On-Premises Kaggle AI Assistant with Qdrant RAG & Ollama

This workflow creates an entirely on-premises AI assistant tailored for Kaggle competitions, such as binary disaster-tweet classification. It leverages n8n for orchestration, Ollama for local LLM hosting (qwen3:8b for chat/coding, mxbai-embed-large for embeddings), and Qdrant for RAG to retrieve competition guidelines, tutorials, and notebooks converted to HTML. A local file trigger watches for new Jupyter notebooks in a specified folder, enabling on-demand ingestion and summarization using ALIENTELLIGENCE/contentsummarizer.

Key features include multi-turn conversational coding assistance with Python recommendations, debugging, and data science best practices; automated workflows for LLM calls and API integrations; and privacy-focused operations with no external data transfer, requiring GPU acceleration for performance. Benefits encompass enhanced productivity for data scientists, secure handling of proprietary competition data, and customizable domain-specific knowledge retrieval.

Ideal use cases: Kaggle participants needing quick code generation or troubleshooting during competitions; teams building private RAG systems for internal ML challenges; or researchers automating notebook processing into searchable study notes. Deploy via containerized kit for easy setup, starting with guideline prompts in initial chats.