QueryPages

• 

  Running Instance of our LLM

Inspiration

Navigating multiple organizations as a student on campus, or a volunteer or an employee is difficult. And for some overwhelming. To ease this process, I introduce you to Query Pages - similar to the Yellow pages but for your organization.... and no it isn't something you will have to carry.

What it does

The Query Pages framework allows for the conversing with organizational data providing accurate responses. It lets you train the well-known llama2 model on your organization's data and policies on secure company system.

How we built it

Stage 1 - Ensure the model can read the various document types - pdf, docx, json and csv. Parse the data into chunks and then vectorize them using Ollama. The values are then stored in a chromadb embedded database.

Stage 2 - Converting the query to vectors and then comparing and retrieving using RAG.

Stage 3 - Integration of Taipy with the framework

Challenges we ran into

Required 8 hours to ensure the working and configuration of the llama2 - Ollama model.

Accomplishments that we're proud of

Able to debug code and didn't lose hope.

What we learned

We learned about idea creation, tabnine and github copilot for development.

What's next for QueryPages

Implementation with more types of files. Usage of better llms and fine-tuning models. Improvement in UI with more tech stacks.

Built With

• chromadb
• langchain
• llama2
• python
• taipy