create-delta-blog-post

name

create-delta-blog-post

description

Scaffolds a new Delta Lake blog post under `src/content/blog/`. Guides the user through generating a URL-safe dated slug directory, front matter (title, description, thumbnail, author, publishedAt) matching the schema in `src/content.config.ts`, optionally creating new author profiles under `src/content/profiles/`, and pasting/grammar-checking initial content. Use when the user says "new blog post", "create a blog", "delta blog", "scaffold a post", or asks to add a post under `src/content/blog`.

Create Delta Blog Post

Scaffold a new blog post in this repo (delta-io/website) end-to-end. Drive the workflow with AskQuestion for every decision point, then write files. The blog and profile schemas live in src/content.config.ts — keep generated front matter in lockstep with them.

Workflow Checklist

Track progress as you go:

- [ ] 1. Generate dated, URL-safe slug from the blog title
- [ ] 2. Confirm and create src/content/blog/<dated-slug>/
- [ ] 3. Collect front matter via dialog (title, description, thumbnail, authors, publishedAt)
- [ ] 4. For each missing author, run the new-author sub-flow
- [ ] 5. Write index.md with front matter
- [ ] 6. Optionally paste content + run grammar/spelling pass
- [ ] 7. Optionally verify links
- [ ] 8. Print final checklist for the user

Step 1 — Slug & Directory

1. Ask the user for the blog title (the long, human-readable form). Use AskQuestion only if the title isn't already in the conversation; otherwise just confirm it.
2. Determine the creation date as today's date in YYYY-MM-DD (use the <timestamp> from the system, not a guess).
3. Slugify the title:
   • lowercase everything
   • strip punctuation: , : . ? ! " ' ( ) [ ] { } ; backticks
   • replace whitespace, &, /, \, and _ with a single -
   • drop any non-ASCII characters (transliterate accents to ASCII first if reasonable)
   • collapse repeated -
   • trim leading/trailing -
4. Final directory name: {YYYY-MM-DD}-{slug} under src/content/blog/.

Worked example:

• Title: Delta Grows Up: Writes, Time Travel, and Unity Catalog
• Date: 2026-05-06
• Result: src/content/blog/2026-05-06-delta-grows-up-writes-time-travel-and-unity-catalog/

Confirm the resulting path with the user before creating it. If a directory with that name already exists, surface that and ask whether to choose a different slug or overwrite.

Step 2 — Front Matter Dialog

The blog collection schema (from src/content.config.ts) requires:

• title: string
• description: string
• author: reference("profiles") | reference("profiles")[]
• publishedAt: date
• thumbnail: image
• updatedAt: date (optional)

Ask each of the following with AskQuestion. Keep questions short and present concrete options when meaningful.

a. Title

The title in front matter does not need to match the slug. Default to the user-provided title; offer a single follow-up to override with an alternative title if they want a different display title than what was used to generate the slug.

b. Description

Two options:

• Write now — collect a description ≤200 characters intended for the preview/social card.
• Defer — insert description: TODO add description before publishing and remind the user this must be filled in before merge.

c. Thumbnail

Ask for the thumbnail filename or path (target ~1200x628):

• If the user provides a file path, copy/move it into the new blog directory and reference it as ./<filename> in front matter.
• If the user has nothing yet, default to thumbnail: ./thumbnail.png and warn that the file must be added before the site builds (the schema requires a real image).

d. Authors

Accept one or more author slugs. For each entered slug:

1. Check whether src/content/profiles/<slug>/index.md exists.
2. If it does not exist, run the New Author Sub-Flow in Step 3 before adding the slug to the list.
3. Always emit the author field as a YAML list, even when there is only one author. This matches existing posts such as src/content/blog/2026-04-17-delta-4-2-released/index.md.

e. publishedAt

Ask for the expected publication date in YYYY-MM-DD form. Default to today if the user has no preference. (updatedAt is optional — only add it if the user explicitly asks.)

Generated index.md template

Write the front matter exactly in this shape (drop the body for now — it's filled in Step 4):

---
title: <title>
description: <description-or-TODO>
thumbnail: ./<thumbnail-filename>
author:
  - <author-slug>
publishedAt: <YYYY-MM-DD>
---

Step 3 — New Author Sub-Flow

Triggered when the user names an author whose profile doesn't exist yet. The profiles schema allows: name (required), and optional photo, role, quote, quoteSource, linkedin, videos, otherReferences.

Reference: src/content/profiles/scott-haines/index.md.

Sub-flow:

1. Ask for the author's full name.
2. Derive a profile slug the same way as the blog slug (lowercase, ASCII, hyphenated). Confirm it with the user — pre-existing slugs in src/content/profiles/ like dennyglee or scottsandre show some authors prefer non-hyphenated forms, so allow a manual override.
3. Confirm src/content/profiles/<slug>/ does not already exist.
4. Ask (each is optional, allow skip):
   • Role — e.g. Staff Software Engineer.
   • LinkedIn URL — must be a valid https://... URL if provided (schema enforces .url()).
   • Headshot — if provided, place at src/content/profiles/<slug>/<slug>.<ext> and reference as photo: ./<slug>.<ext>.
5. Write src/content/profiles/<slug>/index.md containing only the fields the user provided. Always include name. Omit empty optional fields entirely (do not write role: with no value).

Example output for a minimal new profile:

---
name: Jane Doe
role: Senior Engineer
linkedin: https://www.linkedin.com/in/janedoe/
---

After the profile is written, return to Step 2d and add the new slug to the post's author list.

Step 4 — Content & Grammar Pass

Ask the user: "Do you have draft content to paste in now?"

• No → leave the body empty after the front matter and skip to Step 6.
• Yes → wait for the user to paste the markdown body, then:
  1. Insert the body into index.md directly under the closing ---.
  2. Run a grammar/spelling pass only. Rules:
     • Fix misspellings, typos, missing/extra spaces, broken punctuation, obvious subject/verb agreement errors, and capitalization.
     • Do not rewrite sentences, restructure paragraphs, change voice, or "improve" wording.
     • Do not add or remove content. If something is unclear, leave it as a question for the user, don't guess.
  3. Briefly summarize the corrections made (e.g. "Fixed 3 spellings: Catalouge → Catalog, recieve → receive, untill → until").

Step 5 — Optional Link Verification

After content is in place, ask: "Want me to verify the links in the post?" Offer two modes via AskQuestion:

• Automated check — extract every external URL (skip relative paths, anchor-only links like #section, and mailto:), then for each run:

  curl -sI -L -o /dev/null -w "%{http_code}" --max-time 10 "<url>"
  

  Flag any URL whose final status is not in 2xx/3xx. Report results as a table: URL, status code, verdict.

• Manual list — extract every external URL and print them as a checklist for the user to spot-check; do not make any network calls.

Step 6 — Final Checklist

Print this checklist to the user as the last step:

- [ ] Directory created at src/content/blog/<dated-slug>/
- [ ] index.md front matter validates against the blog schema
- [ ] Thumbnail file present in the directory (or TODO noted)
- [ ] Every author slug resolves to src/content/profiles/<slug>/index.md
- [ ] publishedAt is a valid YYYY-MM-DD
- [ ] Description is ≤200 chars or marked TODO
- [ ] Grammar/spelling pass run if content was pasted
- [ ] Links verified (or manual list provided) if content was pasted

Out of Scope

• Running astro check or building the site.
• Image resizing or optimization.
• Creating commits, branches, or PRs (let the user drive Git).