mdb-11.py

from elevenlabs.client import ElevenLabs
from elevenlabs import play
from langchain_community.vectorstores import MongoDBAtlasVectorSearch
from langchain_openai import AzureOpenAIEmbeddings
from langchain_core.messages import HumanMessage
from langchain_openai import AzureChatOpenAI
import pymongo
# PDF loaded into MongoDB Atlas = https://arxiv.org/pdf/2303.08774.pdf
MDB_URI = ""
cluster = pymongo.MongoClient(MDB_URI)
DB_NAME = "extbrain"
COLLECTION_NAME = "demo_collection"
azureEmbeddings = AzureOpenAIEmbeddings(
    deployment="________________",
    model="text-embedding-ada-002",
    azure_endpoint="https://_____________.openai.azure.com",
    openai_api_key="__API_KEY__",
    openai_api_type="azure",
    chunk_size=1,
    disallowed_special=()
)
vector_search = MongoDBAtlasVectorSearch.from_connection_string(
    MDB_URI,
    DB_NAME + "." + COLLECTION_NAME,
    azureEmbeddings,
    index_name="vector_index",
)
model = AzureChatOpenAI(
    deployment_name="________________", # Or another suitable engine  
    azure_endpoint="https://_____________.openai.azure.com",
    openai_api_key="__API_KEY__",
    openai_api_type="azure",
    api_version="2023-03-15-preview",
)
query = "How will the development of more efficient hardware and algorithms impact the future compute requirements for training large language models like GPT-4?"
print("QUESTION:"+query)
results = vector_search.similarity_search(query)
print("# of Results:"+str(len(results)))
retrieved_documents = [result.page_content for result in results[:3]]  
message = HumanMessage(
    content=f"Answer the following question based on these documents:\n**Question:** {query}\n**Documents:** {retrieved_documents} \n**Answer:**"
)
client = ElevenLabs(
  api_key="__API_KEY__" 
)
audio = client.generate(
    text=str(model([message]).content),
    voice="__VOICE_ID__"
)
play(audio)