Add Prompt To AI For Resume Scoring

Signed-off-by: Tal Jacob <taljacob2@gmail.com>

Author: taljacob2

nextstep-backend/src/controllers/resources_controller.ts

@@ -7,7 +7,6 @@ import multer from 'multer';
 import {CustomRequest} from "types/customRequest";
 import {updateUserById} from "../services/users_service";
 import {handleError} from "../utils/handle_error";
-import { scoreResume } from '../services/resume_service';
 
 const createUserImageResource = async (req: CustomRequest, res: Response) => {
     try {

nextstep-backend/src/controllers/resume_controller.ts

@@ -17,15 +17,14 @@ const getResumeScore = async (req: Request, res: Response) => {
             return res.status(404).send('Resume not found');
         }
 
-        const score = await scoreResume(resumePath, jobDescription);
-        return res.status(200).json({ score });
+        const scoreAndFeedback = await scoreResume(resumePath, jobDescription);
+        return res.status(200).send(scoreAndFeedback);
     } catch (error) {
         handleError(error, res);
     }
 };
 
 
-
 export default {
     getResumeScore,
 }; 

nextstep-backend/src/services/chat_api_service.ts

@@ -7,20 +7,20 @@ const cleanResponse = (response: string): string => {
 
 
 
-export const chatWithAI = async (inputUserMessage: string)=> {
+export const chatWithAI = async (userMessageContent: string, systemMessageContent: string)=> {
     try {
         const API_URL = config.chatAi.api_url();
         const API_KEY = config.chatAi.api_key();
         const MODEL_NAME = config.chatAi.model_name();
 
         const systemMessage = {
             role: 'system',
-            content: 'You are an AI assistant tasked with providing the first comment on forum posts. Your responses should be relevant, engaging, and encourage further discussion, also must be short, and you must answer if you know the answer. Ensure your comments are appropriate for the content and tone of the post. Also must answer in the language of the user post. answer short answers. dont ask questions to follow up'
+            content: systemMessageContent
         };
 
         const userMessage = {
             role: 'user',
-            content: inputUserMessage
+            content: userMessageContent
         };
 
 

nextstep-backend/src/services/posts_service.ts

@@ -32,7 +32,7 @@ const addMisterAIComment = async (postId: string, postContent: string) => {
         misterAI = await usersService.addUser('misterai', 'securepassword', 'misterai@example.com', 'local');
     }
 
-    const comment = await chatService.chatWithAI(postContent);
+    const comment = await chatService.chatWithAI(postContent, 'You are an AI assistant tasked with providing the first comment on forum posts. Your responses should be relevant, engaging, and encourage further discussion, also must be short, and you must answer if you know the answer. Ensure your comments are appropriate for the content and tone of the post. Also must answer in the language of the user post. answer short answers. dont ask questions to follow up');
     const commentData: CommentData = { postId, owner: misterAI.id, content: comment };
     const savedComment = await commentsService.addComment(commentData);
     return savedComment;

nextstep-backend/src/services/resume_service.ts

@@ -2,20 +2,73 @@ import { config } from '../config/config';
 import path from 'path';
 import fs from 'fs';
 import axios from 'axios';
+import { chatWithAI } from './chat_api_service';
 
-const scoreResume = async (resumePath: string, jobDescription?: string): Promise<number> => {
+const systemTemplate = `You are a very experienced ATS (Application Tracking System) bot with a deep understanding named Bob the Resume builder.
+You will review resumes with or without job descriptions.
+You are an expert in resume evaluation and provide constructive feedback with dynamic evaluation.
+You should also provide an improvement table, taking into account:
+- Content (Medium priority)
+- Keyword matching (High priority)
+- Hard skills (High priority)
+- Soft skills (High priority)
+- Overall presentation (Low priority)`;
+
+const feedbackTemplate = (resumeText: string, jdText: string) => `
+Resume Feedback Report
+Here is the resume you provided:
+${resumeText}
+And the job description:
+${jdText}
+
+Create the Improvement Table in relevance to the resume and give the consideration and suggestion for each section strictly following 
+the pattern as below and don't just out this guided pattern :
+| Area          | Consideration                                                   | Status | Suggestions |
+| ------------- | --------------------------------------------------------------- | ------ | ----------- |
+| Content       | Measurable Results: At least 5 specific achievements or impact. |  Low   |             |
+|               | Words to avoid: Negative phrases or clichés.                    |        |             |
+| Keywords      | Hard Skills: Presence and frequency of hard skills.             |  High  |             |
+|               | Soft Skills: Presence and frequency of soft skills.             |        |             |
+| Presentation  | Education Match: Does the resume list a degree that matches the job requirements? |  High   |             |
+
+Strengths:
+List the strengths of the resume here.
+
+Detailed Feedback:
+Provide detailed feedback on the resume's content, structure, grammar, and relevance to the job description.
+
+Suggestions:
+Provide actionable suggestions for improvement, including specific keywords to include and skills to highlight.
+
+Based on your analysis, provide a numerical score between 0-100 that represents the overall quality and match of the resume.
+The score should be provided at the end of your response in the format: "SCORE: X" where X is the numerical score.
+`;
+
+const scoreResume = async (resumePath: string, jobDescription?: string): Promise<{ score: number; feedback: string }> => {
     try {
-        // Here you would integrate with an actual ATS system
-        // For now, we'll return a mock score
-        // In a real implementation, you would:
-        // 1. Parse the resume text
-        // 2. Compare it with the job description
-        // 3. Calculate a score based on keywords, skills, experience, etc.
+        // Read the resume file
+        const resumeText = fs.readFileSync(resumePath, 'utf-8');
+        
+        // Prepare the prompt for the AI
+        const prompt = feedbackTemplate(resumeText, jobDescription || 'No job description provided.');
+        
+        let feedback = 'The Chat AI feature is turned off. Could not score your resume.';
+
+        if (config.chatAi.turned_on()) {
+            // Get feedback from the AI
+            feedback = await chatWithAI(prompt, systemTemplate);   
+        }
+        
+        // Extract the score from the feedback
+        const scoreMatch = feedback.match(/SCORE: (\d+)/);
+        const score = scoreMatch ? parseInt(scoreMatch[1]) : 0;
 
-        // Mock implementation
-        const score = Math.floor(Math.random() * 100);
-        return score;
+        return {
+            score,
+            feedback
+        };
     } catch (error) {
+        console.error('Error scoring resume:', error);
         throw new Error('Failed to score resume');
     }
 };