Open In Colab

Chroma Multi-Modal Demo with LlamaIndex

Chroma is a AI-native open-source vector database focused on developer productivity and happiness. Chroma is licensed under Apache 2.0.

Discord    License    Integration Tests

Chroma is fully-typed, fully-tested and fully-documented.

Install Chroma with:

pip install chromadb

Chroma runs in various modes. See below for examples of each integrated with LangChain.

  • in-memory - in a python script or jupyter notebook

  • in-memory with persistance - in a script or notebook and save/load to disk

  • in a docker container - as a server running your local machine or in the cloud

Like any other database, you can:

  • .add

  • .get

  • .update

  • .upsert

  • .delete

  • .peek

  • and .query runs the similarity search.

View full docs at docs.

Basic Example

In this basic example, we take the a Paul Graham essay, split it into chunks, embed it using an open-source embedding model, load it into Chroma, and then query it.

If you’re opening this Notebook on colab, you will probably need to install LlamaIndex 🦙.

!pip install llama-index

Creating a Chroma Index

!pip install llama-index chromadb --quiet
!pip install chromadb==0.4.17
!pip install sentence-transformers
!pip install pydantic==1.10.11
!pip install open-clip-torch
# import
from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext
from llama_index.vector_stores import ChromaVectorStore
from import StorageContext
from llama_index.embeddings import HuggingFaceEmbedding
from IPython.display import Markdown, display
import chromadb
# set up OpenAI
import os
import openai

openai.api_key = OPENAI_API_KEY

Download Images and Texts from Wikipedia

import requests

def get_wikipedia_images(title):
    response = requests.get(
            "action": "query",
            "format": "json",
            "titles": title,
            "prop": "imageinfo",
            "iiprop": "url|dimensions|mime",
            "generator": "images",
            "gimlimit": "50",
    image_urls = []
    for page in response["query"]["pages"].values():
        if page["imageinfo"][0]["url"].endswith(".jpg") or page["imageinfo"][
    return image_urls
from pathlib import Path
import urllib.request

image_uuid = 0

wiki_titles = {
    "Tesla Model X",
    "Pablo Picasso",
    "The Lord of the Rings",
    "The Matrix",
    "The Simpsons",

data_path = Path("mixed_wiki")
if not data_path.exists():

for title in wiki_titles:
    response = requests.get(
            "action": "query",
            "format": "json",
            "titles": title,
            "prop": "extracts",
            "explaintext": True,
    page = next(iter(response["query"]["pages"].values()))
    wiki_text = page["extract"]

    with open(data_path / f"{title}.txt", "w") as fp:

    images_per_wiki = 0
        # page_py =
        list_img_urls = get_wikipedia_images(title)
        # print(list_img_urls)

        for url in list_img_urls:
            if url.endswith(".jpg") or url.endswith(".png"):
                image_uuid += 1
                # image_file_name = title + "_" + url.split("/")[-1]

                    url, data_path / f"{image_uuid}.jpg"
                images_per_wiki += 1
                # Limit the number of images downloaded per wiki page to 15
                if images_per_wiki > MAX_IMAGES_PER_WIKI:
        print(str(Exception("No images found for Wikipedia page: ")) + title)

Set the embedding model

from chromadb.utils.embedding_functions import OpenCLIPEmbeddingFunction

# set defalut text and image embedding functions
embedding_function = OpenCLIPEmbeddingFunction()
/Users/haotianzhang/llama_index/venv/lib/python3.11/site-packages/tqdm/ TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See
  from .autonotebook import tqdm as notebook_tqdm

Build Chroma Multi-Modal Index with LlamaIndex

from llama_index.indices.multi_modal.base import MultiModalVectorStoreIndex
from llama_index.vector_stores import QdrantVectorStore
from llama_index import SimpleDirectoryReader, StorageContext
from chromadb.utils.data_loaders import ImageLoader

image_loader = ImageLoader()

# create client and a new collection
chroma_client = chromadb.EphemeralClient()
chroma_collection = chroma_client.create_collection(

# load documents
documents = SimpleDirectoryReader("./mixed_wiki/").load_data()

# set up ChromaVectorStore and load in data
vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(

Retrieve results from Multi-Modal Index

retriever = index.as_retriever(similarity_top_k=50)
retrieval_results = retriever.retrieve("Picasso famous paintings")
# print(retrieval_results)
from llama_index.schema import ImageNode
from llama_index.response.notebook_utils import (

image_results = []
cnt = 0
for r in retrieval_results:
    if isinstance(r.node, ImageNode):
        if cnt < MAX_RES:
        cnt += 1

display_image_uris(image_results, [3, 3], top_k=2)

Node ID: 13adcbba-fe8b-4d51-9139-fb1c55ffc6be
Similarity: 0.774399292477267
Text: == Artistic legacy == Picasso’s influence was and remains immense and widely acknowledged by his …

Node ID: 4100593e-6b6a-4b5f-8384-98d1c2468204
Similarity: 0.7695965506408678
Text: === Later works to final years: 1949–1973 === Picasso was one of 250 sculptors who exhibited in t…

Node ID: aeed9d43-f9c5-42a9-a7b9-1a3c005e3745
Similarity: 0.7693110304140338
Text: Pablo Ruiz Picasso (25 October 1881 – 8 April 1973) was a Spanish painter, sculptor, printmaker, …

Node ID: 5a6613b6-b599-4e40-92f2-231e10ed54f6
Similarity: 0.7656537748231977
Text: === The Basel vote === In the 1940s, a Swiss insurance company based in Basel had bought two pain…

Node ID: cc17454c-030d-4f86-a12e-342d0582f4d3
Similarity: 0.7639671751819532
Text: == Style and technique ==

Picasso was exceptionally prolific throughout his long lifetime. At hi…