Upload folder using huggingface_hub

conv_app.py

@@ -1,9 +1,41 @@
 import gradio as gr
-from conversation import make_conversation, auth_function
+from conversation import make_conversation
 import random
+import time
+
+global USERNAME
+global PASSWORD
+global INPUT
+global OUTPUT
+global SOURCE
+global DOCS
+
+# def auth_function(username, password):
+#     USERNAME = username
+#     user_name = username
+#     return username == password
+
+
+# def make_conversation(message, history):
+#     INPUT = message
+#     text_, source, docs = run(message)
+#     OUTPUT = text_
+#     SOURCE = source
+#     DOCS = docs
+
+#     # print("INPUT: ", INPUT)
+#     # print("OUTPUT: ", OUTPUT)
+#     # print("SOURCE: ", SOURCE)
+#     # print("DOCS: ", DOCS)
+
+#     for i in range(len(text_)):
+#         time.sleep(0.001)
+#         yield text_[: i+1]
 
 with gr.Blocks(css="style.css") as demo:
-    gr.Markdown("##DR. VAI")
+    gr.Markdown("""
+    # Dr. V AI
+    """)
     gr.ChatInterface(make_conversation).queue()
 
 demo.launch()

conversation.py

@@ -1,3 +1,123 @@
+# from langchain.document_loaders import TextLoader
+# import pinecone
+# from langchain.vectorstores import Pinecone
+# import os
+# from transformers import AutoTokenizer, AutoModel
+# from langchain.agents.agent_toolkits import create_conversational_retrieval_agent
+# from langchain.agents.agent_toolkits import create_retriever_tool
+# from langchain.chat_models import ChatOpenAI
+# import torch
+# from langchain.agents.openai_functions_agent.agent_token_buffer_memory import (AgentTokenBufferMemory)
+# from langchain.agents.openai_functions_agent.base import OpenAIFunctionsAgent
+# from langchain.schema.messages import SystemMessage
+# from langchain.prompts import MessagesPlaceholder
+# import gradio as gr
+# import time
+# from db_func import insert_one
+# from global_variable_module import gobal_input, global_output
+# import random   
+
+
+# def get_bert_embeddings(sentence):
+#     embeddings = []
+#     input_ids = tokenizer.encode(sentence, return_tensors="pt")
+#     with torch.no_grad():
+#         output = model(input_ids)
+#     embedding = output.last_hidden_state[:,0,:].numpy().tolist()
+#     return embedding
+
+# model_name = "BAAI/bge-base-en-v1.5"
+# model = AutoModel.from_pretrained("/Users/aakashbhatnagar/Documents/masters/ophthal_llm/models/models--BAAI--bge-base-en-v1.5/snapshots/617ca489d9e86b49b8167676d8220688b99db36e")
+# tokenizer = AutoTokenizer.from_pretrained("/Users/aakashbhatnagar/Documents/masters/ophthal_llm/models/models--BAAI--bge-base-en-v1.5/snapshots/617ca489d9e86b49b8167676d8220688b99db36e")
+# prompt_file = open("prompts/version_2.txt", "r").read()
+
+# pinecone.init(
+#     api_key=os.getenv("PINECONE_API_KEY"),  # find at app.pinecone.io
+#     environment=os.getenv("PINECONE_ENV"),  # next to api key in console
+# )
+
+# index_name = "ophtal-knowledge-base"
+# index = pinecone.Index(index_name)
+# vectorstore = Pinecone(index, get_bert_embeddings, "text")
+# retriever = vectorstore.as_retriever()
+# tool = create_retriever_tool(
+#     retriever,
+#     "search_ophtal-knowledge-base",
+#     "Searches and returns documents regarding the ophtal-knowledge-base.",
+# )
+# tools = [tool]
+# system_message = SystemMessage(content=prompt_file)
+# memory_key='history'
+
+# llm = ChatOpenAI(openai_api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4", temperature=0.2)
+# prompt = OpenAIFunctionsAgent.create_prompt(
+#     system_message=system_message,
+#     extra_prompt_messages=[MessagesPlaceholder(variable_name=memory_key)],
+# )
+
+# agent_executor = create_conversational_retrieval_agent(llm, tools, verbose=False, prompt=prompt)
+# user_name = None
+
+# def run(input_):
+#     output = agent_executor({"input": input_})
+#     output_text = output["output"]
+#     source_text = ""
+#     doc_text = ""
+#     global_input = input_
+#     global_output = output_text
+        
+
+#     if len(output["intermediate_steps"])>0:
+#         documents = output["intermediate_steps"][0][1]
+#         sources = []
+#         docs = []
+
+#         for doc in documents:
+#             if doc.metadata["source"] not in sources:
+#                 sources.append(doc.metadata["source"])
+#                 docs.append(doc.page_content)
+
+#         for i in range(len(sources)):
+#             temp = sources[i].replace('.pdf',  '').replace('.txt', '').replace("AAO", "").replace("2022-2023", "").replace("data/book", "").replace("text", "").replace("  ", " ")
+#             source_text += f"{i+1}. {temp}\n"
+#             doc_text += f"{i+1}. {docs[i]}\n"
+
+#         # output_text =  f"{output_text} \n\nSources: \n{source_text}\n\nDocuments: \n{doc_text}"
+#         # output_text =  f"{output_text}"
+    
+#     doc_to_insert = {
+#         "user": user_name,
+#         "input": input_,
+#         "output": output_text,
+#         "source": source_text,
+#         "documents": doc_text
+#     }
+
+#     insert_one(doc_to_insert)
+
+#     return output_text
+
+# def make_conversation(message, history):
+#     text_ = run(message)
+#     for i in range(len(text_)):
+#         time.sleep(0.001)
+#         yield text_[: i+1]
+    
+# def auth_function(username, password):
+#     user_name = username
+#     return username == password
+
+# def random_response(message, accuracy, history):
+#     print(type(message))
+#     print(message)
+#     print(accuracy)
+#     out = random.choice(["Yes", "No"])
+#     gobal_input = out
+#     # open a txt file
+#     with open("function hit", "a+") as f:
+#         f.write(message)
+#     return out
+
 from langchain.document_loaders import TextLoader
 import pinecone
 from langchain.vectorstores import Pinecone
@@ -14,6 +134,8 @@ from langchain.prompts import MessagesPlaceholder
 import gradio as gr
 import time
 from db_func import insert_one
+from langchain.agents import AgentExecutor
+
 
 def get_bert_embeddings(sentence):
     embeddings = []
@@ -24,8 +146,8 @@ def get_bert_embeddings(sentence):
     return embedding
 
 model_name = "BAAI/bge-base-en-v1.5"
-model = AutoModel.from_pretrained(model_name)
-tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModel.from_pretrained("models/models--BAAI--bge-base-en-v1.5/snapshots/617ca489d9e86b49b8167676d8220688b99db36e")
+tokenizer = AutoTokenizer.from_pretrained("models/models--BAAI--bge-base-en-v1.5/snapshots/617ca489d9e86b49b8167676d8220688b99db36e")
 prompt_file = open("prompts/version_2.txt", "r").read()
 
 pinecone.init(
@@ -43,21 +165,38 @@ tool = create_retriever_tool(
     "Searches and returns documents regarding the ophtal-knowledge-base.",
 )
 tools = [tool]
-system_message = SystemMessage(content=prompt_file)
+system_message = SystemMessage(content="You are an assistant to ophthamologists and your name is 'Dr.V AI'. Help users answer medical questions. You are supposed to answer only medical questions and not general questions.")
 memory_key='history'
 
 llm = ChatOpenAI(openai_api_key=os.getenv("OPENAI_API_KEY"), model="gpt-4", temperature=0.2)
+# llm = ChatOpenAI(openai_api_key="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", model="gpt-4", temperature=0.2)
 prompt = OpenAIFunctionsAgent.create_prompt(
     system_message=system_message,
     extra_prompt_messages=[MessagesPlaceholder(variable_name=memory_key)],
 )
 
-agent_executor = create_conversational_retrieval_agent(llm, tools, verbose=False, prompt=prompt)
+memory = AgentTokenBufferMemory(memory_key=memory_key, llm=llm)
+
+
+
+# agent_executor = create_conversational_retrieval_agent(llm, tools, verbose=False, prompt=prompt, )
+
+agent = OpenAIFunctionsAgent(llm=llm, tools=tools, prompt=prompt)
+
+agent_executor = AgentExecutor(
+    agent=agent,
+    tools=tools,
+    memory=memory,
+    verbose=True,
+    return_intermediate_steps=True,
+)
+
 user_name = None
 
 def run(input_):
     output = agent_executor({"input": input_})
     output_text = output["output"]
+    print(output_text)
     source_text = ""
     doc_text = ""    
 
@@ -87,19 +226,19 @@ def run(input_):
         "documents": doc_text
     }
 
-    insert_one(doc_to_insert)
+    # insert_one(doc_to_insert)
 
     return output_text
 
-def make_conversation(message, history):
-    text_ = run(message)
-    for i in range(len(text_)):
-        time.sleep(0.001)
-        yield text_[: i+1]
+# def make_conversation(message, history):
+#     text_ = run(message)
+#     for i in range(len(text_)):
+#         time.sleep(0.001)
+#         yield text_[: i+1]
     
-def auth_function(username, password):
-    user_name = username
-    return username == password
+# def auth_function(username, password):
+#     user_name = username
+#     return username == password