app.py

# 🧠💬 Advanced Mental Health Companion AI Chatbot with Gradio UI
# Developed using Gemini 1.5 Pro API with RAG, JSON mode, embeddings, and vector search

# --- 📌 Install Required Libraries ---
import os
import json
import gradio as gr
import google.generativeai as genai
from langchain.vectorstores import FAISS
from langchain.embeddings import OpenAIEmbeddings
from langchain.docstore.document import Document
from langchain.text_splitter import CharacterTextSplitter

# --- 📌 Load Gemini API Key ---
gemini_key = os.getenv("GEMINI_API_KEY")
genai.configure(api_key=gemini_key)

# --- 📌 Initialize Gemini Model ---
model = genai.GenerativeModel(model_name="gemini-1.5-pro")
chat = model.start_chat(history=[])

# --- 📌 Define Core Functionalities ---
def detect_emotion(user_input):
    prompt = f"""
    Analyze the message and return a JSON with:
    {{
      "emotion": "<Emotion>",
      "tone": "<Suggested_Tone>",
      "affirmation": "<One_Line_Affirmation>",
      "response": "<Comforting response>"
    }}

    Message: "{user_input}"
    """
    response = model.generate_content(prompt, generation_config={"response_mime_type": "application/json"})
    return json.loads(response.text)

def get_affirmation():
    prompt = "Provide a unique calming affirmation for someone feeling overwhelmed."
    return model.generate_content(prompt).text

def get_journaling_prompt():
    prompt = "Give me a mental health journaling prompt."
    return model.generate_content(prompt).text

def get_calming_technique():
    prompt = "Suggest a calming breathing or grounding technique."
    return model.generate_content(prompt).text

# --- 📌 Memory & Vector Store (RAG Setup) ---
documents = [
    Document(page_content="Take a deep breath. You are doing your best."),
    Document(page_content="It's okay to not be okay. Give yourself grace."),
    Document(page_content="You are not alone. Many have walked this path and found light again."),
    Document(page_content="Small steps lead to big changes. Just begin.")
]

splitter = CharacterTextSplitter(chunk_size=100, chunk_overlap=0)
split_docs = splitter.split_documents(documents)
embedding = OpenAIEmbeddings(openai_api_key=gemini_key)
vector_db = FAISS.from_documents(split_docs, embedding)

def retrieve_supportive_text(user_input):
    docs = vector_db.similarity_search(user_input, k=2)
    return "\n".join([d.page_content for d in docs])

# --- 📌 Smart Response Generator ---
def generate_companion_response(name, issue):
    emotion_data = detect_emotion(issue)
    rag_support = retrieve_supportive_text(issue)

    system_prompt = f"""
    You're MindMate, a kind mental health assistant.
    Respond personally to {name}.

    Emotion: {emotion_data['emotion']}
    Tone: {emotion_data['tone']}
    Affirmation: {emotion_data['affirmation']}

    Based on the above, and the following supportive texts:
    {rag_support}

    Compose a comforting and warm message to the user:
    Message: {issue}
    """
    response = chat.send_message(system_prompt)
    return response.text, emotion_data['emotion'], emotion_data['affirmation']

# --- 📌 Gradio UI ---
def chatbot_interface(name, issue):
    response, emotion, affirmation = generate_companion_response(name, issue)
    journaling = get_journaling_prompt()
    technique = get_calming_technique()

    return {
        "Emotion": emotion,
        "Affirmation": affirmation,
        "Companion Response": response,
        "Journaling Prompt": journaling,
        "Calming Tip": technique
    }

inputs = [
    gr.Textbox(label="Your Name"),
    gr.Textbox(label="What's troubling you today?", lines=4)
]

outputs = gr.JSON(label="MindMate's Support")

demo = gr.Interface(
    fn=chatbot_interface,
    inputs=inputs,
    outputs=outputs,
    title="🧠 MindMate: Your Mental Health Companion",
    description="Talk to MindMate by sharing your name and what's on your mind. Get emotional insights, affirmations, and comfort."
)

demo.launch()