Skip to content

Marqo Clothing Apparel Simple CLI Demo

Getting Started

  1. Download the Dataset from Clothing Dataset into the directory where the script is found.

  2. Run this command inside the script directory to setup an HTTP server

    python3 -m http.server 8222
    This is for the marqo docker container to read files from local os. For more info on this please visit this link.

  3. Make sure to run the Marqo docker container via the following command:

    docker run --name marqo -it --privileged -p 8882:8882 --add-host host.docker.internal:host-gateway marqoai/marqo:latest

  4. Install Marqo

    pip install marqo

  5. Run the script via the following command:


import marqo
import pprint
import pandas as pd

mq = marqo.Client(url='http://localhost:8882') # Connection to Marqo Docker Container

dataset_path = "http://localhost:8222/" # Place your file path here (directory where http server is setup)

def load_index(index_name: str, number_data: int) -> None:
        shirt_data = pd.read_csv('clothing-dataset/images.csv').head(number_data)[['image','label','kids']].to_dict('records')

        # dataset came from this link:
        # the .csv file has the following headers:
        # image, sender_id, label, kids
        # (image name, id of the sender who sent the pictures from sender_id, what kind of clothing it is, whether or not the clothing is for kids)
        # Dataset Example:.
        # 4285fab0-751a-4b74-8e9b-43af05deee22,124,Not sure,False
        # 70045b01-b350-4918-be74-2f627290ad7a,95,Skirt,False

        for data in shirt_data:
            path = "http://host.docker.internal:8222/clothing-dataset/images/" + data['image'] + ".jpg"
            data['image'] = path

        settings = {
            "treat_urls_and_pointers_as_images":True,   # allows us to find an image file and index it

        mq.create_index(index_name, **settings)


        print("Index successfully created.")

    except Exception as e:
        print("Index already exists.")

def delete_index(index_name: str):
        print("Index successfully deleted.")
    except Exception as e:
        print("Index does not exist.")

def delete_doc_from_index(index_name:str, doc_ids:list[str]):
    results = mq.index(index_name).delete_documents(ids=doc_ids)
    return results

def search_index_text(index_name:str, query_text: str, search_method: str):
    results = mq.index(index_name).search(

    # Marqo also has other features such as searhcing based on a specific attribute field and query fitlering
    # refer to the documentation on how these features work (
    return results

def search_index_image(index_name:str, image_name: str):
    # make sure the image is located inside the directory in which the python http server is running

    image_path = "http://host.docker.internal:8222/" + image_name

    results = mq.index(index_name).search(image_path)

    return results

def get_index_stats(index_name: str) -> dict:
    results = mq.index(index_name).get_stats()
    return results

def main():
    print("Welcome to Marqo Demo!")
    while True:
        action = int(input('''
What would you like to do?
1) Create an Index
2) Delete an Index
3) Search from an Index
4) Show Index Stats
5) Delete a document from an Index
6) Quit

Action: '''))

        if action == 1:
            index_name = input("Index name: ")
            no_of_items = int(input("No. of items in dataset: "))

            load_index(index_name, no_of_items)
        elif action == 2:
            index_name = input("Index name: ")

        elif action == 3:
            index_name = input("Index name: ")
            search_type = input("Search Type (Text, Image): ")

            if search_type == 'Text':
                search_mode = str(input("Search Mode: (Lexical, Tensor)"))
                query_text = str(input("Query Text: "))

                results = search_index_text(index_name, query_text, search_mode.upper())

            elif search_type == 'Image':
                image_name = str(input("Image name (include MIME type .jpg or .png): "))

                results = search_index_image(index_name, image_name)


        elif action == 4:
            index_name = input("Index name: ")

        elif action == 5:
            index_name = input("Index name: ")
            no_of_docs = int(input("No. of documents to delete: "))
            doc_ids = []

            for i in range(no_of_docs):
                doc_id = input("Document ID: ")

            delete_doc_from_index(index_name, doc_ids)



Function References

  • load_index(index_name, number_data)

    Name Type Description
    index_name String name of index
    number_data Integer number of data lines to parse from dataset

    This function reads the data from the images.csv file found in the dataset and creates an index in the Marqo client. The created index name is based on the value of index_name.

  • delete_index(index_name)

    Name Type Description
    index_name String name of index

    This function deletes an index based on the value of index_name. Function returns an error if index does not exist.

  • delete_doc_from_index(index_name, doc_ids)

    Name Type Description
    index_name String name of index
    doc_ids List[String] list of document ids

    This function deletes the documents found in index index_name based on the document IDs found in doc_ids.

  • search_index_text(index_name, query_text, search_method)

    Name Type Description
    index_name String name of index
    query_text String search query text
    search_method String search method (Lexical or Tensor)

    This function runs a TEXT search query query_text on the documents found in index index_name. The search method is based on search_method.

    A Python dictionary is returned containing all the results of the query.

  • search_index_iamge(index_name, image_name)

    Name Type Description
    index_name String name of index
    image_name String name of image used for searching

    This function runs an IMAGE tensor search query based on image_name on the documents found in index index_name.

    A Python dictionary is returned containing all the results of the query.

  • get_index_stats(index_name)

    Name Type Description
    index_name String name of index

    This function returns the index stats based on index_name.


Feel free to checkout the code in order to have a better understanding on how Marqo functions are used :).