Skip to content
fellowcoder

Tutorials · 5 builds

Build it once, understand it forever

Each tutorial ends with working code and an explanation of why it works — including the failure modes nobody mentions until production.

beginner
You can read code, but this stack is new
intermediate
Comfortable shipping; new to this problem
advanced
You've hit the wall the naive version leaves

Series

Retrieval that survives production

3 parts · 26 min total
  1. Chunking that survives contact with real documentsBuild the ingestion half of a retrieval pipeline: structure-aware splitting, contextual headers, and a schema that lets you re-chunk without re-embedding the world.intermediate9 min
  2. Hybrid retrieval: fuse keyword and vector searchVector search alone misses exact identifiers and rare terms. Add Postgres full-text search, fuse the rankings with RRF, and rerank the top slice.intermediate8 min
  3. Measuring retrieval: recall@k, MRR, and a golden set you'll maintainBuild the harness that tells you whether a change helped. Bootstrap a golden set from your own corpus, separate retrieval failures from generation failures, and stop shipping on vibes.advanced9 min

Standalone

One-sitting builds

Tutorial9 minadvanced

Build a streaming tool loop with the Claude API

A production-shaped agent loop in TypeScript: typed tools, streamed output, human approval gates, and the pause_turn case that silently truncates answers.

Start tutorial
Tutorial7 minbeginner

Full-text search in Postgres, without Elasticsearch

Weighted tsvectors, generated columns, ranked results, highlighted snippets, and typo tolerance — a complete search feature in one table.

Start tutorial