This is a submission for the Cloudflare AI Challenge.
What I Built
I have built a chatbot embed for web apps that can be deployed to cloudflare stack. It needs more work currently it's not complete. What I created is similar to sitegpt. Where users visits trains there chatbots and then they can embed it on there websites using an iframe. I wanted to do this for whatsapp but after a bit research they asked for verification and other things.
Demo
Web Link: Botified.chat
My Code
Github: Botified Chat
This time I choose to use complete cloudflare stack with sveltekit and with there recent introduction of platformProxy, I like it somewhat.
Tech Stack:
- Metaframework: Sveltekit
- Database: Cloudflare D1
- ORM: Drizzle
- Vector Database: Vectorize
- Deployment: Cloudflare Workers
Journey
I started very late around a day ago but i did had this idea what i wanna build. So, I took my weapons and sat for 10 hours and completed as much as possible as i mentioned earlier it's not completed. It doesn't matter it gets rejected or anything. I learnt a lot in last 10 hrs about langchain, cloudflares workers issues, D1 migration handling with prisma and Drizzle. It's was fun to sit foe this long to do my own project. Sometimes, we just do the our 9-5 and forget the chances we get and this was my chance. I'll complete my project now.
Multiple Models and/or Triple Task Types
I have used 2 models:
- Embedding: @cf/baai/bge-large-en-v1.5
- Chat: @cf/meta/llama-2-7b-chat-int8 They serve there purpose well good enough. I'll be making changes in the project. I don't know it's allowed or not.
Thanks @cloudflaredev and Dev.to for this fun project.