Update funcs.py

funcs.py

@@ -96,74 +96,144 @@ def get_doc_response_emotions(user_message, therapy_session_conversation):
     print(f"therapy_session_conversation: {therapy_session_conversation}")
     return '', therapy_session_conversation, emotions_msg
 
-from transformers import AutoTokenizer, AutoModelForCausalLM
-import torch
+# from transformers import AutoTokenizer, AutoModelForCausalLM
+# import torch
+
+# # Load model once globally for reuse
+# tokenizer = AutoTokenizer.from_pretrained("tiiuae/Falcon-H1-1.5B-Deep-Instruct")
+# use_cuda = torch.cuda.is_available()
+# model = AutoModelForCausalLM.from_pretrained(
+#     "tiiuae/Falcon-H1-1.5B-Deep-Instruct",
+#     torch_dtype=torch.float16 if use_cuda else torch.float32,
+#     device_map="auto" if use_cuda else None
+# )
+# def generate_falcon_response(prompt, max_new_tokens=300):
+#     inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=1800).to(model.device)
+#     outputs = model.generate(
+#         **inputs,
+#         max_new_tokens=max_new_tokens,
+#         do_sample=True,
+#         temperature=0.7,
+#         top_p=0.9
+#     )
+#     decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
+#     return decoded_output[len(prompt):].strip()
+
+
+# def summarize_and_recommend(therapy_session_conversation):
+
+#     session_time = str(datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
+#     session_conversation = [item[0] for item in therapy_session_conversation]
+#     session_conversation = [x for x in session_conversation if x is not None]
+#     session_conversation.insert(0, "Session_time: " + session_time)
+    
+#     session_conversation_processed = '\n'.join(session_conversation)
+#     print("session_conversation_processed:", session_conversation_processed)
 
-# Load model once globally for reuse
-tokenizer = AutoTokenizer.from_pretrained("tiiuae/Falcon-H1-1.5B-Deep-Instruct")
-use_cuda = torch.cuda.is_available()
-model = AutoModelForCausalLM.from_pretrained(
-    "tiiuae/Falcon-H1-1.5B-Deep-Instruct",
-    torch_dtype=torch.float16 if use_cuda else torch.float32,
-    device_map="auto" if use_cuda else None
-)
-def generate_falcon_response(prompt, max_new_tokens=300):
-    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=1800).to(model.device)
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=max_new_tokens,
-        do_sample=True,
-        temperature=0.7,
-        top_p=0.9
-    )
-    decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    return decoded_output[len(prompt):].strip()
+#     # Summarization prompt
+#     summary_prompt = f"""You are an Expert Cognitive Behavioural Therapist and Precis writer.
+# Summarize STRICTLY the following session into concise, ethical, and clinically meaningful content.
 
+# Session:
+# {session_conversation_processed}
+
+# Format your response as:
+# Session Time:
+# Summary of the patient messages: 
+# Summary of therapist messages: 
+# Summary of the whole session: 
+# Ensure the entire summary is less than 300 tokens."""
+
+#     full_summary = generate_falcon_response(summary_prompt, max_new_tokens=300)
+#     print("\nFull summary:", full_summary)
+
+#     # Recommendation prompt
+#     recommendation_prompt = f"""You are an expert Cognitive Behavioural Therapist.
+# Based STRICTLY on the following summary, provide a clinically valid action plan for the patient.
+
+# Summary:
+# {full_summary}
+
+# Use this format:
+# - The patient is referred to...
+# - The patient is advised to...
+# - The patient is refrained from...
+# - It is suggested that the patient...
+# - Scheduled a follow-up session with the patient...
+
+# Ensure the list contains NOT MORE THAN 7 points and is in passive voice with proper tense."""
+
+#     full_recommendations = generate_falcon_response(recommendation_prompt, max_new_tokens=400)
+#     print("\nFull recommendations:", full_recommendations)
+
+#     chatbox = []
+#     return full_summary, full_recommendations, chatbox
 
-def summarize_and_recommend(therapy_session_conversation):
 
-    session_time = str(datetime.now().strftime("%Y-%m-%d %H:%M:%S"))
-    session_conversation = [item[0] for item in therapy_session_conversation]
-    session_conversation = [x for x in session_conversation if x is not None]
+import requests
+from datetime import datetime
+import os
+
+HF_API_TOKEN = os.getenv('hf_token')  # 🔁 Replace with your HF token
+API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
+HEADERS = {"Authorization": f"Bearer {HF_API_TOKEN}"}
+
+def query_huggingface(prompt):
+    payload = {
+        "inputs": prompt,
+        "parameters": {
+            "max_new_tokens": 400,
+            "temperature": 0.7,
+            "top_p": 0.9,
+            "return_full_text": False
+        }
+    }
+    response = requests.post(API_URL, headers=HEADERS, json=payload)
+    response.raise_for_status()
+    return response.json()[0]['generated_text'].strip()
+
+def summarize_and_recommend(therapy_session_conversation):
+    session_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    session_conversation = [item[0] for item in therapy_session_conversation if item[0]]
     session_conversation.insert(0, "Session_time: " + session_time)
-    
     session_conversation_processed = '\n'.join(session_conversation)
     print("session_conversation_processed:", session_conversation_processed)
 
-    # Summarization prompt
-    summary_prompt = f"""You are an Expert Cognitive Behavioural Therapist and Precis writer.
-Summarize STRICTLY the following session into concise, ethical, and clinically meaningful content.
-
-Session:
-{session_conversation_processed}
+    summary_prompt = f"""
+You are an Expert Cognitive Behavioural Therapist and Precis writer.
+Summarize STRICTLY the below session content ONLY into useful, ethical, relevant and realistic phrases in this format:
 
-Format your response as:
-Session Time:
+Session Time: {session_time}
 Summary of the patient messages: 
-Summary of therapist messages: 
+Summary of therapist messages:
 Summary of the whole session: 
-Ensure the entire summary is less than 300 tokens."""
 
-    full_summary = generate_falcon_response(summary_prompt, max_new_tokens=300)
+Conversation:
+{session_conversation_processed}
+"""
+
+    full_summary = query_huggingface(summary_prompt)
     print("\nFull summary:", full_summary)
 
-    # Recommendation prompt
-    recommendation_prompt = f"""You are an expert Cognitive Behavioural Therapist.
-Based STRICTLY on the following summary, provide a clinically valid action plan for the patient.
+    recommendation_prompt = f"""
+You are an expert Cognitive Behavioural Therapist.
+Based STRICTLY on the following summary ONLY, provide a clinically valid and appropriate action plan for the patient as a bullet list.
+Include both medical and non-medical suggestions in passive voice.
 
 Summary:
 {full_summary}
 
-Use this format:
+Format:
 - The patient is referred to...
 - The patient is advised to...
 - The patient is refrained from...
 - It is suggested that the patient...
 - Scheduled a follow-up session with the patient...
 
-Ensure the list contains NOT MORE THAN 7 points and is in passive voice with proper tense."""
+(Ensure not more than 7 bullet points)
+"""
 
-    full_recommendations = generate_falcon_response(recommendation_prompt, max_new_tokens=400)
+    full_recommendations = query_huggingface(recommendation_prompt)
     print("\nFull recommendations:", full_recommendations)
 
     chatbox = []