Edyx

Free, uncapped AI APIs for everyone

Comment

Inspiration

The idea for Edyx was born out of a universal developer frustration: The Rate Limit. We've all been there. You're in the flow, hacking away at a new idea at 2 AM, and suddenly your API calls start failing. You've hit the free tier ceiling, or worse, you need to enter credit card details just to test a "Hello World" endpoint. We wanted to build the opposite of that. We wanted to create a frictionless sandbox a place where developers, students, and hackers could access powerful AI models without worrying about usage quotas or billing cycles. We asked ourselves: What if we could build an API gateway that just works, stays free, and gets out of your way?

What it does

Edyx is an open AI API gateway that provides free, uncapped access to Large Language Models. We offer three distinct endpoints tailored to different needs:

  • Fast Model: Optimized for speed and low latency (great for real-time apps).
  • Balanced Model: A middle-ground focused on reasoning and accuracy.
  • Conversational Model: Tuned for chat applications and maintaining context. Developers get a dashboard to generate API keys, track their token usage in real-time, and integrate these models into their apps with just a few lines of code.

⚙️ How we built it

We architected Edyx as a distributed system to handle high traffic while maintaining low latency.

The Stack

  • Frontend: Built with React and Vite for performance. We used Three.js (React Three Fiber) to create immersive 3D elements (like the interactive hero blob and floating code particles) that give the platform a premium, "developer-first" aesthetic.
  • Backend: A robust Node.js/Express server handles authentication, user management, and API key generation.
  • Database: We use Supabase (PostgreSQL) for reliable, relational data storage.
  • Edge Computing: This is our "secret sauce." We utilize Cloudflare Workers and KV (Key-Value) Storage for the API gateway.

The Architecture

When a user generates an API key, we don't just save it to the database. We asynchronously push that key to the Cloudflare edge network. $$ \text{User} \rightarrow \text{Backend} \rightarrow \text{Postgres (Persist)} \rightarrow \text{Cloudflare KV (Cache)} $$ When you make a request to an Edyx model, the gateway validates your key directly at the edge nearest to you—without needing to hit our central database. This reduces latency significantly.

Challenges we faced

1. Latency vs. Security Validating an API key usually requires a database lookup, which adds latency.

  • Solution: We implemented the Edge Sync pattern. By syncing valid keys to Cloudflare KV, we perform validation in milliseconds at the edge, ensuring the proxy adds negligible overhead to the AI request. 2. Visualizing "Invisible" Tech APIs are backend tools; they don't have a "look." We struggled to make the project visually exciting.
  • Solution: We focused heavily on Creative Coding. We built custom interactive components—like an Emoji that follows your cursor and reacts to your login input, and dynamic graphs that render usage data in real-time. We turned invisible data into a visual experience. 3. Abuse Prevention Offering "uncapped" APIs is risky.
  • Solution: We built a smart rate-limiting system that monitors usage patterns. While we don't cap legitimate volume, we can detect and throttle bot-like behavior to keep the service fair for everyone.

What we learned

Building Edyx taught us that User Experience (UX) isn't just for buttons and forms it's for APIs too. The "Developer Experience" (DX) of measuring how fast someone can go from signup to first successful API call is a critical metric. Technically, we deepened our understanding of Edge Computing and Distributed Systems. Synchronizing state between a central Postgres database and a globally distributed KV store was a complex challenge that taught us valuable lessons about data consistency and eventual consistency models. Ultimately, Edyx is more than just a proxy; it's a statement that powerful tools should be accessible to everyone who wants to build.

Built With

Share this project:

Updates