Python

At STCC I wrote the core of a RAG app in Python: an API that turned documents into embeddings and answered chat questions using a local model. I use Python mostly for integrations, scripts and talking to language models.

How it worked

The documents came in every format: PDFs, Word files and others. I converted them to Markdown first and only then computed their embeddings. The database was PostgreSQL. The interface was a CodeIgniter 4 app that proxied requests to the Python API; the API handled the embeddings and the chat with the model, which ran on Ollama, adding the relevant passages from the documents to each question.

Not Local RAG Example

I've also written an open-source RAG, Local RAG Example, but it's a different system, written entirely in Go.

A language I keep going deeper into

Python is one of the languages I keep studying in my self-taught path, for its versatility and its role as the de facto standard in AI and data analysis.