All templates By NappAI team CV Scorer
Process where a job position is specified and the available CVs are scored based on each one's value for performing that job.
graph TD
%%{init: {'theme': 'mc','layout': 'elk'}}%%
ChatOutput-h6ga5[<div><img alt="logo" src="/_astro/messages-square.BaSDmT6g.svg" style="height: 20px !important;width: 20px !important"/></div>Chat Output]
style ChatOutput-h6ga5 stroke:#a170ff
TextInput-wm8fq[<div><img alt="logo" src="/_astro/type.Dy26vmDy.svg" style="height: 20px !important;width: 20px !important"/></div>Job Description]
style TextInput-wm8fq stroke:#a170ff
GDriveFilesComponent-ozzfk[<div><img alt="logo" src="/_astro/google_drive.wKmDsV2c.svg" style="height: 20px !important;width: 20px !important"/></div>Google Drive File Manager]
style GDriveFilesComponent-ozzfk stroke:#a170ff
OpenAIModel-sjaw0[<div><img alt="logo" src="/_astro/openAI.BhmuxEs3.svg" style="height: 20px !important;width: 20px !important"/></div>OpenAI]
style OpenAIModel-sjaw0 stroke:#a170ff
AdvancedAgent-qf7fp[Agent]
style AdvancedAgent-qf7fp stroke:#a170ff
TextInput-wm8fq -.- AdvancedAgent-qf7fp
linkStyle 0 stroke:#a170ff
AdvancedAgent-qf7fp -.- ChatOutput-h6ga5
linkStyle 1 stroke:#a170ff
OpenAIModel-sjaw0 -.- AdvancedAgent-qf7fp
linkStyle 2 stroke:#a170ff
GDriveFilesComponent-ozzfk -.- AdvancedAgent-qf7fp
linkStyle 3 stroke:#a170ff
📄 CV Scorer Workflow Documentation
🧩 Overview
The CV Scorer workflow automatically evaluates a list of candidate CVs stored in a Google Drive folder against a specified job description. An AI agent extracts candidate information, applies a weighted scoring rubric, and produces a ranked list of candidates with detailed justification. The final output is displayed in a chat interface, making it easy for recruiters to review the recommendations.
⚙️ Main Features
- Accepts a job description as free‑form text.
- Retrieves all CV files from a designated Google Drive folder.
- Uses a pre‑trained OpenAI model to parse CVs and compare them to the job criteria.
- Calculates a compatibility score (1‑10) and ranks candidates accordingly.
- Generates a structured, markdown‑formatted summary for each candidate.
- Presents the ranked results directly in a chat window.
🔄 Workflow Steps
| Component Name | Role in the Workflow | Key Inputs | Key Outputs |
|---|---|---|---|
| Text Input (Job Description) | Supplies the job description that defines the evaluation criteria. | Text containing the position details (title, responsibilities, qualifications). | Message with the job description text. |
| Google Drive File Manager | Accesses all CV files in a specified Drive folder. | Operation: Get, Folder: CVs, file‑selection parameters. |
Tool that provides file data to the agent. |
| OpenAI Model | Generates language model output and supplies the agent with LLM capabilities. | Prompt text, model name (gpt‑4o‑mini), temperature, token limits. |
Text output (for debugging) and LanguageModel instance. |
| Advanced Agent | Orchestrates the scoring logic, invokes tools, and formulates the final answer. | Job description (as user prompt), LLM, Google Drive tool. | Chat‑formatted response containing candidate rankings and justifications. |
| Chat Output | Displays the agent’s answer in the playground chat interface. | The final response message from the agent. | Visible chat message. |
Sequence: Text Input → Advanced Agent (via LLM and Drive tool) → Chat Output.
🧠Notes
- The agent is powered by a custom system prompt that defines evaluation criteria, scoring weights, and output formatting.
- Google Drive File Manager is used only in “Get” mode; it pulls every file in the designated folder but does not modify or delete any content.
- The scoring rubric is hard‑coded in the agent’s system prompt: technical match (40 %), experience relevance (35 %), education (15 %), and transversal skills (10 %).
- Output is strictly markdown‑formatted to ensure consistency across all candidates.
- The workflow assumes that each CV is a PDF or document that the OpenAI model can read; extremely large files may exceed token limits.
- No personal data is logged beyond the agent’s internal state; however, the tool will transmit file content to the OpenAI endpoint, so sensitive information should be handled in compliance with applicable data‑privacy regulations.
- The agent can be extended with additional tools or adjusted scoring logic by editing the system prompt or adding new tool definitions.