The preservation of the historical heritage of Palestine, a domain which is characterized by its valuable cultural and historical significance, presents an interesting challenge in digital heritage preservation. This study suggests the development of a conversational agent that utilizes state-of-the-art natural language processing methods. Founded on a Retrieval-Augmented Generation, the system utilizes LLMs like Mistral and LLaMA 3. The mitigation of inherent biases in LLMs is one of the major challenges that this project addresses. These biases can have a profound impact on the representation and interpretation of historical events, particularly in politically and culturally sensitive contexts. ChromaDB, a vector database, provides efficient data retrieval by organizing and indexing data in a manner that is optimized for Similarity searches. This provides seamless integration with the system’s (RAG) architecture, while Chainlit provides interactive and user friendly interaction with the conversational agent. To evaluate the chatbot’s performance, a set of 80 questions was meticulously chosen, with an equal distribution of Arabic and English questions. The questions were assessed for their intelligibility, relevance, and thoroughness. The results suggest that Mistral achieved a performance score of 60% in Arabic and 82.5% in English, LLaMA 3 scored 77.5% in Arabic and 79.16% in English. These results highlight the system’s possibilities for multilingual historical preservation. This paper, which opens the path for more developments in multilingual NLP applications, shows the benefit of combining vector based retrieval with LLMs for digital legacy.