dn

ask.py

@@ -4,6 +4,7 @@ from langchain import OpenAI
 import os
 from llama_index.langchain_helpers.agents import IndexToolConfig, LlamaIndexTool, LlamaToolkit, create_llama_chat_agent
 from langchain.chains.conversation.memory import ConversationBufferMemory
+from llama_index import QuestionAnswerPrompt
 
 
 # logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
@@ -13,31 +14,44 @@ pinecone_key=os.environ['PINECONE_KEY']
 def askQuestion(brain, question, prompt, temperature, maxTokens):
     temperature = float(temperature)
     finalQuestion = prompt+question
-    print(brain, finalQuestion, temperature, maxTokens)
+    print(finalQuestion)
+    print(temperature, maxTokens)
+    #print(type(temperature))
+    #print(type(maxTokens))
     Brain_Name = brain.lower()
     print(Brain_Name)
     pinecone.init(api_key=pinecone_key,
                   environment="us-west4-gcp")
     pineconeindex = pinecone.Index(Brain_Name)
+    pineconeindex.describe_index_stats
     index = GPTPineconeIndex([], pinecone_index=pineconeindex)
     # index = GPTSimpleVectorIndex.load_from_disk('index.json')
 
     # For Q-A set this value to 4, For Content-Genration set this value b/w 7-10.
     data_chunks = 5
 
-    # prompt query goes here
-    # query="summarize in full detail the solution that dimetyd is providing, and previous email sequences which can be used as a context knowledge"
-    query = finalQuestion
+    QA_PROMPT_TMPL = (
+    "We have provided context information below. \n"
+    "---------------------\n"
+    "{context_str}"
+    "\n---------------------\n"
+    "Given this information, please answer the question at the end of this main prompt: "+prompt+" {query_str}\n"
+    )
+
+    QA_PROMPT = QuestionAnswerPrompt(QA_PROMPT_TMPL)
+    
+    query = question
     # relevant info from brain goes here
     info = ["pdf"]
 
     llm_predictor = LLMPredictor(llm=OpenAI(
         temperature=temperature, model_name="text-davinci-003", max_tokens=maxTokens))
+
     service_context_gpt4 = ServiceContext.from_defaults(
         llm_predictor=llm_predictor)
 
     response = index.query(query, service_context=service_context_gpt4,
-                           similarity_top_k=data_chunks, response_mode="compact")
+                           similarity_top_k=data_chunks, response_mode="compact",text_qa_template=QA_PROMPT)
     print(question)
     print(response)
     memory = ConversationBufferMemory(memory_key="chat_history")