Inspiration
From the Viral Caveman Speech https://github.com/juliusbrussee/caveman i've decided to do it better in the futuristic way the prompts optimized by precision and structure not reduction to primal sentences!
What it does
Spaceman Speech compresses verbose LLM prompts using local NLP — stripping filler, restructuring into clean ROLE/TASK/OUTPUT format. 30-50% token savings, same meaning. No second LLM needed.
How we built it
Its Python backend fastapi and simple html, js frontend.
Challenges we ran into
NLP and Lemmatization and Cosine Similarity is difficult!
Accomplishments that we're proud of
Full working webapp for prompt optimization. The app vibe is Windows 95 design inspired. An window on classic pixel art background.
What we learned
Classical designs are so funny! We feel like hackers form 90's.
What's next for Spaceman Speech
We will open source it right after judgment! MIT license for free to everyone. We deserved prompts form cosmos!
Log in or sign up for Devpost to join the conversation.