Medical GraphRAG Assistant

#!/usr/bin/env python3 """Load test image with NV-CLIP NIM embedding""" import sys sys.path.insert(0, '.') from src.db.connection import get_connection from src.embeddings.nvclip_embeddings import NVCLIPEmbeddings print('Initializing NV-CLIP NIM embedder...') embedder = NVCLIPEmbeddings() # Generate embedding for text (we don't have actual images on AWS yet) print('\nGenerating text embedding for test...') query = 'chest x-ray pa view' embedding = embedder.embed_text(query) print(f'✓ Generated {len(embedding)}-dim embedding') print(f'Sample values: {embedding[:5]}') # Insert into database print('\nInserting into MedicalImageVectors...') conn = get_connection() cursor = conn.cursor() image_id = 'TEST_NIM_001' patient_id = 'P10000000' study_type = 'Chest X-ray PA' image_path = 'test_dicom.png' embedding_str = '[' + ','.join(map(str, embedding)) + ']' cursor.execute(""" INSERT INTO SQLUser.MedicalImageVectors (ImageID, PatientID, StudyType, ImagePath, Embedding, CreatedAt, UpdatedAt) VALUES (?, ?, ?, ?, TO_VECTOR(?, DOUBLE), CURRENT_TIMESTAMP, CURRENT_TIMESTAMP) """, (image_id, patient_id, study_type, image_path, embedding_str)) conn.commit() print(f'✓ Inserted {image_id}') # Verify cursor.execute("SELECT COUNT(*) FROM SQLUser.MedicalImageVectors") count = cursor.fetchone()[0] print(f'\n✓ Total images in database: {count}') cursor.close() conn.close() print('\n✅ Test image loaded with NV-CLIP NIM embedding!')