Prompt Party

The first collaborative AI workspace.

Comment

Inspiration

We've all been there: staring at an LLM chat, wishing we could easily share our progress or collaborate with a teammate. As founders of small, stealth startups, our team constantly uses AI research tools, but the glaring absence of intuitive collaboration features was a recurring frustration. We wanted to bridge this gap. Drawing inspiration from the real-time document editing power of Google Drive, we envisioned a similar experience for LLM chats – a seamless, shared workspace where AI-powered research could truly become a team effort. This collective frustration sparked the idea that became Prompt Party.

What it does

Prompt Party is our answer to this need. It's a revolutionary platform that empowers multiple users to create a shared workspace for AI chats. Within this collaborative environment, everyone can prompt, edit, and converse about project additions in real-time. Prompt Party brings true, instantaneous collaboration to your AI-powered research and development, transforming individual AI chats into powerful team endeavors.

How we built it

Prompt Party was built with a focus on real-time collaboration leveraging React and Supabase. The application features user authentication, chat functionalities, and integration with various LLM models (Gemini, OpenRouter) for AI-powered conversations. We utilized Supabase for database management, authentication, and real-time subscriptions, ensuring a robust and scalable backend.

Challenges we ran into

  • Gemini API Key Integration: Getting the Gemini API key to work correctly was a significant hurdle, requiring multiple attempts and persistent debugging.
  • Real-time Messaging Reliability: Syncing messages across multiple users reliably proved tricky, with race conditions and authentication delays causing inconsistent updates.
  • Shared Link Multi-User Access: Enabling seamless multi-user access via a single link required creative workarounds due to Bolt's limited native support for anonymous sessions.
  • RevenueCat Payment System Integration: This remained unresolved during the hackathon, plagued by user ID errors, Supabase session timing issues, race conditions, and restrictive RLS policies.
  • Supabase Quirks in Bolt: We frequently observed delays in authentication and database updates, often necessitating manual refreshes and workarounds for RLS policies.

Accomplishments that we're proud of

Functional Collaborative Chat: Our biggest accomplishment is successfully building a robust, real-time collaborative interface. Despite the inherent platform limitations, Prompt Party effectively syncs across users, enabling shared conversations and truly collaborative AI interactions.

What we learned

The Power of Prompting: We quickly learned that Bolt heavily relies on clear, intentional prompting. The quality of our prompts directly correlated with the effectiveness of the outcomes.

Pre-Implementation Discussion is Key: Utilizing Bolt's "Discuss" button to outline our plans before diving into implementation proved incredibly beneficial. This organized our thoughts, saved significant development time, and helped us avoid early-stage errors.

What's next for Prompt Party

Our vision for Prompt Party extends far beyond this hackathon. We're aiming to:

  • Integrate More LLMs: Expand support beyond Gemini to include a wider range of large language models, giving users more flexibility.
  • Become an "All-in-One" AI Hub: Our ultimate goal is to create a unified platform, similar to OpenRouter, where users can seamlessly switch between various LLMs (e.g., GPT, Claude, Gemini).
  • Refine Payment System: Once Supabase issues are resolved, we'll properly integrate RevenueCat for robust subscription and token usage management.
  • Implement Advanced Multi-User Editing: We envision adding collaborative editing features, like a discussion box similar to Gemini Canvas, allowing multiple users to simultaneously edit documents, leave comments, and suggest improvements in real-time.

Built With

Share this project:

Updates