Update app.py

app.py

@@ -3,8 +3,7 @@ from pydantic import BaseModel
 from sentence_transformers import SentenceTransformer
 import faiss
 import pandas as pd
-import random
-from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
 
 app = FastAPI()
 
@@ -14,11 +13,6 @@ embedding_model = SentenceTransformer("all-MiniLM-L6-v2")
 summarization_model = AutoModelForSeq2SeqLM.from_pretrained("google/long-t5-tglobal-base")
 summarization_tokenizer = AutoTokenizer.from_pretrained("google/long-t5-tglobal-base")
 
-# New: Load Local LLM for Dynamic Emotional Responses (Mistral/Llama)
-response_model_name = "mistralai/Mistral-7B-Instruct"
-response_tokenizer = AutoTokenizer.from_pretrained(response_model_name)
-response_model = AutoModelForCausalLM.from_pretrained(response_model_name)
-
 # Load datasets
 recommendations_df = pd.read_csv("treatment_recommendations.csv")
 questions_df = pd.read_csv("symptom_questions.csv")
@@ -40,38 +34,13 @@ class ChatRequest(BaseModel):
 class SummaryRequest(BaseModel):
     chat_history: list  # List of messages
 
-
 @app.post("/get_questions")
 def get_recommended_questions(request: ChatRequest):
-    """Retrieve the most relevant diagnostic questions with a dynamically generated conversational response."""
-    
-    # Step 1: Encode the input message for FAISS search
+    """Retrieve the most relevant diagnostic questions."""
     input_embedding = embedding_model.encode([request.message], convert_to_numpy=True)
     distances, indices = question_index.search(input_embedding, 3)
-
-    # Step 2: Retrieve the top 3 relevant questions
     retrieved_questions = [questions_df["Questions"].iloc[i] for i in indices[0]]
-
-    # Step 3: Use a local LLM to generate context-aware empathetic responses
-    prompt = f"""
-    User: {request.message}
-    
-    You are a compassionate psychiatric assistant. Before asking a diagnostic question, respond empathetically.
-    
-    Questions:
-    1. {retrieved_questions[0]}
-    2. {retrieved_questions[1]}
-    3. {retrieved_questions[2]}
-    
-    Generate a conversational response that introduces each question naturally.
-    """
-
-    inputs = response_tokenizer(prompt, return_tensors="pt")
-    output = response_model.generate(**inputs, max_length=300)
-    enhanced_responses = response_tokenizer.decode(output[0], skip_special_tokens=True).split("\n")
-
-    return {"questions": enhanced_responses}
-
+    return {"questions": retrieved_questions}
 
 @app.post("/summarize_chat")
 def summarize_chat(request: SummaryRequest):
@@ -82,7 +51,6 @@ def summarize_chat(request: SummaryRequest):
     summary = summarization_tokenizer.decode(summary_ids[0], skip_special_tokens=True)
     return {"summary": summary}
 
-
 @app.post("/detect_disorders")
 def detect_disorders(request: SummaryRequest):
     """Detect psychiatric disorders from full chat history at the end."""
@@ -92,7 +60,6 @@ def detect_disorders(request: SummaryRequest):
     disorders = [recommendations_df["Disorder"].iloc[i] for i in indices[0]]
     return {"disorders": disorders}
 
-
 @app.post("/get_treatment")
 def get_treatment(request: SummaryRequest):
     """Retrieve treatment recommendations based on detected disorders."""
@@ -102,3 +69,4 @@ def get_treatment(request: SummaryRequest):
         for disorder in detected_disorders
     }
     return {"treatments": treatments}
+