adding new service for asking questions to a pdf with langchain, vect…

…ordb, and autogen

ai_agency_05_pdf/.env

@@ -0,0 +1,4 @@
+OPENAI_API_KEY=sk-1111
+BASE_URL=http://localhost:1234/v1
+AUTOGEN_USE_DOCKER=False
+TOKENIZERS_PARALLELISM=false

ai_agency_05_pdf/main.py

@@ -0,0 +1,93 @@
+import os
+
+import autogen
+import dotenv
+from langchain.chains import ConversationalRetrievalChain
+from langchain.memory import ConversationBufferMemory
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.document_loaders import PyPDFLoader
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_community.vectorstores import Chroma
+from langchain_openai import OpenAI
+from typing_extensions import Annotated
+
+dotenv.load_dotenv()
+
+# load the pdf file from directory
+loaders = [PyPDFLoader('./ancient_rome.pdf')]
+docs = []
+for file in loaders:
+    docs.extend(file.load())
+# split text to chunks
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000)
+docs = text_splitter.split_documents(docs)
+
+# pages = loaders[0].load_and_split()
+# print(pages[0])
+
+# create a vectorstore
+vectorstore = Chroma(
+    collection_name="full_documents",
+    embedding_function=HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2",
+                                             model_kwargs={'device': 'cpu'})
+)
+vectorstore.add_documents(docs)
+
+qa = ConversationalRetrievalChain.from_llm(
+    OpenAI(temperature=0),
+    vectorstore.as_retriever(),
+    memory=ConversationBufferMemory(memory_key="chat_history", return_messages=True)
+)
+
+# set config for autogen
+config_list = [
+    {
+        # "base_url": os.getenv("BASE_URL"),
+        "api_key": os.getenv("OPENAI_API_KEY"),
+        "model": "gpt-3.5-turbo"
+    }
+]
+
+# set autogen user agent and assistant agent with function calling
+llm_config = {
+    "timeout": 600,
+    "seed": 42,
+    "config_list": config_list,
+    "temperature": 0.7,
+}
+
+# create an AssistantAgent instance "assistant"
+assistant = autogen.AssistantAgent(
+    name="assistant",
+    llm_config=llm_config,
+)
+
+# create a UserProxyAgent instance "user_proxy"
+user_proxy = autogen.UserProxyAgent(
+    name="user_proxy",
+    human_input_mode="NEVER",
+    max_consecutive_auto_reply=1,
+    code_execution_config=False,
+    llm_config=llm_config,
+    system_message="""Reply TERMINATE if the task has been solved at full satisfaction.
+Otherwise, reply CONTINUE, or the reason why the task is not solved yet."""
+)
+
+
+@user_proxy.register_for_execution()
+@assistant.register_for_llm(description="PDF Assistant.")
+def chat_docs(question: Annotated[str, "The question to be asked"]) -> str:
+    response = qa({"question": question})
+    return response["answer"]
+
+
+user_proxy.initiate_chat(
+    assistant,
+    message="""
+Find the answers to the question below from the ancient_rome.pdf and do not write any code.
+
+1. Can you give me a multiple choice question from the pdf with 4 choices and 1 of them being the answer?  Also give the reason for the answer and the page it was found on
+
+Start the work now.
+"""
+)