extract-resume

Extract Resume - Source PDF → Structured Data

Read a resume's uploaded source PDF and produce JSON matching the JobPilot resume schema, then save via the API. Inverse of the editor.

Setup

Follow ../_shared/setup.md. The profile response provides user.primaryResumeId, primaryResumeSourceAbsolutePath, and resumes (every base with id, label, sourceFilename, hasData, isPrimary).

Step 1: Resolve Target

Parse the argument:

• Integer → use that resume id.
• Empty → use user.primaryResumeId. If no primary, stop:

  No primary resume set. Pass an explicit id, or set a primary at <$JOBPILOT_WEB/resumes>.

• --force (anywhere) → overwrite existing structured data. Otherwise refuse to overwrite (Step 3).

Let RESUME_ID be the resolved id, FORCE be true/false.

jobpilot-api GET /api/resumes/$RESUME_ID

If 404, stop and report the id doesn't exist.

Step 2: Verify Source PDF

sourceFilename must be set. If null, stop:

Resume {id} ({label}) has no uploaded source PDF. Upload one at <$JOBPILOT_WEB/resumes/{id}>, then re-run.

Resolve the absolute path:

• Primary resume → prefer primaryResumeSourceAbsolutePath.
• Otherwise → ${JOBPILOT_WORKSPACE_ROOT}/apps/api/storage/resumes/{sourceFilename}.

If sourceMimeType !== "application/pdf", stop and ask the user to re-upload as PDF.

Step 3: Refuse to Clobber

If content is non-null and FORCE === false, stop:

Resume {id} ({label}) already has structured data (version {n}). Edit at <$JOBPILOT_WEB/resumes/{id}>, or re-run with --force to overwrite from the PDF.

If FORCE, proceed and overwrite.

Step 4: Read and Parse

Read the PDF at the path from Step 2. Produce a single JSON object matching:

{
  basics: {
    name:     string,           // required
    headline?: string,          // professional title/headline if present
    email?:   string,
    phone?:   string,
    website?: string,
    linkedin?: string,
    github?:  string,
    location?: string,
  },
  summary?: string,             // 1–3 sentences
  experience: Array<{
    company:   string,
    title:     string,
    location?: string,
    start:     string,          // free-form, e.g. "Jul 2022"
    end?:      string,          // omit or "Present" if current
    bullets:   string[],
  }>,
  projects: Array<{
    name:        string,
    url?:        string,
    description?: string,        // one prose line; omit if there's only a tech-stack line
    bullets:     string[],
    keywords:    string[],       // the tech-stack line (e.g. "Next.js, Prisma, Docker")
    start?:      string,         // only if the PDF states it; same format as experience dates
    end?:        string,         // "Present" if ongoing
  }>,
  skills: Array<{
    group: string,              // e.g. "Languages"
    items: string[],
  }>,
  education: Array<{
    school:  string,
    degree:  string,
    start?:  string,
    end?:    string,
    details: string[],
  }>,
  publications: Array<{
    title:    string,
    authors?: string,           // the citation's author list, verbatim, et al. included
    venue?:   string,           // journal, conference, or publisher
    year?:    string,           // free-form: "2024", "In press", "Under review"
    url?:     string,
    doi?:     string,
  }>,
  awards: Array<{
    title:        string,
    issuer?:      string,
    year?:        string,
    description?: string,
  }>,
  certifications: Array<{
    name:          string,
    issuer?:       string,
    issued?:       string,
    expires?:      string,
    credentialId?: string,
    url?:          string,
  }>,
  sections: Array<{            // anything above doesn't model - see the rule below
    title:   string,           // the PDF's own heading, verbatim ("Grants", "Invited Talks")
    entries: Array<{
      heading:     string,     // the entry's first line - a grant name, a talk title
      subheading?: string,     // funder, host, venue
      meta?:       string,     // year, amount, or other right-aligned detail
      bullets:     string[],
    }>,
  }>,
}