Challenges

Integration of parts

• The biggest problem was to integrate the two parts we made unstructed and structed were made different and many parts were different like Ollama vs Gemini api we decided to go with gemini

Dataset searching

• No dataset was found on kaggle where we could create a dense graph *It was important since we were going to do bfs traversal

Multiple embeaddings getting fetched from FAISS

• For one single query many entries were getting retrived which was a problem since we were doing bfs traversal also and that would result in too much time