The automotive industry is drowning in technical PDFs. From design specs to maintenance manuals, these documents are critical but often difficult to navigate. What if you could simply ask a question and get an immediate, accurate answer? That's the promise of Retrieval-Augmented Generation (RAG) and Large Language Models (LLMs), a powerful combination poised to transform how engineers and technicians access vital information. Imagine a local, offline chatbot trained on your company's specific automotive PDFs, capable of understanding complex technical jargon and multi-column layouts. Researchers have been tackling this challenge, using the open-source Ollama model as a foundation. Ollama allows for rapid deployment of LLMs in low-performance settings, perfect for industrial environments with limited resources and data privacy concerns. This research goes beyond basic RAG, incorporating several key innovations. First, a custom PDF processing technique combines the strengths of PDFMiner and Tabula to accurately extract information from complex document structures, including tables and multi-column layouts. Second, an advanced RAG system built on the Langchain framework uses a combination of BM25 retrieval and the BGE reranker model, a large language model designed specifically for enhanced search relevance in Chinese. Finally, a self-RAG agent powered by LangGraph improves the chatbot’s reasoning abilities by allowing it to “ask itself” clarifying questions, leading to more accurate and coherent responses, especially for complex, multi-step queries common in automotive diagnostics and troubleshooting. Testing across various datasets, including a proprietary collection of automotive documents, revealed substantial improvements in answer accuracy and relevance. The results are promising, pointing towards a future where information access in the automotive industry is seamless and intuitive. Imagine troubleshooting a complex electrical issue with a simple natural language query or quickly retrieving specific safety regulations from a vast collection of documents. This optimized RAG technique powered by Ollama offers a real-world solution for enhancing productivity and innovation in automotive engineering and manufacturing. However, there's more work to be done. Future research aims to refine the real-time performance of these chatbots, essential for fast-paced manufacturing environments. Integration with visual data like diagrams and schematics is also a key focus, along with ensuring data privacy and ethical AI usage. This research lays the foundation for a new era of intelligent information retrieval in the automotive sector, one where technical knowledge is readily available and intuitively accessible to everyone.