Resume Bot

I wanted to try out Langflow to build a little Retrieval Augmented Generation (RAG) Chatbot. Langflow is, in their words, "a visual framework for building multi-agent and RAG applications". Basically, you can hook together different components and run them. In my case, I hooked a PDF parser, chat window, a vector database, and OpenAI. It did not require any coding, but it's definitely an exercise for those who are more technically savy, as you end up bootstrapping it using a Python command.

I used this video as a starting point. Even though as I write this the video is less than two months old, some things are slightly different in the Langflow user interface. To me that shows how quickly the ecosystem is iterating.

Langflow workflow I used
Langflow workflow I used

After I created the workflow, I started the chat. You get presented with a chat interface similar to ChatGPT.

I started to ask it questions about my resume.

Do I know security?
Asking the bot: do I know security?

That wasn't too bad. Next, I asked it if I knew how to write software.

Do I know how to write software?
Asking the bot: do I know how to write software?

This is where ChatGPT's strength really shows. It was able to pick out the tools and languages from my resume, and it correctly structured them. Yes, it's relatively straightforward conceptually, but given that it was able to do it with some local data and a workflow I put together in a few minutes is quite satisfying.

It was a fun little exercise, and it's a good example of how much great, free or low cost software is out there. The workflow is on Github as "Resume Searcher".