Conversational Market Expert
Voice-to-Voice AI Agent with RAG & Real-Time Knowledge Access
180
HOURS
200
TESTS
1
GOAL
Core Features
Voice-to-Voice
Natural conversation with gemini-live-2.5-flash-preview

RAG
Pipeline
Intelligent knowledge access to Gopf's vector database with Qdrant
Evolution of Intelligent
Knowledge Access
■□□□□□ Static base with default report
■■□□□□ Dynamic base with URL Input
■■■■■□ RAG and Qdrant for db import
■■■■■■ Google search tool (by Gopf)

Live Processing
Event-based interaction with Gemini Live API

Production Integration
Docker containerized & deployed on Gopf platform

Architecture
Progress, not Perfection
The leading principle that helped me moving along from a blind start through a chaotic progress, soon through object oriented code, until the goal was reached and the market expert was ready for deployment on the Gopf platform.
FAQ
Explain the Poster in other words
The Poster visualises the Progress. As the preview model was not documented well at that time, I had to make things out by doing: trying, testing, debugging, changing, moving along, rereading the docs, rechecking for code examples, discussing setup, tools and code with my supervisor Marc Bravin and with the Gopf team ongoing.
What means the WHILE on the Poster?
The main bug was a for loop which had to be replaced by a while true
What does Gopf do?
Gopf AG is a market whisperer platform for companies of all branches but focused on manufactory companies Switzerland. Check out & contact Gopf
by Chiara Steindl
Linkedin
Made with