Update app.py

app.py

@@ -66,41 +66,73 @@ def call_zephyr(prompt):
         return f"Zephyr API error: {result['error']}"
     return result[0].get("generated_text", "")
 
-def get_candidates_by_category(category_name, job_titles):
+def filter_candidates_by_category(category_name, job_titles):
+    """
+    Step 1: Filter candidates based on actual work experience roles.
+    """
     data = fetch_json_local(JSON_FILE)
-    candidates = []
+    filtered = []
     for person in data:
         work_exps = person.get("work_experiences", [])
         if len(work_exps) == 0:
             continue
-        if any("full stack developer" in exp.get("roleName","").lower() for exp in work_exps):
+
+        # Exclude candidates who ONLY have Full Stack roles
+        non_fullstack_roles = [exp.get("roleName") for exp in work_exps if "full stack developer" not in exp.get("roleName","").lower()]
+        if not non_fullstack_roles:
             continue
 
+        # Include if any role matches the category
+        if any(role in job_titles for role in non_fullstack_roles):
+            filtered.append(person)
+    return filtered
+
+def get_final_recommendations(category_name, job_titles, top_n=5):
+    """
+    Step 2: Use Zephyr LLM for final recommendation and pick top N candidates.
+    """
+    candidates = filter_candidates_by_category(category_name, job_titles)
+    recommended = []
+
+    for person in candidates:
         prompt = f"""
-You are an HR assistant. Determine if the following candidate is suitable for the category '{category_name}'.
-The category is defined by the job titles: {job_titles}
+You are an HR assistant. Review this candidate and determine if they are suitable for the category '{category_name}'.
+The category includes the following job titles: {job_titles}
 
 Candidate JSON: {json.dumps(person)}
 
-Respond only 'Yes' or 'No'.
+Based on their work experience, skills, and education, respond only 'Yes' if suitable, otherwise 'No'.
 """
-        llm_response = call_zephyr(prompt)
-        if llm_response and "Yes" in llm_response:
-            candidates.append({
+        response = call_zephyr(prompt)
+        if response and "Yes" in response:
+            work_exps = person.get("work_experiences", [])
+            non_fullstack_roles = [exp.get("roleName") for exp in work_exps if "full stack developer" not in exp.get("roleName","").lower()]
+            recommended.append({
                 "Name": person.get("name"),
                 "Email": person.get("email"),
                 "Phone": person.get("phone"),
                 "Location": person.get("location"),
-                "Roles": ", ".join([exp.get("roleName") for exp in work_exps]),
+                "Roles": ", ".join(non_fullstack_roles),
                 "Skills": ", ".join(person.get("skills", [])),
                 "Salary": person.get("annual_salary_expectation", {}).get("full-time", "N/A")
             })
 
-    if len(candidates) == 0:
+    if len(recommended) == 0:
         return pd.DataFrame()
 
-    df = pd.DataFrame(candidates)
-    return df
+    df = pd.DataFrame(recommended)
+
+    # Optional: Sort by salary if available (ascending)
+    def parse_salary(s):
+        if isinstance(s, str) and s.startswith("$"):
+            return float(s.replace("$","").replace(",",""))
+        return float('inf')
+    if "Salary" in df.columns:
+        df["Salary_sort"] = df["Salary"].apply(parse_salary)
+        df = df.sort_values("Salary_sort")
+        df = df.drop(columns=["Salary_sort"])
+
+    return df.head(top_n)  # return top N candidates
 
 # ----------------------------
 # GRADIO INTERFACE
@@ -108,9 +140,7 @@ Respond only 'Yes' or 'No'.
 def run_dashboard(category):
     if category not in CATEGORIES:
         return pd.DataFrame()
-    df = get_candidates_by_category(category, CATEGORIES[category])
-    if df.empty:
-        return pd.DataFrame()
+    df = get_final_recommendations(category, CATEGORIES[category], top_n=5)
     return df
 
 category_options = list(CATEGORIES.keys())
@@ -118,10 +148,10 @@ category_options = list(CATEGORIES.keys())
 demo = gr.Interface(
     fn=run_dashboard,
     inputs=gr.Dropdown(category_options, label="Select Category"),
-    outputs=gr.Dataframe(label="Suitable Candidates"),
+    outputs=gr.Dataframe(label="Top 5 Recommended Candidates"),
     live=False,
     title="Startup Candidate Dashboard - Zephyr-7B-Beta",
-    description="Select a category to view suitable candidates."
+    description="View top 5 final recommended candidates filtered by category using Zephyr LLM."
 )
 
 if __name__ == "__main__":