services/vectordb_manager/vectordb_manager.py

from langchain.chains import RetrievalQA
from langchain.prompts import PromptTemplate
from langchain.embeddings import OpenAIEmbeddings

# from langchain_community.embeddings import OpenAIEmbeddings
from langchain.vectorstores import Chroma
from src.utils import brian_knows_system_message
from uuid import uuid4

import chromadb
from chromadb.config import Settings
from chromadb.utils import embedding_functions

import sys
import os
import openai

import logging

sys.path.append("../..")
from dotenv import load_dotenv, find_dotenv

_ = load_dotenv(find_dotenv())  # read local .env file
openai.api_key = os.environ["OPENAI_API_KEY"]


class VectordbManager:
    def __init__(
        self,
        knowledge_base_name: str,
    ) -> None:
        self.knowledge_base_name = knowledge_base_name
        self.vector_db = None

    def load_vectordb(
        self,
        embedding_function=OpenAIEmbeddings(),
    ):
        client = chromadb.HttpClient(
            host="chroma.brianknows.org",
            port="443",
            ssl=True,
            settings=Settings(allow_reset=True),
        )
        vectordb = Chroma(embedding_function=embedding_function, client=client)
        self.vector_db = vectordb

    def load_collection(self, embedding_function=OpenAIEmbeddings()):

        client = chromadb.HttpClient(
            host="chroma.brianknows.org",
            port=443,
            ssl=True,
            settings=Settings(
                allow_reset=True,
            ),
        )

        collection = client.get_collection(
            self.knowledge_base_name,
            embedding_function=embedding_functions.OpenAIEmbeddingFunction(
                api_key=os.environ["OPENAI_API_KEY"]
            ),
        )
        return collection

    def create_vector_db(self, splits: list, knowledge_base_name: str):
        logging.info("create_vector_db")
        embedding_fn = OpenAIEmbeddings()

        try:
            client = chromadb.HttpClient(
                host="chroma.brianknows.org",
                port=443,
                ssl=True,
                settings=Settings(
                    allow_reset=True,
                ),
            )
            collection = client.get_or_create_collection(
                knowledge_base_name,
                embedding_function=embedding_functions.OpenAIEmbeddingFunction(
                    api_key=os.environ["OPENAI_API_KEY"]
                ),
            )

            ids = []
            metadatas = []
            documents = []

            for split in splits:
                ids.append(str(uuid4()))
                metadatas.append(split.metadata)
                documents.append(split.page_content)
            collection.add(documents=documents, ids=ids, metadatas=metadatas)
            vector_db = Chroma.from_documents(
                documents=splits, embedding=embedding_fn, client=client
            )
            self.vector_db = vector_db

        except Exception as e:
            logging.error(f"error in creating db: {str(e)}")

    def add_splits_to_existing_vectordb(
        self,
        splits: list,
    ):
        for split in splits:
            try:
                self.vector_db.add_documents([split])
                print("document loaded!")
            except Exception as e:
                print(f"Error with doc : {split}")
                print(e)

    def retrieve_docs_from_query(self, query: str, k=2, fetch_k=3) -> list:
        """
        query : Text to look up documents similar to.
        k : Number of Documents to return. Defaults to 4.
        fetch_k : Number of Documents to fetch to pass to MMR algorithm.
        lambda_mult : Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5.
        """
        retrieved_docs = self.vector_db.max_marginal_relevance_search(
            query, k=k, fetch_k=fetch_k
        )
        return retrieved_docs

    def retrieve_qa(
        self,
        llm,
        query: str,
        score_threshold: float = 0.65,
        system_message=brian_knows_system_message,
    ):
        """return llm answer based on docs"""

        # Build prompt
        template = """You are a Web3 assistant. Use the following pieces of context to answer the question at \
            the end. If you don't know the answer, just say: "I don't know". Don't try to make up an \
                answer! Provide a always a detailed and comprehensive response. """

        fixed_template = """ {context}
        Question: {question}
        Detailed Answer:"""

        template = system_message + fixed_template

        QA_CHAIN_PROMPT = PromptTemplate.from_template(template)

        # Run chain
        qa_chain = RetrievalQA.from_chain_type(
            llm,
            retriever=self.vector_db.as_retriever(
                search_type="similarity_score_threshold",
                search_kwargs={"score_threshold": score_threshold},
            ),
            return_source_documents=True,
            chain_type_kwargs={"prompt": QA_CHAIN_PROMPT},
            # reduce_k_below_max_tokens=True,
        )
        result = qa_chain({"query": query})

        return result