Cv MCP server

CV Data Repository

The single source of truth for Francisco Perez-Sorrosal's CV content. YAML data, validation schema, and publishing workflow. Rendered artifacts (PDF, HTML, LaTeX, Typst, Markdown) are generated and released via GitHub Releases, never committed here.

What's here

• cv-data/resume.yaml — structured CV content (experience, education, skills, publications, patents, etc.)
• cv-data/resume-semantics.yaml — semantic overlay (topics, relationships, annotations)
• schemas/ — JSON Schema for validation; mirrored from cv-forge
• .github/workflows/ — CI validation and tag-triggered publishing

Making changes

To propose an edit to the CV data:

1. Clone this repo and create a branch: git switch -c update/your-change
2. Edit cv-data/resume.yaml or cv-data/resume-semantics.yaml
3. Validate locally (see below)
4. Push and open a PR against main

The CI workflow validates automatically; if checks pass, merge and proceed to Publishing.

Validate before you PR

# Install check-jsonschema (standalone CLI)
pip install check-jsonschema

# Validate both files
check-jsonschema --schemafile schemas/resume.schema.json cv-data/resume.yaml
check-jsonschema --schemafile schemas/semantics.schema.json cv-data/resume-semantics.yaml

All green? Ready to PR.

Publishing a release

A release is triggered by pushing a CalVer tag in the form YYYY.MM.DD (with optional .N for same-day re-releases, e.g. 2026.09.13.1).

git tag 2026.09.13
git push origin 2026.09.13

The workflow automatically:

• Renders the CV in all formats (PDF, LaTeX, Typst, Markdown, HTML)
• Publishes eight assets to the GitHub Release
• Deploys the HTML to the live CV site
• Creates a release.json manifest (consumed by the MCP server)

All assets are stable, version-free URLs under releases/latest/download/.

Republish an Existing Release

To re-run the publish workflow without creating a new tag (useful when templates or rendering logic changes):

gh workflow run publish.yml -f tag=2026.09.14

Or via the GitHub Actions UI:

1. Open Actions → Publish CV
2. Click "Run workflow"
3. Enter the tag in the tag input
4. Click "Run workflow"

This regenerates all eight release assets and redeploys the site, keeping the same version number. Assets at releases/latest/download/ are updated in place (overwritten).

Schema Updates

The JSON schemas in schemas/ are mirrored from cv-forge. When cv-forge releases a new version that changes the data models (via export-schemas), update the schemas in this repo:

1. After cv-forge merges a PR that changes src/cv_forge/models/, refresh the mirrored copies:

curl -s https://raw.githubusercontent.com/francisco-perez-sorrosal/cv-forge/v1/schemas/resume.schema.json > schemas/resume.schema.json
   curl -s https://raw.githubusercontent.com/francisco-perez-sorrosal/cv-forge/v1/schemas/semantics.schema.json > schemas/semantics.schema.json
   git add schemas/
   git commit -m "chore: sync schemas from cv-forge v1"
   git push

1. The CI drift-check (.github/workflows/validate.yml) compares the local schemas against the remote cv-forge v1 copies. If they diverge, the check fails and blocks PRs until the sync above completes. A push that lands seconds before cv-forge re-points its v1 tag can fail the drift check with a 404 on the upstream fetch; re-run the failed Validate CV Data job once the tag exists.

Consuming the CV

As data

• YAML sources: resume.yaml and resume-semantics.yaml, published as release assets (see the URL pattern below)
• JSON: the MCP server exposes the same data as JSON resources (fps-cv://resume, fps-cv://semantics) and per-entry lookups

As documents

Shortened. The full README is on GitHub.