AcBud: Applied Prompt Compression

AcBud is a practical AI companion that supports you in reaching your health and wellness goals. Our solution evaluates prompt summary, the truncation of context without losing semantic information.

Comment

About the Accountability Buddy Project

Inspiration

The idea for the Accountability Buddy project came from a desire to build a practical tool to help people stay motivated and achieve their wellness goals. Inspired specifically by the WOOP (Wish, Outcome, Obstacle, Plan) framework discussed in Yale's "The Science of Wellbeing" course on Coursera, we recognized the importance of structured goal-setting and proactive follow-through strategies. We wanted to create an accessible AI tool that would support users throughout their personal development journeys.

What We Learned

Through this project, we deepened our understanding of goal-setting psychology, especially how structured frameworks like WOOP significantly increase the likelihood of achieving personal objectives. We also explored advanced techniques in prompt engineering, learned how to effectively compress and summarize AI-generated content without losing crucial information, and gained valuable experience in creating intuitive conversational interfaces using Gradio. This compression should prove quite valuable for future applications, where the frequent reuse of prompts necessitates such optimization.

How We Built It

We leveraged Anthropic's Claude API to power the conversational aspects of the Accountability Buddy, using carefully designed prompts to guide the model's responses. To handle scenarios where user conversations became lengthy and verbose, we implemented a specialized summarizer layer capable of condensing extensive context into concise yet comprehensive summaries, ensuring critical details remained intact. We built a clean, user-friendly frontend interface using Gradio, emphasizing seamless interaction without overwhelming users.

Built With

Share this project:

Updates