WhichWhip

• 

  Home

• 

  Eval Flow 1

• 

  Eval Flow 2

• 

  Eval Flow 3

• 

  Eval Flow 4

• 

  Past Evals

• 

  Preferences

Inspiration

My dad has been looking to buy a car for weeks now. He usually sends me a facebook marketplace post and asks for my opinion on it. We usually debate on how reliable the car all in a messy whatsapp thread. I'd rather have a place we can look through all the cars we have looked at organized by what we are looking for in a car

What it does

It takes a paste of a car listing's details and extracts the information. It uses this to do further research about what people say about the car and how reliable it is. After this it weighs what is has found against things people usually consider when buying a car like MPG, milage, year, etc. With this is generates a ranking and recommendation based on the user's preferences.

How we built it

Framework

NextJs - easiest to setup to coordinate front/backend

Zod - For memory management

AI

Gemini - to recognize fields from a post description

OpenAI - Further research of the car's details

Jev - to classify the research into a rating

Hosting

Heroku

Challenges we ran into

Gemini rate limiting for reasearch

OpenAI burning exhausting tokens on research (had to by credits)

Heroku - Deployment configs so it wouldn't 503

Could not get the whichwhip.tech domain to route to the Render hosting

Accomplishments that we're proud of

My dad was able to use it this morning and liked it

What we learned

Zod - for local memory management. Fast because it is local but structured poortly.

What's next for WhichWhip

Making state persist, requiring a auth and session management. Better function to judge car on, example: electric, insurance fees, etc.

Built With

• codex
• gemini
• heroku
• jev
• openai