EventPlanner GPT

Inspiration

Event planning is super archaic and time-consuming. Our team wondered if there was a way for us to automate the most difficult part, finding venues + vendors that match a user's exact query (so they do not have to spend hours and hours google searching and calling vendors unnecessarily). And then we pushed our idea to see if we could make that query actionable, so could we use GPT to write an email or directly call the vendors that the user liked from our results.

We know that this technology has the potential to disrupt the entire event planning industry and we want to eventually become a complete end-to-end event planner using AI that uses gpt to automate targeted recommendations for venues + vendors, negotiating contracts, building budgets, developing schedules, and coordinating rsvps.

What it does

A user comes on and they enter their needs so for example "I am looking to host a dinner in San Francisco on 5/17 from 6-8PM for 50 people and my budget is $5,000. I need a venue, catering, and AV services for this event." Our system takes this query and evaluates it into subtasks, then we send Evenplanner GPT (agentGPT) out to fulfill the substasks. It critiques itself to make sure the options it returns is within the parameters the user put in and puts up constraints so it doesn't return more than 3 targeted results. It complies its list together and sends it to the user. The user then decides which vendors it would like to work with and our system sends inquiries out to them via email. All of this data gets stored in users profile so if the user wants to find more vendors or re-query our system, they can.

How we built it

We built using Langchain and Next.js. We built the ai model using langchain, which connects to our frontend on next.js and our dev storage is through supabase. We are using gpt3.5 and Langchain also allows us to create subtasks from users's prompts and then we use websearching to find the relevant vendors of their event (and within their constraints)

Challenges we ran into

setting up our chat server, going through + learning about langchain, rate limits on openai (credits not being applied), navigating thru lang-chaining to connect to our chat frontend and enabling realtime web searching.

Accomplishments that we're proud of

We built an agent that can take all the user's data and query in real-time and produce accurate results (we are super excited about this technology!) we are also happy that we were able to learn about Langchain and implement it into our technology. we are super happy that we have something to give to our users and produce results

What we learned

Langchain is hard to debug. OpenAI and Microsoft take a while to approve your account, so working around rate limits was a learning experience

What's next for EventPlanner GPT

We are rolling this out to our clients! We are currently beta testing with Microsoft, Youtube, Doorvest, Hubspot and a bunch of other startups in the space.

deck - https://drive.google.com/file/d/1l7wVxFLNuu5o2_xC1sQLqiEs4FrOBP1y/view?usp=share_link

loom link- https://www.loom.com/share/e9759be39f3345e8929ec0ae52864c68