Accurate homeopathic remedy recommendation is crucial for timely identification of disease and treatment planning. However, the prevailing approaches in the domain of homeopathic medicines recommendation are challenged due to presence of incomplete symptom information and lack of semantic understanding which lead to imprecise remedy suggestions. To overcome these limitations, the present work proposes Homeocure, a homeopathy remedy recommendation system based on the symptoms provided by means of integrating Large Language Models (LLMs) with a Retrieval-Augmented Generation (RAG) pipeline. To accomplish this task, firstly, the Homeopathic reference text is digitized using PyMuPDF and structured into JSON. Secondly, to mitigate the incomplete symptom description, the model utilizes an interactive re-questioning module. This process provides a well-defined symptom for effective recommendation. Finally, sentenceTransformers have been incorporated for symptom embeddings and indexed using FAISS for rapid semantic similarity search. To conclude, retrieved symptom–remedy pairs are provided as context to a lightweight generative model (TinyLlama-1.1B-Chat-v1.0). To enhance the reliability, three metrics: embedding-based cosine similarity, TF-IDF similarity, and fuzzy string matching are applied to evaluate consistency between user queries and recommended remedies. Experimental results demonstrate that the system delivers interpretable, context-grounded, and computationally efficient recommendations, highlighting the potential of LLM + RAG architectures for decision support in homeopathy and related medical domains.