AIr Chat

Realtime Chat Application With Generative AI For Message Suggestions. Improving Conversation Between Two Humans. Best Used In Professional Settings For Working Class Market. Powered By Google Gemini.

Comment

Inspiration

  1. Engaging in live discussions with friends during project work often proved chaotic, as we navigated between technical concepts, errors, and solutions in chat application and browser. How can we streamline access to answers without toggling between apps?
  2. Ever found yourself in a crucial conversation with someone important and wished for a seamless synchronous exchange? How can you facilitate this?
  3. When on the go, responding to messages seeking information you're unsure of or have partial knowledge about can be challenging. What solution can address this dilemma?
  4. Seeking to maintain an interactive and lively online conversation but unsure of how to do so? Let's explore the features our project offers to tackle these scenarios.

What it does

AIr is a chat application enhanced with generative AI, designed to seamlessly assist in online conversations among humans. By analyzing previous messages within the conversation, it suggests responses to keep the dialogue flowing smoothly. Unsure of how to reply to a message? Simply click on the AI button to receive four suggested responses to choose from.

Additionally, AIr offers other notable features:

  1. AI Assistant: Engage in conversations with an AI assistant.
  2. Image Analyzer: Provides explanations and more for images shared within the chat.
  3. Sentiment Analyzer: Analyzes sentiments such as mood, tone, and coherence within the conversation. Concerned about privacy? We'll address that shortly.

How we built it

The project is built using Django for the server side operations and React.js for front-end development. As per the generative AI part we are using Google Gemini AI.

  1. For text generation we are using gemini-1.5-pro-latest
  2. For Image analyzer we are using gemini-pro-vision

Challenges we ran into

The process of building this project was a roller coaster for us with lots of learning, mistakes and corrections.

Some challenges we ran into:

  1. Initially we used some other AI model that we were comfortable with working and had knowledge about it as we used it in our previous projects. Replacing it with Gemini was a challenging
  2. Integrating Gemini Pro Vision for image analyzer, also storing images temporarily for that purpose
  3. In the sentiment analyzer some negative messages were not accepted by Gemini so we had to handle that. Challenges that initially were hurdles led to the new features and improvements

Accomplishments that we're proud of

Building this project from scratch with just an idea to a working model was an achievement that couldn't have been possible without this hackathon. Our project demo is exactly what we intended to do, so this thing is very important for us.

What's next for AIr Chat

So you were concerned about privacy, because the messages are being sent to AI model which is hosted by companies of that model.
So we have a solution, we will use Gemini Nano which will run on the user's device therefore the messages are with user. This might help in privacy. This project can be used as a standalone app or can be integrated in applications and services that are used for communication like Slack, LinkedIn, Gmail and more.

Share this project:

Updates