Running this workshop self-paced

This workshop is designed to run in a room with an instructor, but everything you need to finish it alone is already in this playbook. If you are working through the repository on your own - or someone shared it with you after the event - this page is your map.

Choose macOS or Windows in the page header before setup. Windows learners run all workshop commands in Ubuntu on WSL 2; PowerShell appears only in explicitly labelled Windows bootstrap and recovery blocks.

What changes without an instructor

Short answer: nothing in modules 00 through 07. The primary path is fully self-serve. You create your own ClickHouse Cloud trial, your own Postgres managed by ClickHouse with clickhousectl (module 03), and your own API keys - there is no shared resource you depend on. Modules 04 through 09 are the same.

An instructor normally provides two things, and each has a self-serve replacement built into these materials:

• The staffed setup clinic (module 00) is replaced by ./preflight.sh in ClickHouse_Demos/workshops/build_workshop/app, which checks everything the live bring-up trips over, and the Troubleshooting page, which lists every failure seen in testing with its exact fix.
• Live unsticking during a module is replaced by two things: your coding agent acting as instructor (below), and the instructor track as your answer key (below).

The instructor track is your answer key

The Instructor track is published in this same playbook. Each module has a matching instructor page with its timing, the failures that actually happen and how to unstick them, and the reset steps to get back to a known-good state. When a module does not behave as its verify step expects, open the matching instructor page - it is written for exactly that moment.

Your coding agent is your instructor

You already loaded the ClickHouse agent skills and connected the remote ClickHouse MCP to your coding agent in module 00. That same agent makes a capable workshop instructor. Paste the block below into it, filling in the module number and what you are seeing:

You are my workshop instructor. The workshop lives in this repo: the app is in workshops/build_workshop/app/, infrastructure notes are in workshops/build_workshop/infra/README.md, and the playbook markdown (learner and instructor tracks) is in site/content/docs/build-workshop. I am on module NN and stuck: <describe what you see>. Read the relevant module page and the troubleshooting page, inspect my local state (docker compose ps, container logs, and .env.workshop WITHOUT printing any secret values), and walk me to the module's verify step.

Because it has the skills and the MCP from module 00, the agent can also query your Cloud service directly while it helps - so it can check whether your tables, pipe, or telemetry actually landed, not just read the docs.

If you want the instructor framing to persist across sessions instead of re-pasting it, drop it into your tool's project-instructions file: CLAUDE.md for Claude Code, AGENTS.md for Codex CLI, and .cursor/rules/ (or AGENTS.md, which Cursor also reads) for Cursor. Each tool loads that file automatically at the repo root, so the guidance is in context every time you open the workshop.

Pacing advice

The per-module time budgets in the overview assume a room where the account signups were done ahead of time. On your own, module 00 dominates: the three browser signups (ClickHouse Cloud, Langfuse, OpenAI) are 5 to 10 minutes each, mostly waiting on email or phone verification. The "do the account signups before the day" callout in 00 Setup applies doubly when you are solo - do all three signups first, come back with keys in hand, and the rest of module 00 is quick. Modules 01 through 09 then run close to their stated budgets.

You are never stranded: every module's Starting point section states exactly what must be in place, so you can stop, come back later, and pick up from there.

Where to go next