I’m currently a university student and my latest project aims to develop an LLM specifically designed to assist students in navigating our university’s course catalog. The goal is to create a tool that is not only cost-effective but also straightforward to maintain and implement, while ensuring it delivers accurate and helpful information.
At present, I’m exploring RAG. I’m very much open to suggestions and insights from the community regarding the best tech stack and architecture to use.
So far, I’ve been considering various tools and frameworks like Flowise AI, Langchain, Pinecone, Colbert. I’m even thinking about using numpy to convert vector embeddings to store in memory every time. If anyone has experience or suggestions about integrating these tools, or if there are other options I should look into, I would greatly appreciate your input.
I’m excited to hear your thoughts and recommendations. This project is an opportunity to create something that could genuinely enhance the educational experience for students like me.
I have fond impression with LangChain ReAct agent tools and its RAG tech stack. However, I do not have experience to speak about it. So you may take a bit research into its documentation. If you are comfortable with reading source code, that would be even better:
Though I am experience with Llama Index, Deep Lake Vector Database, and Open AI Agents functional calling tools. I am happy to help, please send me direct message, we can discuss some strategies to implement this project of yours.
Thank you for sharing your thoughts! I’m glad to hear you have a positive impression of the LangChain ReAct agent tools and its RAG (Retrieval-Augmented Generation) tech stack.
Since you mentioned not having hands-on experience yet, I think reviewing the official documentation is a great first step. It will give you a clear overview of the architecture, use cases, and how the tools interact.
If you’re comfortable with reading source code, diving into the repository could provide even deeper insights, such as how the agents are implemented, how RAG pipelines work, and how to customize them for your own projects.
Here are a few additional ideas to explore:
Experiment with a small demo project using LangChain + RAG to see the workflow in action.
Look into the examples directory in the GitHub repo—they often show practical applications.
Compare how LangChain RAG differs from other RAG implementations in terms of retriever integration, prompt design, and memory management.
Take note of common challenges or limitations mentioned in issues or discussions in the repo—it will help you anticipate problems if you start building with it.
When it comes to navigating the course catalog, consider creating a topological ordering based on pre-requisites. It’ll help students figure out the path required to achieve their desired outcome. Here’s an example.