# -*- coding: utf-8 -*- """utils.ipynb Automatically generated by Colab. Original file is located at https://colab.research.google.com/drive/1mBC7DV-7zgoOK7ie6CXwhRvQyqsgJjiy """ # -*- coding: utf-8 -*- """utils.ipynb Automatically generated by Colab. Original file is located at https://colab.research.google.com/drive/13uVA8g6FZq1KLvDxbVtgbp1trtKd_0W8 """ import os from langchain.chat_models import ChatOpenAI from langchain.document_loaders import TextLoader from langchain.embeddings import OpenAIEmbeddings from langchain.vectorstores import FAISS from langchain.chains.summarize import load_summarize_chain from langchain.tools import Tool from langchain.agents import initialize_agent, AgentType from langchain.document_loaders import TextLoader from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter import geopy import openai from geopy.geocoders import Nominatim import requests import googlemaps from langchain.schema import SystemMessage # Using Google maps API #gmaps = googlemaps.Client(key=os.getenv("GOOGLE_MAPS_API_KEY")) gmaps = googlemaps.Client(key="AIzaSyDw18m127Esf_Phj-oaIcUN_2C3Ax2lXs4") # Load API key from environment variable (Hugging Face Secrets) #openai_api_key = os.getenv("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") openai_api_key = "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" # your current key # os.environ["OPENAI_API_KEY"] = openai_api_key # Load and process docs loader = TextLoader("./combined_text.txt") documents = loader.load() # Split text into chunks text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100) docs = text_splitter.split_documents(documents) embedding = OpenAIEmbeddings(openai_api_key=openai_api_key) vectorstore = FAISS.from_documents(docs, embedding) retriever = vectorstore.as_retriever() system_message = SystemMessage( content=""" You are a helpful mental health assistant. Always use the available tools to answer questions. If the user asks about clinics or hospitals, call the clinic search tool and return its result directly (don't summarize). Return full markdown responses (bold names, bullet points, maps, links). """ ) # llm = ChatOpenAI(model_name="gpt-4", temperature=0) # summary_chain = load_summarize_chain(llm, chain_type="map_reduce") llm = ChatOpenAI(model_name="gpt-3.5-turbo", openai_api_key=openai_api_key, temperature=0) #llm = llm.bind_messages([system_message]) summary_chain = load_summarize_chain(llm, chain_type="map_reduce") def suggest_helpline(query): return """📞 Helplines: - Crisis Text Line: Text HOME to 741741 - National Suicide Prevention Lifeline: 1-800-273-TALK - Trevor Project (LGBTQ+): 1-866-488-7386""" helpline_tool = Tool( name="EmergencyHelpline", func=suggest_helpline, description="Offer emergency mental health helpline numbers based on distressing keywords." ) def summarize_retrieved(query): docs = retriever.get_relevant_documents(query) return summary_chain.run(docs) summarized_tool = Tool( name="SummarizedDocSearch", func=summarize_retrieved, description="Use this tool to retrieve and summarize relevant information from symptoms, precausion, suggested food and activities from the knowledge base." ) # def find_clinics(city, radius=10000): # location = gmaps.geocode(city)[0]['geometry']['location'] # lat, lng = location['lat'], location['lng'] # places_result = gmaps.places_nearby( # location=(lat, lng), # radius=radius, # keyword="mental health center", # type="health" # ) # results = places_result.get("results", [])[:5] # limit to 5 clinics # details = [] # for place in results: # place_id = place["place_id"] # place_details = gmaps.place(place_id=place_id).get("result", {}) # name = place_details.get("name", "N/A") # address = place_details.get("formatted_address", "N/A") # phone = place_details.get("formatted_phone_number", "N/A") # website = place_details.get("website", "N/A") # rating = place_details.get("rating", "N/A") # hours = place_details.get("opening_hours", {}).get("weekday_text", []) # maps_link = f"https://maps.google.com/?cid={place_details.get('url', '').split('cid=')[-1]}" if "url" in place_details else "N/A" # hours_md = "\n".join(hours) if hours else "No hours listed" # website_md = f"[Website]({website})" if website != "N/A" else "N/A" # details.append(f"""🏥 **{name}** # 📍 {address} # 📞 {phone} # 🌐 {website_md} # ⭐ Rating: {rating} # 🕒 Working Hours:\n{hours_md} # 📌 [View on Google Maps]({maps_link})""") # return "\n\n---\n\n".join(details) or "No clinics found." def find_clinics(city, radius=10000): location = gmaps.geocode(city)[0]['geometry']['location'] lat, lng = location['lat'], location['lng'] places_result = gmaps.places_nearby( location=(lat, lng), radius=radius, keyword="mental health center", type="health" ) results = places_result.get("results", [])[:5] clinics = [] for place in results: place_id = place["place_id"] details = gmaps.place(place_id=place_id).get("result", {}) clinics.append({ "name": details.get("name", "N/A"), "address": details.get("formatted_address", "N/A"), "phone": details.get("formatted_phone_number", "N/A"), "lat": details.get("geometry", {}).get("location", {}).get("lat"), "lng": details.get("geometry", {}).get("location", {}).get("lng"), }) return clinics clinic_finder_tool = Tool( name="NearbyMentalHealthClinics", func=lambda q: find_clinics(q), description=( "Use this tool when a user asks to find mental health clinics, hospitals, or centers " "in a specific city or location. It returns detailed clinic info including: " "name, full address, contact number, website, Google rating, opening hours, map link and suggested helpline numbers" ) ) tools = [summarized_tool, helpline_tool, clinic_finder_tool] agent = initialize_agent( tools=tools, llm=llm, agent=AgentType.STRUCTURED_CHAT_ZERO_SHOT_REACT_DESCRIPTION, verbose=True )