flowmarg / scripts /verify_knowledge.py
akash
Deploy to Hugging Face (source code only, no binary files)
2379054
Raw History Blame Contribute Delete
4.7 kB
"""
KNOWLEDGE BASE VERIFICATION
Check if chunks are correctly created and retrievable
"""
import os
import sys
import json
from pathlib import Path
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
def verify_knowledge_base():
print("\n" + "=" * 60)
print(" 🔍 KNOWLEDGE BASE VERIFICATION")
print("=" * 60 + "\n")
# 1. Check ChromaDB
print("1️⃣ Checking ChromaDB...")
try:
import chromadb
client = chromadb.PersistentClient(path="knowledge/chroma_db")
collection = client.get_collection("track_design_knowledge")
count = collection.count()
print(f" ✅ Collection exists with {count} documents")
except Exception as e:
print(f" ❌ Error: {e}")
return False
# 2. Sample query - Network Rail standards
print("\n2️⃣ Testing search - 'Network Rail ballast depth'...")
try:
from sentence_transformers import SentenceTransformer
model = SentenceTransformer('all-MiniLM-L6-v2')
query = "Network Rail ballast depth requirement"
embedding = model.encode([query])
results = collection.query(
query_embeddings=embedding.tolist(),
n_results=3
)
if results and results.get('documents'):
print(f" ✅ Found {len(results['documents'][0])} results")
for i, doc in enumerate(results['documents'][0][:3], 1):
print(f"\n Result {i}:")
print(f" Source: {results['metadatas'][0][i-1].get('source', 'Unknown')}")
print(f" Preview: {doc[:200]}...")
else:
print(" ⚠️ No results found")
except Exception as e:
print(f" ❌ Error: {e}")
# 3. Test S&C query
print("\n3️⃣ Testing search - 'S&C switch crossing NR56'...")
try:
query = "NR56 switch and crossing details"
embedding = model.encode([query])
results = collection.query(
query_embeddings=embedding.tolist(),
n_results=3
)
if results and results.get('documents'):
print(f" ✅ Found {len(results['documents'][0])} results")
for i, doc in enumerate(results['documents'][0][:3], 1):
print(f"\n Result {i}:")
print(f" Source: {results['metadatas'][0][i-1].get('source', 'Unknown')}")
print(f" Preview: {doc[:200]}...")
else:
print(" ⚠️ No results found")
except Exception as e:
print(f" ❌ Error: {e}")
# 4. Test SOP query
print("\n4️⃣ Testing search - 'SOP track design procedure'...")
try:
query = "track design SOP procedure"
embedding = model.encode([query])
results = collection.query(
query_embeddings=embedding.tolist(),
n_results=3
)
if results and results.get('documents'):
print(f" ✅ Found {len(results['documents'][0])} results")
for i, doc in enumerate(results['documents'][0][:3], 1):
print(f"\n Result {i}:")
print(f" Source: {results['metadatas'][0][i-1].get('source', 'Unknown')}")
print(f" Preview: {doc[:200]}...")
else:
print(" ⚠️ No results found")
except Exception as e:
print(f" ❌ Error: {e}")
# 5. Check chunk distribution
print("\n5️⃣ Checking chunk distribution...")
try:
# Get all documents
all_results = collection.get()
if all_results and all_results.get('metadatas'):
sources = {}
for meta in all_results['metadatas']:
source = meta.get('source', 'Unknown')
# Clean source name
if 'REPW' in source:
source = source[:15] # Show first 15 chars for REPW files
sources[source] = sources.get(source, 0) + 1
print(f" Total chunks: {len(all_results['metadatas'])}")
print(f" Unique sources: {len(sources)}")
# Show top sources
print("\n Top sources:")
sorted_sources = sorted(sources.items(), key=lambda x: x[1], reverse=True)
for source, count in sorted_sources[:10]:
print(f" - {source}: {count} chunks")
except Exception as e:
print(f" ❌ Error: {e}")
print("\n" + "=" * 60)
print(" ✅ VERIFICATION COMPLETE")
print("=" * 60)
return True
if __name__ == "__main__":
verify_knowledge_base()