Hacktoberfest Weekend Challenge: Build for a Friend Submission 🤝
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
OmniCall is for my friend Daniel, who teaches evening classes. Parents message during the lesson about fees, a broken class link, or a child leaving the class. Reading those messages means stopping the lesson.
The teacher pastes what the parent wrote and presses Hear the reply. Gemma decides the topic and how urgent it is. ElevenLabs speaks a short reply they can send back. MongoDB Atlas keeps the call for after class.
Demo
Watch the walkthrough
One run used this parent message: “The class link is not working and my child cannot join.”
The page said: “This is about the class link. high priority.”
The voice said: “I’m sorry the link isn’t working. Please try refreshing your browser or check your email for a new link.”
That trace shows model gemma-4-26b-a4b-it, 1 model call, 232 tokens, 0 errors, and a cost under $0.01. The tools on the trace are save_support_log and elevenlabs_tts.
Code
https://github.com/shuchi111/omnicall-AI
How I Built It
The open model is Gemma, gemma-4-26b-a4b-it, served through Google’s Gemma API. FastAPI sends the parent message to Gemma and asks for intent, priority, and two sentences written to be spoken.
Atlas stores each call. The page lists the calls from that point on. ElevenLabs turns the sentences into audio, and the browser plays them. Sentry records the run as one agent trace: the Gemma chat, the database write, and the speech.
Why Does Open Innovation Matter?
Gemma makes the judgment. The teacher does not download a multi-gigabyte model, and the message is not sent to a closed chat product just to decide whether it is about fees or a class link. Changing the model is a name in the request, not a rewrite. The trace shows the exact model, the token count, and the cost of that decision.
ElevenLabs is only the voice. Atlas is only the memory.
Prize Categories
- Best Use of Gemma
- Best Use of ElevenLabs
- Best Use of MongoDB Atlas
- Best Use of Sentry Agent Tracing