Gemini-API-Developer-Competition

organized by Google

Overview

Gemini is a family of generative AI models developed by Google DeepMind that is designed for multimodal use cases. The Gemini API gives you access to the Gemini Pro Vision and Gemini Pro models.

Gemini 1.5 Pro is a next-generation AI model that delivers a breakthrough in long context understanding. It can process upto 1M tokens across different modalities, enabling richer and more complex interactions.

Files API is an easy way for developers to upload files for multimodel use cases. It’s the most stable and least error-prone method of unlocking multimodel support in your applications using Gemini.

Known Limitations:

Maximum request size: The Gemini 1.5 Pro API currently has a maximum request size of 20MB. Use the File API for requests larger than 20MB.

TECHNICAL DETAILS ABOUT THE IMPLEMENTATION

Gemini 1.5 Pro is a next-generation AI model that delivers a breakthrough in long context understanding. It can process upto 1M token window across different modalities, enabling richer and more complex interactions. To put things into Prespective, one can use it to:

extract insights from 30k+ lines of code
1 hour video
extract information from 11 hours of audio files
even from a document with 700k words

Known Limitations:

Maximum request size: The Gemini 1.5 Pro API currently has a maximum request size of 20MB. Use the File API for requests larger than 20MB.

Tech Stack Components

Data Framework - LlamaIndex for extended RAG(Retrieval Augmented Generation) functionality Frontend Framework - Streamlit
LLM(Large Language Model) - Gemini 1.5 Pro language model from Google Programming Language - Python
Vector Database - Qdrant

# Creates a virtual environment called venv
python -m venv google-ai

# Activate the virtual environment
source google-ai/bin/activate

ashish-soni08 / gemini-api-developer-competition Goto Github PK