Code Crew
HomeAboutServicesPortfolioBlogContact
Start Your Project
Code Crew

We turn your ideas into digital reality.

Company

  • About
  • Portfolio
  • Blog
  • Contact

Services

  • Services

Contact Information

  • info@codecrew.com.tr

Subscribe to our newsletter

Be the first to hear about new posts and announcements.

© 2026 Code Crew. All rights reserved.

Privacy PolicyTerms of ServiceCookie Policy
Blog
AI

LLM + RAG Architecture: Production Notes

Lessons from building a customer assistant with RAG: vector DB selection, chunking strategies and evaluation.

Serhat Kıran
April 11, 20267 min read

Vector DB Choice

We compared pgvector, Qdrant and Weaviate. Teams already running PostgreSQL find pgvector a frictionless start.

Chunking

Semantic chunking outperforms fixed-size chunking significantly on accuracy.

Evaluation

We run regression tests on every PR using Ragas and an internal golden set.

Tags:
LLM