memory-less sub-agent will build, evaluate, and save the best one!
Budget Planning: Set maximum cost/token budgets so your optimizations stay cheap.
Local Versioning: No Git required! Every change creates a snapshot you can rollback to using version history and version restore.

Type help in the terminal for all commands. Enjoy!