Hacker News new | ask | show | jobs
Show HN: PromptMage – Simplify and Manage Your LLM Workflows (promptmage.io)
10 points by oainstaller 659 days ago
Hello HN community,

I'm excited to share a project we've been working on called PromptMage, an open-source tool designed to simplify the process of creating and managing complex workflows for large language models (LLMs).

Backstory

As developers and researchers working with LLMs, we found ourselves constantly grappling with the challenges of prompt iteration, testing, and version control. Existing tools didn't quite fit our needs for an integrated, intuitive workflow management solution. So, we decided to build one. PromptMage was born out of a desire to streamline these processes and make LLM technology more accessible to teams of all sizes.

What is PromptMage?

PromptMage is a self-hosted python tool that provides a user-friendly interface for developing, testing, and managing LLM prompts. It offers several key features:

- Prompt Playground: A space to rapidly test, compare, and refine prompts.

- Integrated Version Control: Track changes and collaborate on prompt development with ease.

- Auto-generated API: A FastAPI-powered API is automatically created for seamless integration and deployment.

- Evaluation Mode: Conduct both manual and automated testing to ensure prompt reliability before deployment.

Why PromptMage?

We built PromptMage to fill a gap we felt keenly in our own work—an all-in-one solution that integrates prompt testing and version control directly into the workflow. Unlike other tools, PromptMage is designed with developers in mind, offering a straightforward setup and intuitive interface to foster collaboration and iteration.

How to Try It Out

PromptMage is currently in its alpha state and is under active development. We encourage you to install it, set it up (which should take about 5 minutes), and give it a spin. We’re eager for feedback from the community to help shape its future. You can deploy it locally or on your server.

Check it out here: <https://promptmage.io/getting-started/>

Don't forget to leave a star on github: <https://github.com/tsterbak/promptmage>

What’s Next?

We have a roadmap full of exciting features and enhancements. We're aiming to expand PromptMage’s capabilities to better support developers, researchers, and organizations as they navigate the rapidly evolving landscape of AI and LLM technology.

We’d love to hear your thoughts and feedback! Whether it's bug reports, feature requests, or contributions, all are welcome. Let's work together to make PromptMage a valuable tool for the community.

Feel free to ask any questions or share your experiences in the comments below!

Thanks :)

3 comments

This seems interesting. Most of the times I test prompts by directly calling the API and making the changes. This can be interesting if we can choose LLM endpoints. I would love to contribute to this repository as open source contribution.
Contributions would be welcome! :) Give it a try and see if it fits your needs.
That's cool. I tried to use Dify and Coze to manage my LLM-flows. But these are too complicated for me. And had bugs.
I would love to get to know your usecases and things that might be missing :)
Is it mostly for people to embed within their code?

I’m looking for something to manage my prompt collections and how well I find them to work with different models - but I mostly use existing software for inference (e.g. Ollama, Open-WebUI, BigAGI, Bolt).

This rules out most of the tooling I’ve found as it seems most are focused on “APM” for LLMs - rather than a prompt library + playground solution.

Hey, thanks for you interest in the project :)

It includes a prompt library and a playground and works with every model backend you set up (I tested a lot with Ollama). You can see screenshots here: https://promptmage.io/getting-started/#flow-playground

Promptmage is a python package meant to support building complex, multi-step llm workflows. Have a look at the examples here to get an idea: https://github.com/tsterbak/promptmage/tree/main/examples