Skip to content

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

AI for the messy bits

Parsing, cleaning, normalizing: the hands-on workshop from Extract Summit 2026. Everything here runs offline.

Setup

  1. Install uv.
  2. git clone https://github.com/zytedata/messy-bits-workshop && cd messy-bits-workshop
  3. uv sync

Run uv sync as soon as you can: it downloads Python and all dependencies, which takes a while on conference Wi-Fi.

Running things

From the directory of the script:

cd demo && uv run 01_parsing.py
cd exercise && uv run starter.py

For the # %% cells in VS Code, open this folder and pick .venv as the interpreter.

Running the notebooks with Docker

Prefer Jupyter without installing uv/Python? You only need Docker:

cd dublin
docker compose up --build

Open http://localhost:8888 (no token). Stop with Ctrl+C. Edits are saved to your local folder.

Optional, for hook/with_checklist.ipynb:

ANTHROPIC_API_KEY=sk-ant-... docker compose up

On Linux, if your user isn't 1000:1000, run UID=$(id -u) GID=$(id -g) docker compose up.

Layout

  • hook/: two LLM answers to the same prompt, quoted on the opening and closing slides, plus the prompt that produced them.
  • demo/: the three live-coding scripts and the sample pages they read.
  • exercise/: the hands-on. Start from starter.py, see README.md there.
  • dublin/: the Dublin version of the workshop: the same demos and exercise as Jupyter notebooks, with the Docker setup and the Scrapy + Jev demo.
  • slides.md: the deck, in Marp format.
  • tests/: uv run pytest checks that every script still prints what the slides say it does.

Governance model

This repository is governed. When the portal created it, it set the org custom property governed=true (plus production, contains_code, and lifecycle). Org-level rulesets in your GitHub org then apply automatically, based on those properties — you don't configure branch protection per repo:

Ruleset Applies when Effect on the default branch
baseline governed=true No force-push, no branch deletion
require-pr-production governed=true AND production=true PRs required, ≥1 approval, stale reviews dismissed
require-secrets-scan governed=true AND contains_code=true The secrets / detect-secrets check must pass

Secrets scanning

.github/workflows/secrets.yml calls the central reusable workflow in zytedata/governance. The required status check is secrets / detect-secrets. You cannot weaken this:

  • The scan logic lives in the governance repo, not here.
  • The check is required by the org ruleset, so a missing/failed check blocks merge.
  • New secrets must be explicitly audited into .secrets.baseline (and reviewed) before a PR can merge.

Run the scan locally before pushing:

pip install detect-secrets pre-commit
pre-commit install
pre-commit run --all-files
# If you add a legitimate, reviewed value that trips the scanner, audit it:
#   detect-secrets scan > .secrets.baseline   # then review the diff in your PR

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages