The paper introduces Telco-RAG, an open-source Retrieval-Augmented Generation (RAG) framework tailored for telecommunications standards, particularly 3GPP documents. The challenge lies in the complexity and rapid evolution of telecom standard documents, which make it difficult for standard LLMs to understand and process. Telco-RAG addresses these challenges by optimizing RAG pipelines for technical content, enabling more accurate and efficient retrieval and generation of information related to telecom standards. The framework includes a dual-stage pipeline: a query enhancement stage and a retrieval stage. The query enhancement stage uses a custom glossary of technical terms and a neural network (NN) router to refine queries and select relevant documents. This improves the accuracy and efficiency of the retrieval process. The retrieval stage then uses the NN router to select documents for the second retrieval, enhancing the final response quality. Telco-RAG also includes hyperparameters optimization, query augmentation, and RAM usage reduction techniques. The framework uses a specialized NN router to predict relevant 3GPP series based on queries, reducing RAM usage by selectively loading embeddings. The prompt engineering in Telco-RAG enhances the LLM's ability to understand and respond to queries by structuring the prompt with definitions, abbreviations, and context. The experimental results show that Telco-RAG significantly improves the accuracy of LLMs in answering telecom-related questions. It outperforms traditional RAG systems and standalone LLMs in terms of accuracy and efficiency. The framework is designed to be generalizable, providing guidelines for RAG implementation in other technical domains. The Telco-RAG system is publicly available as an open-source chatbot for 3GPP standards, contributing to the integration of AI in the telecommunications field.The paper introduces Telco-RAG, an open-source Retrieval-Augmented Generation (RAG) framework tailored for telecommunications standards, particularly 3GPP documents. The challenge lies in the complexity and rapid evolution of telecom standard documents, which make it difficult for standard LLMs to understand and process. Telco-RAG addresses these challenges by optimizing RAG pipelines for technical content, enabling more accurate and efficient retrieval and generation of information related to telecom standards. The framework includes a dual-stage pipeline: a query enhancement stage and a retrieval stage. The query enhancement stage uses a custom glossary of technical terms and a neural network (NN) router to refine queries and select relevant documents. This improves the accuracy and efficiency of the retrieval process. The retrieval stage then uses the NN router to select documents for the second retrieval, enhancing the final response quality. Telco-RAG also includes hyperparameters optimization, query augmentation, and RAM usage reduction techniques. The framework uses a specialized NN router to predict relevant 3GPP series based on queries, reducing RAM usage by selectively loading embeddings. The prompt engineering in Telco-RAG enhances the LLM's ability to understand and respond to queries by structuring the prompt with definitions, abbreviations, and context. The experimental results show that Telco-RAG significantly improves the accuracy of LLMs in answering telecom-related questions. It outperforms traditional RAG systems and standalone LLMs in terms of accuracy and efficiency. The framework is designed to be generalizable, providing guidelines for RAG implementation in other technical domains. The Telco-RAG system is publicly available as an open-source chatbot for 3GPP standards, contributing to the integration of AI in the telecommunications field.