Johannes Wachter
Blog Shorts Topics Projects About
// RAG

RAG beyond hello world

Retrieval is a pipeline, not a lookup.

A series on building retrieval that survives contact with real documentation. It starts where most tutorials stop, with the naive version, and walks through the decisions that actually move quality: what you index, what you search for, which candidates you generate, how you rank them, and how you know any of it got better. Built in the open with symfony/ai.

article
RAG beyond hello world: retrieval is a pipeline →
September 24, 2026 · 7 min
The mental model. Why the naive version breaks on real docs, and the four questions retrieval really is.
article
Indexing sets the ceiling for retrieval →
September 30, 2026 · 7 min
The indexing side: loading, chunking and enriching the source before anyone searches it.
article
How embeddings and vector stores actually work →
October 15, 2026 · 7 min
What a vector really is, how embeddings are produced, and what separates a vector store from a plain database. The foundation the earlier pieces leaned on.
tutorial
Building a minimal RAG pipeline with Symfony AI →
October 28, 2026 · 8 min
The deliberate hello world: index, embed, retrieve, answer. Then see where it breaks.

All writing →

© 2026 Johannes Wachter · johanneswachter.dev · Privacy · RSS: everything, writing, shorts