The SymptomWise framework separates language understanding from diagnostic logic to eliminate hallucinations in medical AI. It maps free-text inputs to validated symptoms before using a deterministic reasoning module to rank diagnoses. This architecture restricts LLMs to symptom extraction only. Practitioners gain a traceable, verifiable pipeline for safety-critical clinical decision support systems.