Initial clean commit with hosted model

.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -0,0 +1,21 @@
+# Base image
+FROM python:3.10-slim
+
+# Set working directory
+WORKDIR /app
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y git
+
+# Copy files
+COPY . .
+
+# Install Python dependencies
+RUN pip install --upgrade pip
+RUN pip install -r requirements.txt
+
+# Expose Streamlit default port
+EXPOSE 8501
+
+# Run the app
+CMD ["streamlit", "run", "app.py", "--server.port=8501", "--server.enableCORS=false"]

README.md

@@ -0,0 +1,122 @@
+---
+title: Counselor Assistant
+emoji: 👀
+colorFrom: red
+colorTo: blue
+sdk: streamlit
+sdk_version: 1.44.1
+app_file: app.py
+pinned: false
+license: mit
+---
+# 🧠 Mental Health Counselor Assistant
+
+This project is a **counselor-facing AI assistant** designed to support mental health professionals by offering helpful suggestions in response to patient messages. It detects crisis and violent messages, generates supportive or informative suggestions, and provides logging and export features.
+
+---
+
+## 📘 How to use this app
+
+1. Enter a message from a patient in the text box.
+2. The app will generate helpful suggestions for how the counselor might reply.
+3. If a safety risk is detected (e.g., crisis or violence), a safe response will be shown instead.
+4. You can save the conversation history as a CSV file.
+
+---
+
+## 🧩 Key Features
+
+- ✅ Fine-tuned classifier (DistilBERT) for predicting intent (advice, question, validation, information)
+- 🔁 Streaming text generation using Flan-T5-XL
+- 🔍 Safety detection for crisis, violence, and toxicity
+- 💡 Suggestion prompts tailored to each response type
+- 💾 Option to save conversation logs
+- 📸 Screenshots for walkthrough
+
+---
+
+## 🧠 Model Architecture
+
+- **Response Type Classifier**: `DistilBERT` fine-tuned on a labeled response dataset.
+- **Text Generator**: `google/flan-t5-xl` used to generate helpful, human-like counselor suggestions.
+- **Safety Filter**: Includes custom regex patterns and `Detoxify` toxicity detector.
+
+---
+
+## 🧪 Datasets Used
+
+- `counselchat-data.csv` – Labeled examples of counselor replies
+- `pair_data.csv` – Counselor-patient message pairs
+- `Kaggle_Mental_Health_Conversations_train.csv`
+- `distilbert_labeled_responses.csv` – Augmented training set
+- `cleaned_combined_mental_health_dataset.csv`
+
+---
+
+## 📁 Project Structure
+
+```bash
+counselor-assistant/
+├── app.py                     # Main Streamlit app
+├── Dockerfile                 # Containerization for deployment
+├── requirements.txt           # Python dependencies
+├── pedal.yaml                 # Optional: Pedal config
+├── datasets/                  # Training + testing datasets
+├── models/                    # Fine-tuned model + tokenizer
+├── notebooks/                 # Data prep + training notebooks
+├── logs/                      # Saved chat logs
+├── screenshots/               # App walkthrough examples
+├── cache/                     # Prompt-response cache
+└── alert/                     # Alert logs (for flagged inputs)
+```
+
+---
+
+## 🐳 Docker Deployment
+
+To build and run the app locally:
+
+```bash
+docker build -t counselor-assistant .
+docker run -p 7860:7860 counselor-assistant
+```
+
+Then navigate to `http://localhost:7860` in your browser.
+
+---
+
+## ☁️ Hugging Face Deployment
+
+You can deploy to [Hugging Face Spaces](https://huggingface.co/spaces/) by:
+
+1. Adding this repo to a Space with SDK = `streamlit`
+2. Uploading the model folder to the Space (or hosting it via Hugging Face Hub)
+3. Pushing your app via `git push`
+
+---
+
+## 📸 Screenshots
+
+### 💬 Classifier Prediction + Suggestions
+![example1](screenshots/example1.png)
+
+### 🚨 Safe Response for Crisis Input
+![example2](screenshots/example2.png)
+
+### 🧠 Multiple Suggestions
+![example3](screenshots/example3.png)
+
+### 💾 Logging the Conversation
+![example4](screenshots/example4.png)
+
+---
+
+## 📦 Requirements
+
+Install dependencies:
+
+```bash
+pip install -r requirements.txt
+```
+
+

app.py

@@ -0,0 +1,226 @@
+import streamlit as st
+from datetime import datetime
+import torch
+import json
+import os
+import csv
+import random
+import re
+from detoxify import Detoxify
+from transformers import pipeline, DistilBertTokenizerFast, DistilBertForSequenceClassification
+from sklearn.preprocessing import LabelEncoder
+
+# === Load models ===
+#bert_model_dir = "models/distilbert_response_type_balanced"
+#bert_tokenizer = DistilBertTokenizerFast.from_pretrained(bert_model_dir)
+#bert_model = DistilBertForSequenceClassification.from_pretrained(bert_model_dir)
+
+bert_model_repo = "scdong/distilbert-response-type"
+bert_tokenizer = DistilBertTokenizerFast.from_pretrained(bert_model_repo)
+bert_model = DistilBertForSequenceClassification.from_pretrained(bert_model_repo)
+
+
+bert_model.eval().to("cpu")
+
+generator = pipeline("text2text-generation", model="google/flan-t5-xl", device=0 if torch.cuda.is_available() else -1)
+tox_model = Detoxify("original")
+
+label_names = ["advice", "information", "question", "validation"]
+le = LabelEncoder().fit(label_names)
+
+# === Improved prompt templates ===
+prompt_table = {
+    "advice": "You are a licensed mental health counselor preparing to support a client who said: \"{msg}\". Provide a thoughtful and supportive suggestion that the counselor might offer to the client.",
+    "validation": "You are an empathetic therapist helping a client who shared: \"{msg}\". Write a kind, emotionally validating reflection the counselor might use.",
+    "information": "You are a knowledgeable therapist assisting a client who said: \"{msg}\". Provide accurate, relevant information that could help the counselor respond knowledgeably.",
+    "question": "You are a licensed therapist preparing to support a client who said: \"{msg}\". Generate an insightful, open-ended question to help the client reflect on their experience and move the conversation forward."
+}
+
+response_bank = {
+    "advice": [
+        "You're doing more than enough — even if it doesn't feel like it right now.",
+        "Try to be as kind to yourself as you would to a close friend.",
+        "Even small steps count. You're showing up, and that matters."
+    ],
+    "validation": [
+        "It makes total sense you'd feel that way given what you're going through.",
+        "Your feelings are valid. Many would feel similarly.",
+        "That sounds really tough — thank you for sharing."
+    ],
+    "information": [
+        "Anxiety can cause a racing heart or shortness of breath — that’s very common.",
+        "Panic attacks are intense but temporary. They usually pass within minutes.",
+        "Mindfulness and deep breathing can help calm your nervous system."
+    ],
+    "question": [
+        "Can you tell me more about what’s been making you feel this way?",
+        "What do you think might be behind these feelings?",
+        "How long have you been feeling this way?"
+    ]
+}
+
+# === Safety detection ===
+CRISIS_PATTERN = re.compile("|".join([
+    r"\bi want to (kill|hurt) myself\b",
+    r"\bi (feel|am|have) suicidal\b",
+    r"\bsuicidal thoughts\b",
+    r"\bsuicidal\b",
+    r"\bi want to die\b",
+    r"\bend my life\b",
+    r"\bno reason to live\b",
+    r"\bgive up on life\b",
+    r"\bcan.?t go on\b",
+    r"\bhopeless\b",
+    r"\bhelpless\b",
+    r"\bi hate myself\b",
+    r"\bkill me\b", r"\bself[- ]?harm\b", r"\bi need help now\b",
+    r"\bi never feel good enough\b", r"\bi feel like giving up\b", r"\bi feel worthless\b",
+    r"\bmy (client|patient) .* (is|has|feels|wants to).* (suicidal|end their life|kill (himself|herself))\b"
+]), flags=re.IGNORECASE)
+
+
+
+
+VIOLENCE_PATTERN = re.compile("|".join([
+    r"\bi want to (hurt|kill) (someone|others|people|them|him|her)\b",
+    r"\bi will kill\b", r"\bi feel like attacking\b", r"\bhomicide\b",
+    r"\bviolence against\b", r"\bi want to cause harm\b"
+]), flags=re.IGNORECASE)
+
+CRISIS_RESPONSE = "I'm really sorry you're feeling this way. You're not alone. Please consider calling 988 or speaking with a mental health professional right away."
+VIOLENCE_RESPONSE = "If you're having thoughts of harming others, it's critical to speak with a professional immediately. Please reach out to emergency services or a crisis line."
+
+# === Cache ===
+CACHE_FILE = "cache.json"
+if os.path.exists(CACHE_FILE):
+    with open(CACHE_FILE, "r") as f:
+        response_cache = json.load(f)
+else:
+    response_cache = {}
+
+def save_cache():
+    with open(CACHE_FILE, "w") as f:
+        json.dump(response_cache, f, indent=2)
+
+def is_crisis(text): return bool(CRISIS_PATTERN.search(text))
+def is_violent_threat(text): return bool(VIOLENCE_PATTERN.search(text))
+def is_toxic(text): return tox_model.predict(text)["toxicity"] > 0.7
+
+def predict_response_type(msg):
+    inputs = bert_tokenizer(msg, return_tensors="pt", truncation=True, padding=True, max_length=128)
+    with torch.no_grad():
+        logits = bert_model(**inputs).logits
+    return le.inverse_transform([torch.argmax(logits).item()])[0]
+
+def clean_response(text):
+    text = text.strip()
+    text = re.sub(r'^["“”\'-]*', '', text)
+    text = re.sub(r'\s+', ' ', text)
+    if text and not text[0].isupper(): text = text[0].upper() + text[1:]
+    if text and text[-1] not in ".!?": text += "."
+    return text
+
+def is_high_quality(text):
+    return len(text.split()) >= 5 and "kill" not in text.lower() and "hurt" not in text.lower()
+
+# === Streamlit App ===
+st.set_page_config(page_title="Mental Health Assistant", layout="centered")
+st.title("🧠 Mental Health Counselor Assistant")
+
+# 📘 Instructions
+st.markdown("### 📘 How to use this app")
+st.markdown("""
+1. Enter a message from a patient in the text box.
+2. The app will generate helpful suggestions for how the counselor might reply.
+3. If a safety risk is detected (e.g., crisis or violence), a safe response will be shown instead.
+4. You can save the conversation history as a CSV file.
+""")
+
+# 💬 Patient input
+msg = st.chat_input("💬 Enter a message from the patient")
+
+if "chat_log" not in st.session_state:
+    st.session_state.chat_log = []
+
+# === Response Generation ===
+if msg:
+    timestamp = datetime.now().isoformat()
+
+    if is_violent_threat(msg):
+        safe_responses = [
+            "It’s important to take violent thoughts seriously. If you or someone is in danger, call 911 immediately.",
+            "Please seek professional help right away. Safety is the top priority.",
+            "If your client has expressed intent to harm others, it's urgent to contact crisis services immediately."
+        ]
+        st.error("🚨 Violence detected")
+        st.markdown(f"**🧍 Patient:** {msg}")
+        for r in safe_responses:
+            st.markdown(f"**🧠 Suggestion:** {r}")
+        st.session_state.chat_log.append((timestamp, msg, "VIOLENCE", safe_responses[0]))
+
+    elif is_crisis(msg):
+        safe_responses = [
+            "It sounds like you're in a lot of pain right now. Please know that you're not alone.",
+            "Your life matters. I encourage you to speak with a mental health professional or call 988.",
+            "Thank you for sharing this. There is help available, and you deserve support and care."
+        ]
+        st.error("🆘 Crisis detected")
+        st.markdown(f"**🧍 Patient:** {msg}")
+        for r in safe_responses:
+            st.markdown(f"**🧠 Suggestion:** {r}")
+        st.session_state.chat_log.append((timestamp, msg, "CRISIS", safe_responses[0]))
+
+    elif is_toxic(msg):
+        safe_responses = [
+            "Thank you for expressing yourself. Let's approach this with compassion and care.",
+            "It’s okay to be upset. I'm here to help you explore these feelings in a safe way.",
+            "Let’s take a deep breath together and talk through what's on your mind."
+        ]
+        st.warning("⚠️ Toxic message detected")
+        st.markdown(f"**🧍 Patient:** {msg}")
+        for r in safe_responses:
+            st.markdown(f"**🧠 Suggestion:** {r}")
+        st.session_state.chat_log.append((timestamp, msg, "TOXIC", safe_responses[0]))
+
+    else:
+        response_type = predict_response_type(msg)
+        prompt = prompt_table[response_type].format(msg=msg)
+
+        if prompt in response_cache:
+            responses = response_cache[prompt]
+        else:
+            outputs = generator(
+                prompt,
+                max_new_tokens=150,
+                do_sample=True,
+                top_p=0.9,
+                temperature=0.7,
+                num_return_sequences=3
+            )
+            responses = [
+                clean_response(o["generated_text"])
+                for o in outputs
+                if is_high_quality(clean_response(o["generated_text"]))
+            ]
+            if not responses:
+                responses = random.sample(response_bank[response_type], 3)
+            response_cache[prompt] = responses
+            save_cache()
+
+        st.markdown(f"**🧍 Patient:** {msg}")
+        for r in responses:
+            st.markdown(f"**🧠 Suggestion:** {r}")
+        st.session_state.chat_log.append((timestamp, msg, response_type, responses[0]))
+
+# === Save History ===
+if st.session_state.chat_log:
+    if st.button("💾 Save Conversation History"):
+        os.makedirs("logs", exist_ok=True)
+        filename = f"logs/counselor_log_{datetime.now().strftime('%Y%m%d_%H%M%S')}.csv"
+        with open(filename, "w", newline='', encoding="utf-8") as f:
+            writer = csv.writer(f)
+            writer.writerow(["timestamp", "message", "type", "response"])
+            for row in st.session_state.chat_log:
+                writer.writerow(row)
+        st.success(f"Saved to `{filename}`")
+

cache.json

@@ -0,0 +1,77 @@
+{
+  "You are a compassionate counselor. Ask an open-ended question: \"i feel super sad \"": [
+    "What would you advise the person to do?",
+    "What would you suggest I do next?"
+  ],
+  "You are an empathetic counselor. Validate their emotions: \"i feel depressed \"": [
+    "The feelings of sadness are valid.",
+    "You can talk with them about their feelings and express empathy.",
+    "They might feel like they have lost their motivation, or feel they are not as important as others."
+  ],
+  "You are an empathetic counselor. Validate their emotions: \"i feel very sad \"": [
+    "I understand, that's very sad.",
+    "I can understand, I was just sad for someone in my family."
+  ],
+  "You are a compassionate counselor. Ask an open-ended question: \"i never feel good enough\"": [
+    "What would you do to the narrator?",
+    "What would be the best response to this client?",
+    "How do you help someone feel good enough about themselves?"
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel depresssed \". Write a kind, emotionally validating reflection the counselor might use.": [
+    "You're not alone in feeling this way.",
+    "I understand that you feel depressed, and I am here to help.",
+    "I know how you feel. I have experienced depression myself."
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel sad \". Write a kind, emotionally validating reflection the counselor might use.": [
+    "You are right that you feel sad.",
+    "I can see how sad you feel.",
+    "I can see that you are feeling sad."
+  ],
+  "You are a licensed therapist preparing to support a client who said: \"i have sucidal thoughts\n\". Generate an insightful, open-ended question to help the client reflect on their experience and move the conversation forward.": [
+    "Why do you think the client is having suicidal thoughts?",
+    "What would you like to discuss with this person?",
+    "What can you do for the client to help them feel better?"
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel bad\". Write a kind, emotionally validating reflection the counselor might use.": [
+    "You are not alone. I have been there. I know how it feels.",
+    "You are right, you do feel bad.",
+    "I understand how you feel."
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel very sad \n\". Write a kind, emotionally validating reflection the counselor might use.": [
+    "You are not alone. I have felt very sad many times.",
+    "You are right, you do feel very sad.",
+    "I can tell that you are feeling sad."
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel very sad \". Write a kind, emotionally validating reflection the counselor might use.": [
+    "I know how you feel about this. I have been in this situation before.",
+    "You are very sad. You have been through a lot.",
+    "I can see that you are feeling sad right now."
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel very depressed\". Write a kind, emotionally validating reflection the counselor might use.": [
+    "You are not alone. Many people are depressed. It is just a feeling.",
+    "You do have a very strong sense of sadness."
+  ],
+  "You are a licensed therapist preparing to support a client who said: \"i want to kll myself \". Generate an insightful, open-ended question to help the client reflect on their experience and move the conversation forward.": [
+    "What are you feeling in the moment?",
+    "How might this be a sign that you are in a rut?"
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel sad\". Write a kind, emotionally validating reflection the counselor might use.": [
+    ": I can see that you feel sad.",
+    "I am sorry to hear that. Is there anything I can do to help you?",
+    "I know you feel sad. I've been there myself."
+  ],
+  "You are an empathetic therapist helping a client who shared: \"i feel depressed\". Write a kind, emotionally validating reflection the counselor might use.": [
+    "You're not alone. I've felt that way myself.",
+    "I am sorry you are feeling depressed.",
+    "You are not alone. You are not alone. Depression can be a very debilitating emotion."
+  ],
+  "You are a licensed therapist preparing to support a client who said: \"i have sucidal thoughts\". Generate an insightful, open-ended question to help the client reflect on their experience and move the conversation forward.": [
+    "What is the best way to help the client?",
+    "How did you feel about this?"
+  ],
+  "You are a licensed therapist preparing to support a client who said: \"sucidal thoughts\". Generate an insightful, open-ended question to help the client reflect on their experience and move the conversation forward.": [
+    "What can I do to help you feel less suicidal?",
+    "What is a possible response to your client's suicidal thoughts?",
+    "What might be different if you were not having suicidal thoughts?"
+  ]
+}

requirements.txt

@@ -0,0 +1,5 @@
+streamlit
+torch
+transformers
+scikit-learn
+detoxify