PromptHub

Discover and Share Ready-to-Use AI Prompts

Comment

Inspiration

Prompt engineering is an art, but it’s often hard to know what works where, especially with so many models and environments available. I wanted to create a space where developers and creators could share prompts that actually work, explore what others are using successfully, and build off each other’s experience.

What it does

PromptHub is a platform where users can discover, share, and contribute effective AI prompts. It helps developers and creators find the best prompts tailored to different models and tasks, improving prompt engineering workflows. The app supports browsing, filtering, and organizing prompts by categories, LLMs used, accuracy, and user feedback, making prompt discovery simple and practical.

How we built it

  1. Designed the frontend in React.js, with a clean UI for browsing and contributing prompts
  2. Used Supabase for database, auth, storage, and serverless functions
  3. Implemented full-text and semantic search using PostgreSQL + vector embeddings
  4. Created a form system to standardize metadata across different prompt submissions
  5. Deployed the app via Netlify

Challenges we ran into

  1. **High token consumption: **Bolt’s discussion and build modes helped reduce usage, but syncing with Supabase still burned through tokens quickly.
  2. **Repeated bug-fix loops: **Bolt occasionally suggested the same fix multiple times, even when it didn’t resolve the issue, leading to token waste. This was addressed by manually providing additional context when the error persisted, which helped maintain tighter control over debugging and intervention.

Accomplishments that we're proud of

  1. Watching Bolt transform a simple description into a fully functional MVP within minutes was genuinely impressive.
  2. The “Enhance Prompt” tool brought clarity and structure to even the vaguest user inputs.
  3. Core integrations were set up and running in no time, speeding up development significantly.
  4. The one-click deploy feature made pushing updates effortless and fast.

What we learned

  1. How to manage and optimize token usage effectively during iterative builds in Bolt, balancing automation with manual control.
  2. Leveraging Bolt’s discussion mode to boost fix accuracy and minimize repetitive or redundant corrections.
  3. The value of providing clear, structured context to AI tools for better prompt enhancement and debugging.

What's next for PromptHub

  1. **Connect models: **Integrate with live models to create a richer, more interactive chatbot experience to allow users to test and compare prompt outputs across multiple LLMs side-by-side for real-time insights.
  2. **User Profiles & Contributions: **Build a community where contributors get recognized, with upvotes, favorites, and threaded discussions.
  3. **Longform Prompt Chains: **Support multi-step prompt sequences designed for complex workflows like agent orchestration or retrieval-augmented generation.

Built With

Share this project:

Updates