Weavel

Weavel automates prompt engineering, delivering the best prompts 50x faster than humans. Simply input your prompt and receive optimized prompts with highest accuracy. Boost your prompt's accuracy by an average 20% in less than 5 minutes.@mwseibel thank you for the hunt!

Hi Product Hunt makers 👋

I'm Andrew, co-founder of Weavel. I'm excited to introduce our prompt engineering automation platform for LLM apps @junpark_314 @hyunjiejung @tobykim .

As foundation models get smarter, more and more tasks can be solved by prompting LLMs in the right way. But going from 80% to 100% is hard and requires days, even weeks of prompt engineering. We are building Weavel to help makers ship fast and with confidence, by automating the hassle of manual trial-and-error of prompt engineering with LLM-based algorithms.

Weavel is designed for developers, data scientists, and AI enthusiasts who need to maximize efficiency and accuracy in their LLM apps. Weavel delivers the prompts with highest accuracy, 50x times faster than humans - optimized prompts surpasses DSPy-optimized prompts by 4% and CoT by 7%.

We are rethinking prompt engineering to make the process fast, accurate, and reliable. Here's how we help the engineering cycle:

💾 Dataset curation:

• Add one line of code to log LLM calls in your product. We automatically curate datasets from the logs - user representative dataset and out-of-distribution dataset for edge cases.
• Enrich your dataset with LLM call logs from production.

🧪 Prompt optimization:

• From your base prompt and provided dataset, Ape (our AI prompt engineer) will iterate on diverse prompts, and find the mixture of instructions and few shot examples with the highest scores.
• It takes only 4 minutes, and improves accuracy by an average 20% 🚀

We'd love to hear your thoughts and feedback. Thanks for checking out Weavel!

PS1: The code that powers Ape (AI prompt engineer) is open source at https://github.com/weavel-ai/Ape

PS2: We are working on a prompt engineering playground that also automatically curates datasets from the runs - will launch soon! Real soon!!🚀🚀🚀@mwseibel @junpark_314 @hyunjiejung @tobykim @aschung01 Hi Andrew! Love the vision behind Weavel! 🚀 Automating prompt engineering is a smart move to boost efficiency. The dataset curation feature sounds particularly valuable for developers. Excited to see how it evolves!That's great! Congratulations! Many solutions aim to optimize prompts for assistant, RAG, and other typical use cases, but fewer offer help with optimizing AI Agent instructions and prompts. Do you have any opportunities to use it for the Agent?Congratulations on the launch! This definitely is promising! QQ: Do you store prompt data internally - was just checking for compliance. All the best!@anchal_r Thank you Anchal!! Yes we do store prompts on our cloud, but with the OSS library you can keep all data to yourself.I appreciate the emphasis on automating evaluations and integration with existing tools. It makes the process more streamlined and less manual. However, I wonder about the learning curve for new users who might not be familiar with this kind of setup.