How this site works

This site is a small full-stack system. The content is Markdown and YAML. At build time it becomes a SQLite database and a set of JSON endpoints. In the browser, sql.js loads the database on demand so you can query it directly. One of the three server functions, /api/ask, sends your question and the schema to a language model, checks the SQL it returns by preparing it against the real database, and hands the SQL back to your browser to run. The model never touches the data.

Architecture

ArchitectureMarkdown and YAMLat build timeSQLite databaseJSON endpointson demandyour browsersql.jsyour question/api/askpreparing it againstthe real databasethe schemaSQLlanguage modelThe model never touches the data.ArchitectureMarkdown and YAMLat build timeSQLitedatabaseJSONendpointson demandyour browsersql.jsyour question/api/askpreparing it againstthe real databasethe schemaSQLlanguage modelThe model nevertouches the data.
curl -A curl/8.0 https://alexkachur.com

Build-time database

/data/portfolio.sql

The ask pipeline

The model behind /api/ask is one environment variable, and Anthropic publishes each model's earliest retirement date, so a swap is a change of that variable and a re-recorded eval.