Guide Shows How to Build a Free Local RAG Chatbot for Trading Research on Android

A developer tutorial published on DEV Community outlines how to build a Retrieval-Augmented Generation (RAG) chatbot that runs entirely on an Android phone using Termux and Ollama, with no paid APIs or cloud services involved. The system is designed for retail traders and quant researchers who want to query their own documents — such as PDFs, markdown notes, and option-chain exports — without sending data to third-party servers. The pipeline consists of four steps: ingesting documents, generating local embeddings, storing vectors in a file-based index, and retrieving relevant chunks to answer queries using a local language model. The entire implementation requires roughly 200 lines of Python and about 2 GB of free storage, with models like Llama 3.2 and nomic-embed-text running locally via Ollama. On a mid-range phone, the setup delivers around 3–8 tokens per second, which the author describes as slow but functional for document question-and-answer tasks.