Django.Pub #5: Adding AI to Django Without Breaking the Product
The fifth Django.Pub meetup at LNTU covered data science, MLOps, RAG and AI agents in Django, with 40+ attendees on site and an online stream.
Author
SP-Lutsk Team
Date published

On 29 October, SP-Lutsk and Lutsk National Technical University (LNTU) held the fifth Django.Pub meetup for the Python and Django community in Lutsk. This edition focused on one practical question: how to add AI to a Django product without breaking the product itself.
The programme had three technical blocks: data science and classical ML, deep learning and MLOps, and LLMs, RAG and AI agents. Two open discussions with the audience covered data tooling and DevOps architecture. More than 25 tools were on the agenda, from NumPy and Polars to MLflow, Kubernetes, pgvector and LangGraph.
Attendees also worked with a live AI assistant built on Django, pgvector and LangGraph and loaded with every talk of the day. They left with the repository and a three-day API key to run it locally.
Three practical takeaways
Never run inference inside a view: queue it with Celery and Redis and return a task_id. Keep embeddings in PostgreSQL with pgvector before adding a separate vector database. Stream LLM output through Django Channels, because a 15-second spinner kills the user experience.
Because of the security situation, the meetup ran in a hybrid format: more than 40 people attended on site, and others followed the online stream.
Django.Pub is part of the long-term partnership between SP-Lutsk and LNTU's Faculty of Computer and Information Technologies, so students learn production practice from engineers who use it every day. Thanks to Dean Inna Kondius and the faculty team, and to the speakers from SP-Lutsk: Orest Novosad, Vlad Hetmanenko and Oleksandr Shypulin.

