do it

kb.py

@@ -1,16 +1,59 @@
 
 from langchain_pinecone import PineconeVectorStore
-from langchain_core.vectorstores import VectorStore
-from langchain_core.vectorstores import VectorStoreRetriever
 from langchain_aws import BedrockEmbeddings
+from langchain.chains.llm import LLMChain
+from langchain.prompts import PromptTemplate
+from langchain_aws import ChatBedrock
 
 INDEX_NAME = "zeteo-health"
 NAMESPACE = "knowedgebase"
 EMBEDDING_DIMENSION = 1536
 EMBEDDING_MODEL_NAME = "amazon.titan-embed-text-v1"
 
+PROMPT = """
+Your name is Zeteo Assistant.
+You are answering questions about health.
+Only use simple words and descriptions. 
+Respond with an empathetic tone.
+Be informative but concise. 
+You will only respond in English.
+Address your response in second person. 
+Use the first name if it is given to address the response.
+Be friendly but formal, avoid phrases like "Hey there".
+If the question is chit chat such as "Hello", "How are you?", "Good morning", "thank you" or "Howdy" respond with a simple greeting and do not include any other information, especially related to health.
+Limit your response only to information contained in the <context>. When referring to the context, use the term "information".
+If the question is not about healthcare and can't be found in the <context> respond with, "I am sorry, I don't have information to answer that question."
+
+<context>
+{context}
+</context>
+
+<profile>
+{profile}
+</profile>
+
+<question>
+{question}
+</question>
+"""
+
+prompt = PromptTemplate.from_template(PROMPT)
+
 def getkb():
     vectorstore = PineconeVectorStore(index_name=INDEX_NAME, 
                             embedding=BedrockEmbeddings(model_id=EMBEDDING_MODEL_NAME), 
                             namespace=NAMESPACE)
     return vectorstore
+
+
+def get_response(question, context, name):
+    llm = ChatBedrock(model_id='anthropic.claude-3-sonnet-20240229-v1:0')
+    chat_chain = LLMChain(llm=llm, prompt=prompt)
+    result = chat_chain.invoke(
+            {
+                "context": context, 
+                "question": question,
+                "profile": f"name: {name}"
+            }
+        )['text']
+    return result

main.py

@@ -2,7 +2,7 @@ from fastapi import FastAPI, Request, Query
 import logging
 from whatsapp_client import WhatsAppWrapper
 import json
-from kb import getkb
+from kb import getkb, get_response
 
 logger = logging.getLogger()
 
@@ -41,11 +41,13 @@ async def callback(request: Request):
         message = value['messages'][0]
         phone_num = message['from']
         mesg_body = message['text']['body']
+        name = value['contacts'][0]['profile']['name']
         print("SEARCH: " + str(mesg_body))
-        documents = '\n'.join([doc[0].page_content for doc in vectorstore.similarity_search_with_score(mesg_body, k=6)])
-        print(f"CONTENT: {documents}")
+        context = '\n'.join([doc[0].page_content for doc in vectorstore.similarity_search_with_score(mesg_body, k=6)])
+        response = get_response(mesg_body, context, name)
+
         client = WhatsAppWrapper()
-        client.send_text_message("hello, right back at ya ..." + mesg_body,  phone_num)
+        client.send_text_message(response,  phone_num)
 
     return {"status": "success"}
     

test.py

@@ -1,6 +1,7 @@
 import json
-from kb import getkb
+from kb import getkb, get_response
 from dotenv import load_dotenv
+from whatsapp_client import WhatsAppWrapper
 
 load_dotenv()
 
@@ -31,7 +32,7 @@ data = """{
                 "id": "wamid.HBgLMTUxMjc0MDQ2MjAVAgASGBQzQTEyQTM0OURGQUI4QzZDNTg2MAA=",
                 "timestamp": "1716933996",
                 "text": {
-                  "body": "hello"
+                  "body": "how do i get tested for prostate cancer?"
                 },
                 "type": "text"
               }
@@ -49,18 +50,17 @@ value = data['entry'][0]['changes'][0]['value']
 
 def main():
     vectorstore = getkb()
-    print('here i am!')
     if 'messages' in value and 'from' in value['messages'][0] and 'text' in value['messages'][0]:
-        print('i got here!')
         message = value['messages'][0]
+        name = value['contacts'][0]['profile']['name']
         phone_num = message['from']
         mesg_body = message['text']['body']
-        print(f"SEARCH: {mesg_body}, PHONE: {phone_num}")
         
-        documents = '\n'.join([doc[0].page_content for doc in vectorstore.similarity_search_with_score(mesg_body, k=6)])
-        print(f"CONTENT: {documents}")
-        #client = WhatsAppWrapper()
-        #client.send_text_message("hello, right back at ya ..." + mesg_body,  phone_num)
+        context = '\n'.join([doc[0].page_content for doc in vectorstore.similarity_search_with_score(mesg_body, k=6)])
+        response = get_response(mesg_body, context, name)
+        print(f"RESPONSE: {response}")
+        client = WhatsAppWrapper()
+        client.send_text_message("hello, right back at ya ..." + response,  phone_num)
 
 if __name__ == "__main__":
     main()