Inspiration
In Kenya, micro-businesses and students run entirely on WhatsApp. However, the official WhatsApp Business API (Twilio/Meta) charges per-conversation fees that are economically unviable for the informal sector. I wanted to democratize access to Google's Gemini AI by building a zero-cost bridge that works on any standard WhatsApp number.
What it does
Kigo acts as a self-hosted proxy. It allows any user to text a standard WhatsApp number and interact directly with the Gemini model. Whether it's a student needing coding help or a vendor needing text translated, Kigo delivers Gemini's reasoning capabilities directly into their chat without costing the host a cent in messaging fees.
How we built it
I ignored standard API providers. Instead, I engineered a custom backend using Node.js, Puppeteer, and the Baileys library to emulate a WhatsApp Web client. This handles the raw WebSocket connection directly.
When a message is received, the Node.js server extracts the payload and routes the text directly to the Gemini API. Gemini processes the prompt, generates the response, and my backend pushes it back through the Baileys socket to the user in real-time.
Challenges we ran into
Managing the WebSocket state and handling session disconnects without crashing the Node.js server. I had to build custom retry logic to ensure the Puppeteer instance remained stable.
Accomplishments that we're proud of
Successfully building a bypass architecture that reduces API messaging overhead to $0, while bringing state-of-the-art Gemini AI to users who don't have the bandwidth to run heavy web apps.
Built With
- gemini-api
- google-sheets
- javascript
- node.js
- puppeteer
- websockets
Log in or sign up for Devpost to join the conversation.