follow-up

ask.py

@@ -2,18 +2,21 @@ from llama_index import GPTPineconeIndex, LLMPredictor, ServiceContext
 import pinecone
 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
 
 
 # logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
 # logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))
+pinecone_key=os.environ['PINECONE_KEY']
 
-
-def askQuestion(brain, question, temperature, maxTokens):
+def askQuestion(brain, question, prompt, temperature, maxTokens):
     temperature = float(temperature)
-    print(brain, question, temperature,maxTokens)
+    finalQuestion = prompt+question
+    print(brain, finalQuestion, temperature, maxTokens)
     Brain_Name = brain.lower()
     print(Brain_Name)
-    pinecone.init(api_key=os.environ['PINECONE_KEY'],
+    pinecone.init(api_key=pinecone_key,
                   environment="us-west4-gcp")
     pineconeindex = pinecone.Index(Brain_Name)
     index = GPTPineconeIndex([], pinecone_index=pineconeindex)
@@ -24,7 +27,7 @@ def askQuestion(brain, question, temperature, maxTokens):
 
     # 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 = question
+    query = finalQuestion
     # relevant info from brain goes here
     info = ["pdf"]
 
@@ -37,11 +40,14 @@ def askQuestion(brain, question, temperature, maxTokens):
                            similarity_top_k=data_chunks, response_mode="compact")
     print(question)
     print(response)
-    return response
+    memory = ConversationBufferMemory(memory_key="chat_history")
+    memory.chat_memory.add_user_message(question)
+    memory.chat_memory.add_ai_message(response.response)
+    return response, memory
 
 
 def getBrains(name):
-    pinecone.init(api_key=os.environ['PINECONE_KEY'],
+    pinecone.init(api_key=pinecone_key,
                   environment="us-west4-gcp")
     active_indexes = pinecone.list_indexes()
     print(active_indexes)
@@ -50,3 +56,32 @@ def getBrains(name):
         return True
     else:
         return False
+
+
+def runAgent(memory, question, temperature, maxTokens):
+    if (memory == False):
+        return "Please Initiate the Chat first.."
+    temperature = float(temperature)
+    pinecone.init(api_key=pinecone_key,
+                  environment="us-west4-gcp")
+    pineconeindex = pinecone.Index("dimetyd-test")
+    index = GPTPineconeIndex([], pinecone_index=pineconeindex)
+    # memory = ConversationBufferMemory(memory_key="chat_history")
+    # print(memory.chat_memory)
+    llm = OpenAI(
+        temperature=temperature, model_name="text-davinci-003", max_tokens=maxTokens)
+    tool_config = IndexToolConfig(
+        index=index,
+        name="Vector Index",
+        description="Use this tool if you can't find the required Information in the previous message history",
+        index_query_kwargs={"similarity_top_k": 4, "response_mode": "compact"},
+        tool_kwargs={"return_direct": True}
+    )
+
+    toolkit = LlamaToolkit(index_configs=[tool_config])
+
+    agent_chain = create_llama_chat_agent(
+        toolkit, llm, memory=memory, verbose=True)
+    response = agent_chain.run(question)
+    # print(memory.chat_memory)
+    return response, memory