Chromadb: how to convert dict result from "Chroma.get()" to langchain.vectorstores.chroma.Chroma object

[RickyGunawan09]

hi guys,
so i have a condition where i have to maintain chromadb for each document at any time by id and only query the selected documents with the required id.

but I found a problem when after I get data from chromadb it becomes to dictionary not the class object. so how can I return the data that I have gotten into a langchain.vectorstores.chroma.Chroma object. i'm using pymongo to store the ids and query back to chromadb.

this is maybe the simple code to get the dictionary from Chroma.get():

from langchain.vectorstores import Chroma
import uuid
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.document_loaders import TextLoader

vectorstore = Chroma(persist_directory="./db",embedding_function=HuggingFaceEmbeddings())

loader = TextLoader('./test2.txt')
print("Loader : ", loader)
document = loader.load()
print("document : ", document)

uuids = [str(uuid.uuid4()) for x in range(len(document))]

# multidocs = vectorstore.from_documents(document, embedding=HuggingFaceEmbeddings(),persist_directory="./db")
# print("multidocs",multidocs)
# >>> multidocs <langchain.vectorstores.chroma.Chroma object at 0x000001CDCF10B700>

vectorstore.from_documents(document, embedding=HuggingFaceEmbeddings(),persist_directory="./db",ids=uuids)
vectorstore.persist()
test_get=vectorstore.get(ids=[uuids[0]])

so my question, how to process the output so it becomes back to langchain.vectorstores.chroma.Chroma object so i can run .as_retriever() for QAchain.
thank you.

[dosubot[bot]]

🤖

Hello,

From your code, it seems like you're trying to convert the dictionary returned by Chroma.get() back into a Chroma object. However, the Chroma.get() method is designed to return a dictionary, not a Chroma object. The dictionary contains the data of the documents, metadatas, distances, and optionally embeddings, which are the results of the query.

If you want to use the .as_retriever() method for QAchain, you would typically use it on a Chroma object, not on the results of a Chroma.get() call. The Chroma object represents the entire vectorstore, while the results of Chroma.get() are just a subset of the data in the vectorstore.

Here's an example of how you might use the .as_retriever() method:

from langchain.vectorstores import Chroma
from langchain.retrievers import QAChain

# Initialize the Chroma vectorstore
vectorstore = Chroma(persist_directory="./db",embedding_function=HuggingFaceEmbeddings())

# Use the vectorstore as a retriever for QAChain
retriever = QAChain(vectorstore.as_retriever())

In this example, vectorstore.as_retriever() is used to create a retriever that can be used with QAChain. The retriever uses the entire vectorstore, not just the results of a single Chroma.get() call.

If you need to work with a subset of the data in the vectorstore, you might consider creating a new Chroma object and adding only the documents you need to it. However, this would involve re-embedding the documents, which could be computationally expensive.

I hope this helps! If you have further questions, feel free to ask.

Sources

• libs/langchain/langchain/vectorstores/chroma.py
• libs/langchain/tests/integration_tests/vectorstores/test_chroma.py
• docs/extras/integrations/vectorstores/chroma.ipynb

---

This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.

Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.

If you want to continue the conversation, start your reply with @dosu-bot.