UPDATE: antigravity test

prompts.yaml

@@ -2,7 +2,7 @@ prompt: |
   Act as a Senior Resume Writer and ATS Optimization Specialist with 20+ years of experience in recruitment.
 
   I am providing you with:
-  1. A Current Resume (text or details below).
+  1. A Current Resume (attached as a file).
   2. A Target Job Description (JD).
   3. The Current Date.
 
@@ -26,6 +26,7 @@ prompt: |
      - Email Rule: You MUST replace the candidate's email domain with @atomicmail.io.
        Example: john.doe@gmail.com becomes john.doe@atomicmail.io.
      - Do not alter any other personal contact details.
+     - ONLY include LinkedIn, GitHub, or Portfolio links if they explicitly appear in the original resume. Do not invent or add them. If they are missing, omit the corresponding `\mbox{...}` and `\AND` separators in the header.
 
   4. Formatting Constraints:
      - The output must be ONLY a code block containing the LaTeX code.
@@ -238,8 +239,7 @@ prompt: |
   Target Job Description:
   {{JOB_DESCRIPTION}}
 
-  Current Resume Details:
-  {{CURRENT_RESUME}}
+
 
   ### OUTPUT INSTRUCTIONS
   Generate ONLY the LaTeX code in a code block. Replace all placeholders with actual content from the resume. Ensure everything is aligned to the job description and immediately compilable.

pyproject.toml

@@ -1,14 +0,0 @@
-[project]
-name = "raftecresumetool"
-version = "0.1.0"
-description = "Add your description here"
-readme = "README.md"
-requires-python = ">=3.13"
-dependencies = [
-    "gradio>=6.1.0",
-    "langchain>=1.1.3",
-    "langchain-core>=1.2.0",
-    "langchain-google-genai>=4.0.0",
-    "pdfplumber>=0.11.8",
-    "pyyaml>=6.0.3",
-]

resumeRewritingService.py

@@ -1,72 +1,59 @@
-from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain_core.messages import HumanMessage
+import google.generativeai as genai
 from utils import readPrompts
 import os
-import base64
-import pdfplumber
-from io import BytesIO
-
+import re
 
 class ResumeEditService:
     def __init__(self):
         prompts = readPrompts(path=os.path.join(os.getcwd(), "prompts.yaml"))
         self.prompt_template = prompts.get("prompt")
-        self.llm = self._initialize_llm()
-    
-    def _initialize_llm(self):
-        """Initialize the Google Generative AI model"""
-        return ChatGoogleGenerativeAI(
-            model="gemini-3-flash-preview",
-            temperature=0.3,
-            google_api_key=os.getenv("GOOGLE_API_KEY")
-        )
-    
-    def _extract_text_from_pdf_base64(self, base64_pdf: str) -> str:
-        """
-        Extract text from base64 encoded PDF
-        
-        Args:
-            base64_pdf: Base64 encoded PDF string
-        
-        Returns:
-            Extracted text from the PDF
-        """
-        # Decode base64 to bytes
-        pdf_bytes = base64.b64decode(base64_pdf)
         
-        # Create a BytesIO object
-        pdf_file = BytesIO(pdf_bytes)
-        
-        # Extract text using PdfPlumber
-        text = ""
-
-        with pdfplumber.open(pdf_file) as pdf:
-            for page in pdf.pages:
-                text += page.extract_text() or ""
-        
-        return text.strip()
+        # Configure the Gemini API
+        genai.configure(api_key=os.getenv("GOOGLE_API_KEY"))
+        self.model = genai.GenerativeModel('gemini-1.5-flash')
     
     def generate_optimized_resume(self, resume_base64: str, job_description: str, current_date: str) -> str:
         """
-        Generate optimized resume based on job description
+        Generate optimized resume based on job description using Gemini with file attachment.
         
         Args:
-            resume_base64: Base64 encoded PDF resume
+            resume_base64: Base64 encoded PDF/DOCX resume
             job_description: The target job description
             current_date: Current date for timeline adjustments
         
         Returns:
-            Raw response from the LLM (LaTeX code in markdown format)
+            Optimized resume in LaTeX format (code fences stripped)
         """
-        # Extract text from PDF
-        current_resume = self._extract_text_from_pdf_base64(resume_base64)
         
-        # Manual string replacement (avoiding template parsing issues)
+        # Prepare the prompt
         prompt = self.prompt_template.replace("{{CURRENT_DATE}}", current_date)
         prompt = prompt.replace("{{JOB_DESCRIPTION}}", job_description)
-        prompt = prompt.replace("{{CURRENT_RESUME}}", current_resume)
         
-        # Invoke the LLM directly
-        response = self.llm.invoke([HumanMessage(content=prompt)])
+        # Prepare the inline data for the file attachment
+        # Assuming PDF for now based on app.py, but could support others.
+        # The app.py reads the file and base64 encodes it without a mime prefix.
+        # We need to construct the part for the model.
+        
+        file_part = {
+            "mime_type": "application/pdf",
+            "data": resume_base64
+        }
         
-        return response.content[0]["text"]
+        # Generate content
+        response = self.model.generate_content([prompt, file_part])
+        
+        # Extract text
+        response_text = response.text
+        
+        # Strip markdown code fences if present
+        cleaned_text = self._strip_code_fences(response_text)
+        
+        return cleaned_text
+
+    def _strip_code_fences(self, text: str) -> str:
+        """Removes markdown code fences from the text."""
+        # Remove opening fence (e.g. ```latex, ```tex, ```)
+        text = re.sub(r'^```\w*\s*', '', text.strip())
+        # Remove closing fence
+        text = re.sub(r'\s*```$', '', text.strip())
+        return text.strip()