CV Builder

A static CV publishing pipeline that:

• Pulls CV records/items from Airtable into normalized JSON.
• Renders CVs for web (/cv?cv=<slug>), print (/cv-print.html), and external embed (cv-embed.js).
• Builds PDF files into dist/<slug>.pdf using Vivliostyle.
• Optionally generates AI feedback for a single triggering CV and writes it back to Airtable.

---

Repository Structure

• scripts/build-cv-json.js — Airtable ➜ data/cv/*.json exporter.
• scripts/build-cv-pdf.js — JSON ➜ paginated HTML ➜ dist/*.pdf builder.
• scripts/run-cv-ai-feedback.js — Airtable CV ➜ OpenAI feedback ➜ Airtable update (Overall AI Feedback).
• cv-render.js — browser renderer shared by viewer/print/embed flows.
• cv.html — viewer page for a single CV with a PDF link.
• cv-feedback.html — CV + AI feedback review page (overall and item-level feedback side-by-side).
• cv-print.html + cv-print.css — print-focused paginated view.
• cv-embed.js + cv-embed.html — non-iframe embeddable widget + usage/instructions page.
• .github/workflows/cv-ai-feedback.yml — issue/manual-trigger workflow for AI feedback runs.

---

High-Level Flow

1. Data ingest: Airtable records are fetched and normalized into data/cv/<slug>.json.
2. Render paths:
   • Web viewer (cv.html) renders CV content client-side.
   • Print page (cv-print.html) renders paginated layout client-side.
   • Embed script (cv-embed.js) renders inside Shadow DOM on external sites.
3. PDF build: scripts/build-cv-pdf.js transforms JSON into print HTML and runs Vivliostyle to generate dist/<slug>.pdf.
4. AI feedback (optional): workflow/script fetches the triggering CV, sends structured prompt to OpenAI, writes output to Airtable Overall AI Feedback.

---

Prerequisites

• Node.js 20+ (ESM project).
• npm.
• Airtable PAT + base/table identifiers for Airtable scripts.
• Optional: OPENAI_API_KEY for AI feedback flow.

Install dependencies:

npm install

---

npm Commands

npm run build:json    # Build CV JSON files from Airtable
npm run pdf:build     # Build PDFs from data/cv/*.json
npm run pdf:preview   # Open Vivliostyle preview for cv-print.html
npm run ai:feedback   # Run AI feedback script for the triggering CV

---

Airtable JSON Build (build:json)

What it does

• Reads CV header-level data from the CVs table.
• Reads itemized content from the CV Items table.
• Normalizes rich-text field variants.
• Writes canonical JSON to data/cv/<slug>.json.

Trigger targeting logic

The script can resolve a specific CV record ID from:

• Issue title pattern like CV_UPDATE recXXXXXXXXXXXXXX.
• Issue body line like recordid recXXXXXXXXXXXXXX.

If configured, it can also build all published CVs.

Environment variables

Required for Airtable access:

• AIRTABLE_PAT
• AIRTABLE_BASE_ID

Optional/config:

• CVS_TABLE_NAME (default: CVs)
• CV_ITEMS_TABLE_NAME (default: CV Items)
• CVS_TABLE_ID
• CV_ITEMS_TABLE_ID
• ISSUE_TITLE
• ISSUE_BODY
• BUILD_ALL_CVS=true|false

CV Footer behavior

build-cv-json reads footer content from the CV table using CV_Footer (also supports fallback matching), and emits it in JSON as section footer for downstream renderers.

---

Web Viewer (cv.html)

• Route format: /cv?cv=<slug>.
• Loads data/cv/<slug>.json.
• Renders CV content using CvRender.renderStandardCv.
• Provides a single action to open dist/<slug>.pdf in a new tab.

---

Print View (cv-print.html + cv-print.css)

• Route format: /cv-print.html?cv=<slug>.
• Uses paginated rendering (CvRender.renderPaginatedCv).
• CSS defines print-friendly multi-block layout and footer area.
• Intended for previewing/validating print layout behavior and as a design reference for PDF generation.

---

PDF Build (pdf:build)

What it does

• Reads each data/cv/*.json file.
• Generates an HTML document with the same layout model as print rendering.
• Calls npx vivliostyle build to output dist/<slug>.pdf (primary five-block layout) and dist/<slug>.secondary.pdf (secondary split layout).

Output

• PDFs are written to dist/ (e.g., dist/coleman-davis.pdf).
• Secondary layout output is written to dist/<slug>.secondary.pdf.

Important notes

• This environment may fail during Playwright Chromium download if external mirrors return 403.
• Script-level parse/runtime checks can still be verified with node --check scripts/build-cv-pdf.js.
• Set CV_PDF_LAYOUTS=primary|secondary|both (default: both) to control which PDF layout(s) are generated.
• Secondary layout behavior: page 1 is single-column, while page 2 uses main+rail with main ordering Work Experience (Cont.), Technical + IT, Recent Projects, and rail ordering Skills, then Education.

---

Embeddable CV (cv-embed.js)

A non-iframe embed that renders directly in the host page and isolates styles using Shadow DOM.

Basic snippet

<div id="cv-embed-target"></div>
<script
  src="https://cpd4f.github.io/cv-builder/cv-embed.js"
  data-cv="coleman-davis"
  data-target="cv-embed-target"
></script>

Supported data-* attributes

• data-cv (required): slug to load (data/cv/<slug>.json).
• data-target (optional): target container ID.
• data-base-url (optional): override base URL for self-hosting.

Embed docs page

• cv-embed.html provides:
  • Copy-to-clipboard snippet.
  • Usage/options guidance.
  • Live preview.

---

AI Feedback Automation

Script (scripts/run-cv-ai-feedback.js)

• Targets one triggering CV record (resolved from issue/body inputs).
• Converts CV JSON/content into structured prompt text.
• Calls OpenAI using OPENAI_API_KEY.
• Writes response to Airtable field: Overall AI Feedback.
• Then writes targeted CV-table feedback to AI Feedback Intro, AI Feedback Core Competencies, and AI Feedback Footer when those source blocks exist.
• Finally (last step) generates item-level feedback for eligible CV Items and writes it to each item's AI Feedback field in the CV Items table (excluding Education and Second Page Rail).

Model

• Configured for gpt-5.4 per project requirement for overall, CV-table field, and item-level feedback generation.

Required environment

• AIRTABLE_PAT
• AIRTABLE_BASE_ID
• OPENAI_API_KEY
• Issue context (ISSUE_TITLE/ISSUE_BODY) when running issue-driven targeting.

---

GitHub Actions: cv-ai-feedback.yml

Workflow triggers:

• workflow_dispatch
• issues events: opened / edited / reopened

Behavior:

1. Validates guard conditions for issue-triggered runs (title prefix + expected issue author + expected event actor).
2. Honors feature toggle AI_FEEDBACK_ENABLED (GitHub Actions variable, default true) so AI feedback can be turned on/off without changing code.
3. Installs dependencies.
4. Runs npm run ai:feedback.
5. Posts success/failure comment back to the triggering issue.

Feature toggle setup:

• Configure repository variable AI_FEEDBACK_ENABLED in Settings → Secrets and variables → Actions → Variables.
• Truthy values (true, 1, yes, on) keep AI feedback enabled.
• Any other value disables AI feedback execution in the workflow.

---

Local Development Checklist

1. npm install
2. Configure .env (or export shell vars) for Airtable/OpenAI secrets.
3. npm run build:json
4. npm run pdf:build
5. Serve repository root with any static server and test:
   • /cv?cv=coleman-davis
   • /cv-print.html?cv=coleman-davis
   • /cv-feedback.html?cv=coleman-davis
   • /cv-embed.html

---

Troubleshooting

npm run pdf:build fails with Playwright download 403

This is typically an environment/network limitation when Vivliostyle tries to fetch Chromium. Re-run in an environment with Playwright download access or preinstalled browser binaries.

Missing footer content

Confirm the source Airtable field is CV_Footer on the CVs table and rerun npm run build:json. Footer content is optional and only renders when present.

Triggered workflow ran but nothing updated

• Check issue title/body includes valid record ID (rec...).
• Confirm secrets are available to workflow (AIRTABLE_PAT, AIRTABLE_BASE_ID, OPENAI_API_KEY).
• Review workflow logs and issue comments for guard/failure reasons.

---

Deployment Notes

This repo is designed for static hosting (e.g., GitHub Pages). Ensure the following are published:

• data/cv/*.json
• dist/*.pdf
• cv.html, cv-print.html, cv-embed.html, cv-embed.js, cv-render.js, cv-print.css