Update app/ranker.py

app/ranker.py

@@ -2,26 +2,31 @@
 import os
 from sentence_transformers import SentenceTransformer, util
 from .utils import extract_text_from_pdf
+import PyPDF2
 
 # Set safe cache directories for Hugging Face models (especially for read-only environments like Hugging Face Spaces)
 os.environ["HF_HOME"] = "/tmp/hf"
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/hf"
 os.environ["SENTENCE_TRANSFORMERS_HOME"] = "/tmp/hf"
 
-# Load the model globally
-model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
+
+model = SentenceTransformer('all-MiniLM-L6-v2')
+
+def extract_text_from_pdf(file_storage):
+    reader = PyPDF2.PdfReader(file_storage)
+    return " ".join([page.extract_text() or '' for page in reader.pages])
 
 def rank_resumes_by_semantic_similarity(resume_files, job_description):
-    resume_texts = [extract_text_from_pdf(f) for f in resume_files]
     job_embedding = model.encode(job_description, convert_to_tensor=True)
-
     results = []
-    for file, resume_text in zip(resume_files, resume_texts):
-        resume_embedding = model.encode(resume_text, convert_to_tensor=True)
-        similarity = util.cos_sim(job_embedding, resume_embedding).item()
+
+    for resume_file in resume_files:
+        text = extract_text_from_pdf(resume_file)
+        resume_embedding = model.encode(text, convert_to_tensor=True)
+        score = util.cos_sim(job_embedding, resume_embedding).item()
         results.append({
-            "filename": file.filename,
-            "score": round(similarity, 4)
+            "filename": resume_file.filename,
+            "score": round(score, 4)
         })
 
     results.sort(key=lambda x: x['score'], reverse=True)