cv_analysis_and_candidate_evaluation.json

Overview

The cv_analysis_and_candidate_evaluation.json file defines a conversational workflow designed to assist Human Resources (HR) professionals in evaluating candidate resumes against a specified job description (JD). This workflow facilitates batch processing of resumes submitted through a chat interface by leveraging a specialized AI agent to extract standardized candidate information and perform JD matching analysis.

The core functionality is to streamline large-scale resume screening by automatically extracting key data points from each resume, normalizing this data, and scoring the alignment between each candidate’s resume and the job description. The output is a clear, hierarchical text analysis in Chinese, formatted to separate individual resume evaluations.

Detailed Components and Their Explanations

The workflow is modeled using a domain-specific language (DSL) that describes components, their parameters, and relationships. The key components are:

1. `begin` (Component: Begin)

Purpose: Entry point for the workflow where HR inputs the Job Description.
Parameters:
- enablePrologue (boolean): Enables an introductory message.
- inputs:
  - JD (line, required): The Job Description text.
- mode (string): "conversational" - enables chat-based interaction.
- prologue (string): Welcoming message guiding HR to upload JD and resumes.
Usage: The HR user starts the process here by submitting the job description.

2. Iteration:PetiteBanksWarn (Component: Iteration)

Purpose: Iterates over the collection of uploaded resumes (sys.files) to process each resume individually.
Parameters:
- items_ref (reference): Points to the list of resumes uploaded into system files.
- outputs:
  - evaluation (Array): Collects evaluation results from the Agent's output.
Interaction: Receives the Job Description from begin and processes each resume through IterationItem and Agent.

3. `IterationItem:EagerGiftsOpen` (Component: IterationItem)

Purpose: Represents a single resume item in the iteration.
Parameters:
- outputs:
  - index (integer): The position of the current resume in the list.
  - item (unknown type): The actual resume data.
Interaction: Sends the current resume to the Agent component for evaluation.

4. Agent:AfraidBearsShare (Component: Agent)

Purpose: The core AI-powered resume evaluation agent.
Parameters:
- llm_id: Specifies the large language model identifier (deepseek-chat@DeepSeek).
- temperature: Set to 0.1 for focused, less random output.
- max_tokens: 4096 tokens max for response length.
- prompts: Template prompt for the AI, injecting the HR query, Job Description, and current resume.
- sys_prompt: Detailed system instructions guiding the agent behavior.
- outputs:
  - content (string): The agent's evaluation result for the resume.
- Various penalty and retry controls for managing AI response quality.
Functionality:
- Extracts 6 key standardized data points from each resume.
- Normalizes candidate information.
- Scores the JD-match on a 10-point scale.
- Provides a brief matching analysis highlighting alignments and mismatches.
- Responds exclusively in Chinese.
- Ensures output formatting complies with length and separation requirements.

Usage Example:

HR is asking about: {sys.query}

Job description is {begin@JD}

Resume is {IterationItem:EagerGiftsOpen@item}

The agent processes this prompt and returns the structured evaluation.

5. `Message:TenLizardsShake` (Component: Message)

Purpose: Outputs the evaluation results from the Agent component back to the user.
Parameters:
- content: References the agent’s content output, presenting the evaluation text.
Interaction: Receives processed evaluation content from the Agent and displays it as the final message output.

Important Implementation Details and Algorithms

Batch Processing via Iteration: The workflow employs an iteration component to handle multiple resumes in sequence, ensuring isolated evaluation for each resume.
AI-Prompt Engineering: The agent is driven by a comprehensive system prompt that defines mission, capabilities, output format, and workflow steps. This ensures consistent and high-quality extraction and evaluation.
Resume Data Extraction and Normalization: The agent extracts six key data points per resume (e.g., full name, contact info, education, experience years, residence city) with confidence levels to ensure data quality.
JD Matching Scoring: The agent scores alignment on a 10-point scale, detailing strengths and gaps between the resume and job description.
Language and Formatting Constraints: The agent always responds in Chinese, separates resume information with paragraphs, avoids subheadings, and respects strict line formatting to keep individual resumes distinct.
Error Handling: The agent includes parameters for retrying requests and error delays to improve robustness.

Interaction with Other System Components

Input: HR uploads a Job Description and multiple resumes (files) into the system.
Processing: The iteration component loops through resumes, sending each to the AI agent for analysis.
Output: The evaluation results are compiled and presented via message components.
Integration Potential: The agent can save results to external services such as Google Sheets via MCP (Multi-Channel Pipeline) for record-keeping or further analysis.
User Interface: The workflow is designed for a conversational chat interface, allowing HR users to submit files and receive evaluations interactively.

Usage Workflow Summary

Begin: HR uploads the Job Description.
Iteration: The system iterates over all uploaded resumes.
Agent Evaluation: Each resume is analyzed by the AI agent for data extraction and JD matching.
Message Output: The evaluation result for each resume is output separately.
Optional Storage: Results can be forwarded to Google Sheets or other systems.

Visual Diagram: Component Flowchart of the Workflow

flowchart LR
    Begin["Begin (Job Description Input)"]
    Iteration["Iteration (Process Resumes)"]
    IterationItem["IterationItem (Single Resume)"]
    Agent["Agent (Resume Evaluation)"]
    Message["Message (Output Evaluation)"]

    Begin --> Iteration
    Iteration --> IterationItem
    IterationItem --> Agent
    Agent --> Message

Explanation: The flowchart illustrates the sequential flow: starting with the Job Description input, iterating over resumes, passing each resume to the AI agent for evaluation, and finally delivering the evaluation message.

Summary

The cv_analysis_and_candidate_evaluation.json is a carefully structured workflow designed to automate and standardize the process of resume evaluation against a job description using AI-powered language models. It supports batch processing, ensures consistent data extraction and matching analysis, and outputs user-friendly evaluations in Chinese within a conversational interface. This file serves as a critical component for HR teams looking to scale and optimize candidate screening.

Appendix: Key Parameters and Concepts

Component	Key Parameters/Outputs	Description
Begin	`JD` (Job Description input)	Starting point for the workflow
Iteration	`items_ref` = `sys.files`, `evaluation` output	Iterates over all uploaded resumes
IterationItem	index, `item`	Represents each resume in the iteration
Agent	`llm_id`, `prompts`, `sys_prompt`, `content` output	AI agent performing resume data extraction and evaluation
Message	`content` referencing Agent output	Displays the evaluation results

If you require further integration details or examples of AI prompt usage, please ask!