ERPNext job applicants scored against the opening by AI

ERPNext job applicants scored against the opening by AI

An n8n workflow for ERPNext. When a job applicant is added, it reads the PDF resume, compares it with the job opening and has
Google Gemini write a fit level, a score, a rating and a reason. It saves them on the applicant, and a score under 80 sets the
status to Rejected while 80 or more sets Accepted.

What is inside

• workflow.json: the workflow as n8n exports it, 38 nodes. Fifteen are sticky notes that explain its steps. It holds no
  credential.
• README.md: these notes.

What we changed from the original copy

• The model was gemini-2.0-flash-exp. It is now models/gemini-3.6-flash, which Google names as the replacement for its 2.0
  flash family.
• The prompt asked for FitLevel with no colon, while Convert to Fields looks for FitLevel:. The prompt now has the colon,
  so the fit level is found.
• Three HTTP nodes called its author's own ERPNext address. They now say <YOUR_ERPNEXT_HOST>, so put in your own.
• The webhook path carried its author's site name. It is now erpnext-job-applicant-shortlist.
• A sticky note with its author's credit, a payment link and an advert is left out.
• Every credential link is gone, so select your own in each node that needs one.
• Nothing else. Microsoft Outlook, WhatsApp Business Cloud and Txt File to Text (Example) are as the original left them.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An ERPNext site with the Job Applicant doctype, an API credential for n8n, and permission to add a webhook and custom fields.
• A Google Gemini API key.
• Resumes stored as PDF links that open without a login. The download node sends no credential.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. In ERPNext, add three custom fields to the Job Applicant doctype, with these field names: custom_fit_level,
   custom_score and custom_justification_by_ai. applicant_rating is already there.
3. Add an ERPNext API credential and select it in ERPNext - Reject if Resume not Attached, ERPNext - Hold Applicant,
   Get Job Opening, Reject Applicant, Update Applicant Data and Accept Applicant.
4. In Reject Applicant, Update Applicant Data and Accept Applicant, replace <YOUR_ERPNEXT_HOST> in the address with your
   ERPNext host, such as erp.example.com.
5. Add a Google Gemini credential and select it in Google Gemini Chat Model.
6. Fill in Microsoft Outlook and WhatsApp Business Cloud, or delete one or both. As they stand, neither has a recipient or a
   message.
7. In ERPNext, add a webhook on the Job Applicant doctype that fires on insert. Its JSON body must carry name, Job_opening
   (the opening applied for, or the text None) and resume_link (a link ending in .pdf). Use the Webhook test address first.
8. Add a test applicant with a PDF link and an opening that has a description. Read the result in ERPNext, press Publish, then
   point ERPNext at the production address.

How an applicant runs

1. Webhook receives the applicant. Code adds a field called myNewField that nothing uses, and ApplicantData keeps the
   request body.
2. Resume Link Provided checks that body.resume_link starts with http. If not, ERPNext - Reject if Resume not Attached
   sets the status to Rejected and the run ends.
3. Applied Against Job checks that body.Job_opening is not the text None. If it is, ERPNext - Hold Applicant sets the
   status to Hold and the run ends.
4. Reume Attachment Link copies the link into body.resume_attachment, and File Type sorts it by its ending. Only the pdf
   output leads on.
5. Download PDF Resume fetches the file, PDF to Text reads its text and Merge1 passes it on.
6. Get Job Opening fetches the opening from ERPNext by name.
7. Recruitment AI Agent sends Gemini the job title, the description and the resume text, and asks for FitLevel, Score,
   Rating and Justification.
8. Convert to Fields reads the four values out of the answer, and Update Applicant Data saves them on the applicant.
9. If score less than 80 decides. Reject Applicant sets Rejected, and Accept Applicant sets Accepted. The empty
   Microsoft Outlook and WhatsApp Business Cloud nodes come after them.

Change the result

• The scoring: the prompt in Recruitment AI Agent. It says what to compare and how to grade.
• The pass mark: the 80 in If score less than 80.
• What a low score does: Reject Applicant. Set its status to Hold if a person should decide.
• What counts as a resume: Resume Link Provided, and the three rules in File Type.
• The model: Google Gemini Chat Model.
• A message to the applicant: fill in Microsoft Outlook or WhatsApp Business Cloud.

Good to know

• Google's deprecations page, read on 12 November 2026, shows the 2.0 flash family shut down on 1 June 2026, with
  gemini-3.6-flash as its replacement. It does not list the experimental gemini-2.0-flash-exp by name.
• A score under 80 sets Rejected with nobody reading it. Rules on automated hiring decisions differ by country and city, so
  check yours first, or point Reject Applicant at Hold.
• Microsoft Outlook and WhatsApp Business Cloud are empty. They have no recipient, subject or message, so fill them in or
  delete them before the first test.
• Frappe's Rating field stores a share from 0 to 1 (Frappe's rating control, read on 12 November 2026), and `Update Applicant
  Data` sends the model's 0 to 5 as it is. If the stars all show full, divide the rating by 5 in that node.
• Only .pdf links go on, and the match is case sensitive. A .doc or .jpg link ends the run with no change to the
  applicant, and a scanned PDF reaches Gemini with no text to score.
• Txt File to Text (Example) has no input. It is a sketch of a second file type.
• Webhook has no authentication, so anyone who learns its address can post a fake applicant. Add Header Auth in the node and
  in ERPNext.
• The resume text and the job description go to Google Gemini. Both can hold personal data.
• Each applicant that gets this far costs at least one Gemini call under your key's plan.