import gradio as gr from huggingface_hub import InferenceClient from typing import List, Tuple import fitz # PyMuPDF from sentence_transformers import SentenceTransformer, util import numpy as np import faiss client = InferenceClient("HuggingFaceH4/zephyr-7b-beta") # Placeholder for the app's state class MyApp: def __init__(self) -> None: self.documents = [] self.embeddings = None self.index = None self.load_pdf("THEDIA1.pdf") self.build_vector_db() def load_pdf(self, file_path: str) -> None: """Extracts text from a PDF file and stores it in the app's documents.""" doc = fitz.open(file_path) self.documents = [] for page_num in range(len(doc)): page = doc[page_num] text = page.get_text() self.documents.append({"page": page_num + 1, "content": text}) print("PDF processed successfully!") def build_vector_db(self) -> None: """Builds a vector database using the content of the PDF.""" model = SentenceTransformer('all-MiniLM-L6-v2') self.embeddings = model.encode([doc["content"] for doc in self.documents]) self.index = faiss.IndexFlatL2(self.embeddings.shape[1]) self.index.add(np.array(self.embeddings)) print("Vector database built successfully!") def search_documents(self, query: str, k: int = 3) -> List[str]: """Searches for relevant documents using vector similarity.""" model = SentenceTransformer('all-MiniLM-L6-v2') query_embedding = model.encode([query]) D, I = self.index.search(np.array(query_embedding), k) results = [self.documents[i]["content"] for i in I[0]] return results if results else ["No relevant documents found."] app = MyApp() def respond( message: str, history: List[Tuple[str, str]], system_message: str, max_tokens: int, temperature: float, top_p: float, ): system_message = ( "You are a knowledgeable DBT (Dialectical Behavior Therapy) coach. You greet users warmly and ask questions like a real counselor. " "You are concise, respectful, and a good listener. You use the DBT book to guide users through DBT exercises and provide helpful information. " "When needed, you ask one follow-up question at a time to guide the user to ask appropriate questions. " "You avoid giving suggestions if any dangerous act is mentioned by the user and refer them to call someone or emergency services. " "Your responses are accurate and concise, and you maintain a professional and supportive tone throughout." ) messages = [{"role": "system", "content": system_message}] for val in history: if val[0]: messages.append({"role": "user", "content": val[0]}) if val[1]: messages.append({"role": "assistant", "content": val[1]}) messages.append({"role": "user", "content": message}) # RAG - Retrieve relevant documents try: retrieved_docs = app.search_documents(message) context = "\n".join(retrieved_docs) messages.append({"role": "system", "content": "Relevant documents: " + context}) response = "" response_buffer = [] for message in client.chat_completion( messages, max_tokens=max_tokens, stream=True, temperature=temperature, top_p=top_p, ): token = message.choices[0].delta.content response += token response_buffer.append(token) if token.endswith('.') or token.endswith('?'): yield ''.join(response_buffer) response_buffer = [] if response_buffer: yield ''.join(response_buffer) except Exception as e: yield f"An error occurred: {str(e)}" demo = gr.Blocks() with demo: gr.Markdown("🧘‍♀️ **Dialectical Behaviour Therapy**") gr.Markdown( "‼️Disclaimer: This chatbot is based on a DBT exercise book that is publicly available. " "We are not medical practitioners, and the use of this chatbot is at your own responsibility.‼️" ) chatbot = gr.ChatInterface( respond, examples=[ ["I feel overwhelmed with work."], ["Can you guide me through a quick meditation?"], ["How do I stop worrying about things I can't control?"], ["What are some DBT skills for managing anxiety?"], ["Can you explain mindfulness in DBT?"], ["I am interested in DBT exercises"], ["I feel restless. Please help me."], ["I have destructive thoughts coming to my mind repetitively."] ], title='Dialectical Behaviour Therapy Assistant👩‍⚕️' ) if __name__ == "__main__": demo.launch()