tests/verify_integrated_data.py

import asyncio
import os
import sys

# Disable tokenizers parallelism for stability
os.environ["TOKENIZERS_PARALLELISM"] = "false"

# Add server directory to path
sys.path.append(os.path.join(os.path.dirname(__file__), '../server'))

from app.services.medical_retriever import get_medical_retriever
from app.services.vector_operations import VectorOperations

async def verify_data():
    print("🔍 VERIFYING INTEGRATED HEALTHCARE DATA...")
    
    # SYSTEM STABILITY: Use same safe embedding as ingestion for verification
    async def safe_get_embedding(text: str):
        return VectorOperations._simple_embedding(text).tolist()
    
    # Mock it on VectorOperations before initializing retriever if possible, 
    # but retriever usually calls it. Let's patch it globally for this script.
    VectorOperations.get_embedding = safe_get_embedding
    
    retriever = get_medical_retriever()
    
    print(f"\n📊 TOTAL DOCUMENTS INDEXED: {len(retriever.documents)}")
    for i, doc in enumerate(retriever.documents):
        source = retriever.doc_metadata[i].get('source') if i < len(retriever.doc_metadata) else "Unknown"
        print(f"   [{i}] Source: {source} | Text: {doc[:80]}...")
    
    test_queries = [
        ("treatment for dummyitis", "MedQuAD"),
        ("is X effective", "PubMedQA"),
        ("what is flu", "WHO/CDC")
    ]
    
    all_passed = True
    for query_text, expected_source in test_queries:
        print(f"\n🔎 Searching for: '{query_text}'...")
        embedding = await VectorOperations.get_embedding(query_text)
        results = retriever.search_medical(query_text, embedding, top_k=3)
        
        if results:
            match = results[0]
            print(f"✅ Found Match!")
            print(f"   Source: {match.get('source')}")
            print(f"   Retrieved Text: {match.get('text')[:100]}...")
            print(f"   Answer: {match.get('answer', match.get('text', ''))[:100]}...")
            print(f"   Confidence: {match.get('confidence'):.4f}")
            
            if match.get('source') != expected_source:
                print(f"⚠️  Warning: Source mismatch. Expected {expected_source}, got {match.get('source')}")
        else:
            print(f"❌ No results found for '{query_text}'")
            all_passed = False
            
    if all_passed:
        print("\n🎉 VERIFICATION SUCCESS: All datasets are correctly integrated and searchable!")
    else:
        print("\n❌ VERIFICATION FAILED: Some data could not be retrieved.")

if __name__ == "__main__":
    asyncio.run(verify_data())