GemInnovator

• 

  Generative AI Apps using Gemini

Inspiration

Main inspiration was the opportunity to help end users boost productivity through content processing, achieve automations by automatically managing interactions with Gemini and allowing end users build impactful Business Use cases.

What it does

GemInnovator provides an innovative Self-Service No-Code/Low-Code Gen-AI App building framework for content generation, synthesis, automated insights using Gemini API unlocking great values from data.

How we built it

We have built a Gen-AI Framework using javascript for all the citizen data scientists to democratize access to Gen-AI. The backend server of the App is built using Java. The goal is to provide a No-Code/Low-Code Self-Service User Experience where users can create Gemini Connection.

Users can install this platform Sparkflows Fire-Insights as mentioned here - https://docs.sparkflows.io/en/latest/installation/installation/index.html

Users can create the Gemini connection as mentioned here - https://docs.sparkflows.io/en/latest/installation/connection/gen-ai-connection/gemini.html

Users can build the Apps as explained here - https://docs.sparkflows.io/en/latest/user-guide/web-app/index.html.

The main strength of this App Builder Framework is that users can orchestrate data cleaning, wrangling, processing, analytics, nlp, gen-ai using different stages by leveraging no-code/low-code workflows. The workflows use a large no of automated processors which can run in serverless dataproc.

Challenges we ran into

The key challenge is to provide a smooth user experience for multi-modal data handling involving video and image analysis. Another challenge is to handle the latency using innovative approaches.

Accomplishments that we're proud of

We are really proud of offering the App Building framework which can abstract out calls to Gemini API and help users build intuitive UI using Drag-n-Drop UI widgets which can allow users capture prompts and show response.

What we learned

We learnt about various capabilities of Gemini App and orchestrating results from Gemini with insights from business domain data. We also learnt about integrating multiple GCP services like Serverless Dataproc, BigQuery and Gemini.

What's next for GemInnovator

We want to integrate AlloyDB and BigQuery as Vector Database along with Gemini. We want to build agents using Vertex-ai and Gemini using API. We want to enhance the Conversational App UI.

Built With

• java
• javascript