MyCase.AI

Something bad happen? Ask MyCase.AI if you have a viable legal claim against your violators.

Comment

Inspiration

Spending time talking to an attorney when you don't have a case is time-consuming and can be expensive. Even if the initial consultation is free, the attorney will not provide "legal advice" or do research about the case until you retain them. We want legal information to be accessible and affordable through a B2C application. The everyday person can now consult with MyCase.AI about something bad or harmful that happened to them and learn the same case law their lawyer would know concerning their incident.

What it does

MyCase.AI analyzes the facts of your case and tells you if there is relevant case law that supports your filing of a lawsuit. MyCase.AI will also search the web for the news of related lawsuits to your case. If you do have a case and you want to file a lawsuit, MyCase.AI will recommend an attorney to contact to file a claim.

How we built it

LlamaIndex is the framework we use for our context augmented AI application using agents and workflows. Chunks of a data set of U.S. federal case law is stored in a vector database, Pinecone. We use Mistral AI for embedding the chunks of data stored in Pinecone.

Agentic RAG is used to add context to the LLMs (Llama 3.1 and OpenAI's GPT-4o) for the function calling feature. The following agents are used: 1) Orchestrator Agent who decides which agent to pick to answer the user's inquiry; 2) RAG Search Agent who finds similar legal cases from the data set we are using (Llama 3.1 as the LLM); and 3) Web Search Agent who searches the web for the latest news on legal cases and to suggest attorneys who would help the user with their situation (GPT-4o as the LLM).

Vessal.AI is used for deploying our agentic workflow.

Challenges we ran into

While working on the backend, we were using the HuggingFace Llama model as the LLM but it does not support function calling so we switched to OpenAI's GPT-4o LLM. During deployment, we were unable to upload the docker image to deploy the entire project, which is why we do not have a live link to demo the project.

Accomplishments that we're proud of

We know a lot more about developing with LlamaIndex, which will be useful for future development projects.

What we learned

Using LlamaIndex as the framework to build a context augmented AI application using agents and workflows.

What's next for MyCase.AI

Deployment of the entire project.

Built With

  • gpt-4o
  • llama3.1
  • llamaindex
  • mistralai
  • openai
  • pinecone
  • reflex
  • vessalai
Share this project:

Updates