Rho | Devpost

Inspiration

Some of our team members have used comparison programs in order to find the plan flight or hotel booking to make the ideal purchase in a sea of options. Making a purchase by comparing as many options as possible while also doing so effectively can be tricky; that's why there are programs to do it for us. We felt that with the ever-growing AI environment, it's difficult to determine which model to use. There are so many factors to consider, such as cost, performance, scalability, etc. Sifting through docs and forums without a clear view/goal or understanding of a model's version history can lead to wasted time. We believed that we could provide a solution that is better than just asking ChatGPT to decide for us, an approach driven by data in order to compare various APIs.

What it does

An API comparison platform that helps developers benchmark AI APIs across cost, latency, accuracy, and scalability, while using an AI component to analyze each API's history and evolution. Developers can view trends like pricing changes, model improvements, and downtime frequency to choose APIs based on both current performance and long-term reliability. Find the perfect APIs for their projects. Built with a focus on beautiful design, powerful filtering, and data-driven decisions.

How we built it

Chroma for semantic understanding: ingests docs/features/tags and powers natural‑language search. Elasticsearch for live ops: indexes time‑series (response time, uptime, error rate), aggregates per API. Node/Express backend: merges Chroma matches with Elastic metrics and computes combined scoring. Minimal agent: periodic sampler that writes performance samples into Elastic. Frontend (one‑pager): fast static client hitting the backend, polished UI with a premium teal theme.

Challenges we ran into

One of our biggest challenges was actually brainstorming ideas for our project. We had several ideas; however, many of the ideas we had covered topics we weren't well-informed on. Additionally, we had to deal with the WiFi issue at the venue. This led to spending time commuting to areas with WiFi instead of programming. The WiFi issue stalled a lot of progress we could have made. Through the process of making the program there were issues the following:

Scope triage: balancing beautiful UX with robust data plumbing on a tight timeline.
Data variety: normalizing pricing/feature vocab across providers; adding synonyms to improve recall.
Venue Wi‑Fi: intermittent connectivity slowed installs and indexing; we built offline fallbacks.

Accomplishments that we're proud of

Clean, explainable ranking: semantic “why” + live metrics, not a black box.
Crisp, professional UI that makes complex trade‑offs obvious at a glance.
Pluggable pipeline: easy to add new providers (Hume, Composio, Replicate, Stability, Shopify, Klarna, etc.).

What we learned

Mentors brought up “what if” questions, which challenged us to pause and think more thoroughly.
Consequentially helping us build in the long run.
Troubleshooting is 100x easier when done in collaboration. People
Decision quality improves when you show both “fit” and “operability” together.
Lightweight agents + time‑series aggregations are enough to unlock high‑value comparisons.
Synonym/field‑weighted search drastically improves relevance without heavyweight infrastructure.

What's next for Rho

Real embeddings pipeline (provider docs scraping + scheduled refresh) and broader provider coverage.
Deeper cost modeling (usage‑based estimators) and scenario‑based scoring (e.g., batch vs real‑time).
Alerting and drift detection (when an API degrades or pricing changes).
Team features: saved comparisons, shared notes, and procurement‑ready reports.