[Question]: Adding documents to existing chroma index

[srsingh24]

Question Validation

• I have searched both the documentation and discord for an answer.

Question

I was trying to figure out how to add documents to ChromaVectorStore after the initial initialization. I came across the following answer on StackOverFlow
that I should set the mode argument to append. While this was helpful, I wanted to cross check the docs and the source code but could not find any mention of the mode argument. Where can I find more information on it?

[dosubot]

Hello there, @srsingh24! I'm a bot here to help you with bug fixes, questions, and contributing. Let's tackle this issue together!

The ChromaVectorStore class does not have a mode argument mentioned in the documentation or source code related to adding documents. Instead, you can add documents to an existing ChromaVectorStore using the add method. Here's an example of how to do this:

import chromadb
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core.schema import TextNode

# Creating a Chroma client
chroma_client = chromadb.EphemeralClient()
chroma_collection = chroma_client.create_collection("quickstart")

# Construct vector store
vector_store = ChromaVectorStore(
    chroma_collection=chroma_collection,
)

# Defining some sample data
nodes = [
    TextNode(
        text=(
            "Michael Jordan is a retired professional basketball player,"
            " widely regarded as one of the greatest basketball players of all"
            " time."
        ),
        metadata={
            "category": "Sports",
            "country": "United States",
        },
    ),
    # Add more TextNode instances as needed
]

# Adding documents to the ChromaVectorStore
for node in nodes:
    vector_store.add(node)

This method allows you to add new documents to the vector store without needing a mode argument ^[1][2][3].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[srsingh24]

@dosu thank you. How do I initialize the ChromaVectorStore with an OpenAI embedding model?

[dosubot]

To initialize the ChromaVectorStore with an OpenAI embedding model, follow these steps:

1. Set up the OpenAI API key.
2. Install the Chroma library using pip install chromadb.
3. Create a Chroma client and collection.
4. Define the embedding model using OpenAI.
5. Load your documents.
6. Set up the ChromaVectorStore and load the data.

Here's a code snippet to guide you:

# Set up OpenAI
import os
import getpass

os.environ["OPENAI_API_KEY"] = getpass.getpass("OpenAI API Key:")
import openai

openai.api_key = os.environ["OPENAI_API_KEY"]

# Install Chroma
# pip install chromadb

# Create client and a new collection
import chromadb
from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core import StorageContext

# Load documents
documents = SimpleDirectoryReader("./data/paul_graham/").load_data()

# Initialize client, setting path to save data
db = chromadb.PersistentClient(path="./chroma_db")

# Create collection
chroma_collection = db.get_or_create_collection("quickstart")

# Define embedding function using OpenAI
embed_model = openai.Embedding.create  # Assuming you have a function to create embeddings

# Assign Chroma as the vector_store to the context
vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
storage_context = StorageContext.from_defaults(vector_store=vector_store)

# Create your index
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context, embed_model=embed_model
)

# Query Data
query_engine = index.as_query_engine()
response = query_engine.query("What did the author do growing up?")
print(response)

This setup configures the ChromaVectorStore with an OpenAI embedding model, allowing you to persist data and avoid re-indexing ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[srsingh24]

@dosu I want to initialize an empty chroma vector store with the LlamaIndex OpenAI Embeddings: https://docs.llamaindex.ai/en/stable/examples/embeddings/OpenAI/

Next, I will to use the following that you provided to add documents as I receive them:

import chromadb
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core.schema import TextNode

# Creating a Chroma client
chroma_client = chromadb.EphemeralClient()
chroma_collection = chroma_client.create_collection("quickstart")

# Construct vector store
vector_store = ChromaVectorStore(
    chroma_collection=chroma_collection,
)

# Defining some sample data
nodes = [
    TextNode(
        text=(
            "Michael Jordan is a retired professional basketball player,"
            " widely regarded as one of the greatest basketball players of all"
            " time."
        ),
        metadata={
            "category": "Sports",
            "country": "United States",
        },
    ),
    # Add more TextNode instances as needed
]

# Adding documents to the ChromaVectorStore
for node in nodes:
    vector_store.add(node)

[dosubot]

To initialize an empty Chroma vector store with LlamaIndex OpenAI embeddings and add documents as they are received, follow these steps:

1. Install Required Packages:

Ensure you have the necessary packages installed:

pip install chromadb llama-index-embeddings-openai

2. Initialize the Chroma Client and Collection:

import chromadb
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core import StorageContext, VectorStoreIndex
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.core.schema import TextNode

# Initialize Chroma client
chroma_client = chromadb.EphemeralClient()
chroma_collection = chroma_client.create_collection("quickstart")

# Define embedding model
embed_model = OpenAIEmbedding(model="text-embedding-3-large")

# Assign Chroma as the vector_store to the context
vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
storage_context = StorageContext.from_defaults(vector_store=vector_store)

3. Initialize an Empty VectorStoreIndex:

# Initialize an empty VectorStoreIndex
index = VectorStoreIndex([], storage_context=storage_context, embed_model=embed_model)

4. Add Documents to the Index as They Are Received:

# Assuming `new_documents` is a list of TextNode instances you receive
for node in new_documents:
    index.insert(node)

This setup allows you to initialize an empty Chroma vector store with OpenAI embeddings and dynamically add documents to it as they are received ^[1][2].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[srsingh24]

@dosu, thanks! And then once i have loaded in documents, how do I load my chromadb without having to re-initialize the vector index without losing all my old documents?

[dosubot]

To load your ChromaDB without re-initializing the vector index and without losing your old documents, you can directly load the stored vectors without loading your documents or creating a new VectorStoreIndex. Here's how you can do it:

import chromadb
from llama_index.core import VectorStoreIndex
from llama_index.vector_stores.chroma import ChromaVectorStore
from llama_index.core import StorageContext

# Initialize client
db = chromadb.PersistentClient(path="./chroma_db")

# Get collection
chroma_collection = db.get_or_create_collection("quickstart")

# Assign Chroma as the vector_store to the context
vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
storage_context = StorageContext.from_defaults(vector_store=vector_store)

# Load your index from stored vectors
index = VectorStoreIndex.from_vector_store(
    vector_store, storage_context=storage_context
)

# Create a query engine
query_engine = index.as_query_engine()
response = query_engine.query("What is llama2?")
print(response)

This approach allows you to load the existing vector index without re-initializing it or losing your previously stored documents ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[srsingh24]

@dosu, I tried your approach of adding documents to the chromavectorstore:

# Defining some sample data
nodes = [
    TextNode(
        text=(
            "Michael Jordan is a retired professional basketball player,"
            " widely regarded as one of the greatest basketball players of all"
            " time."
        ),
        metadata={
            "category": "Sports",
            "country": "United States",
        },
    ),
    # Add more TextNode instances as needed
]

# Adding documents to the ChromaVectorStore
for node in nodes:
    vector_store.add(node)

and I get the following error:

TypeError                                 Traceback (most recent call last)
Cell In[16], line 19
      1 # Defining some sample data
      2 # nodes = [
      3 #     TextNode(
   (...)
     16 
     17 # Adding documents to the ChromaVectorStore
     18 for node in nodes:
---> 19     vector_store.add(node)

File ~site-packages/llama_index/vector_stores/chroma/base.py:278, in ChromaVectorStore.add(self, nodes, **add_kwargs)
    275 node_chunks = chunk_list(nodes, max_chunk_size)
    277 all_ids = []
--> 278 for node_chunk in node_chunks:
    279     embeddings = []
    280     metadatas = []

File ~site-packages/llama_index/vector_stores/chroma/base.py:107, in chunk_list(lst, max_chunk_size)
     95 def chunk_list(
     96     lst: List[BaseNode], max_chunk_size: int
     97 ) -> Generator[List[BaseNode], None, None]:
     98     """Yield successive max_chunk_size-sized chunks from lst.
     99 
    100     Args:
   (...)
    105         Generator[List[BaseNode], None, None]: list of nodes with embeddings
    106     """
--> 107     for i in range(0, len(lst), max_chunk_size):
    108         yield lst[i : i + max_chunk_size]

TypeError: object of type 'TextNode' has no len()

[dosubot]

The error you're encountering is due to passing a single TextNode object to the add method, which expects a list of BaseNode objects. To resolve this, you need to wrap your TextNode instances in a list before passing them to the add method. Here's how you can modify your code:

# Defining some sample data
nodes = [
    TextNode(
        text=(
            "Michael Jordan is a retired professional basketball player,"
            " widely regarded as one of the greatest basketball players of all"
            " time."
        ),
        metadata={
            "category": "Sports",
            "country": "United States",
        },
    ),
    # Add more TextNode instances as needed
]

# Adding documents to the ChromaVectorStore
vector_store.add(nodes)  # Pass the list of nodes directly

By passing the entire list of TextNode objects to the add method, you should avoid the TypeError ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}