One tiny change made my local LLMs more useful than ChatGPT for real work

Running local LLMs can be significantly enhanced by adding retrieval-augmented generation (RAG), which lets models pull context from your own documents instead of relying solely on static training data. LM Studio offers a built-in RAG plugin supporting up to five documents (30MB total), though the default 4096-token context window needs to be increased for practical use. For self-hosted tools, dedicated embedding models like Nomic Embed v1 convert documents into vector representations for semantic search. This approach keeps data private while making local 9B models competitive with cloud-based AI for document-heavy tasks.