Discussion about RAG model using Chroma DB

Hi everyone,
I am using Langchain RetrievalQA chain to QA over a JSON document. The document is related to the organization’s portfolio. I tried it for one-on-one module, the chatbot results are good for that but when I try it on a complete portfolio it does not return correct answer.

import os
from langchain.vectorstores import Chroma
from langchain.embeddings import OpenAIEmbeddings
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.llms import OpenAI
from langchain.chains import RetrievalQA
from langchain.document_loaders import JSONLoader

os.environ['OPENAI_API_KEY'] = 'sk-my-openai-api-key'
document_path = '/path/to/document/portfolio.json'

def load_docs(directory):
  loader = JSONLoader(document_path, jq_schema=".", text_content=False)
  documents = loader.load()
  return documents

documents = load_docs(document_path)

def split_docs(documents, chunk_size=500, chunk_overlap=10):
  text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=chunk_size,
    chunk_overlap=chunk_overlap,
    length_function=len,
    add_start_index=True
  )
  docs = text_splitter.split_documents(documents)
  return docs

docs = split_docs(documents)

persist_directory = 'chroma_db'
embedding = OpenAIEmbeddings()

vectordb = Chroma.from_documents(
    documents=docs,
    embedding=embedding,
    persist_directory=persist_directory,
    collection_name='semp6_portfolio'
)
vectordb.persist()
vectordb = None

prsstdb = Chroma(
    persist_directory=persist_directory,
    embedding_function=embedding,
    collection_name='semp6_portfolio'
)
qa = RetrievalQA.from_chain_type(
    llm=OpenAI(),
    chain_type="stuff",
    retriever=prsstdb.as_retriever(),
)

qa.run(query)

Am I doing something wrong here? Need help.

Did u get a solution. I too am stuck with similar issue.

Hi, its been a while. Did you find anything?
Are the results coming out good?