​

Web UI vs. Python

With Oxen.ai you can either perform your labeling workflows directly in the web interface, or build your own using the Python Library.

​

Using the UI

Oxen.ai has a data frame interface built in that allows your to view, query and edit any csv, json, or parquet file. The data can be multimodal, allowing you to see images and text side by side.

If you are using google sheets or excel for your datasets this is a great place to get started before building out more powerful workflows.

​

Upload Your Dataset

Navigate to the repository you want to work with, or create a new one. If you need starter data, you can find example datasets on our explore page. To follow along with the example, you can clone the Synthetic Political Spam Dataset we are using.

oxen clone https://hub.oxen.ai/Laurence/political-spam --all
oxen create-remote --name "your-username/political-spam"
oxen config --set-remote origin "your-username/political-spam"
oxen push origin main

​

Edit File

Open the file you want to edit and press the Edit button above your data.

​

Editing Columns

To edit columns, go to the Schema section on the left of your dataset. Here, there are four actions you can take:

• Add a Column: Click the Add Column button.

• Delete a Column: Click the Trash icon next to the column name you would like to delete.

• Edit a Column: Click the pencil icon to change the column name or the data type.

• Hide a Column: Click the Eye icon to the left of the column name you would like to hide for the edit.

​

Editing Rows

To edit your rows, there are two actions you can take:

• Add a Row: Click the Add Row button to get a new blank row at the end of your dataset.

• Delete a Row: Click one of the cells in the row you would like to delete. Then click the large red “delete” button on the right of the screen.

To undo a deleted row, click the revert button.

​

Editing Cells

To edit cells, click on the cell you would like to edit and make any changes you would like to the data. Then click the Save button to save progress.

To undo any saved changes, click the revert button.

​

Committing Changes

To commit your changes, click on the Commit button, write your commit message, choose the branch, and click “Commit changes”.

​

Returning to Data Frame

To return to the original dataframe, click the Return to dataframe button on the left of the commit button.

Congratulations! You’ve just seen how easy it is to edit your datasets without downloading on Oxen.AI. For more examples of different uses, click here!

​

Using the Python Library

The web interface is built on top of HTTP APIs that are also exposed through Oxen.ai’s Python Library. This makes it easy to interact with data frames programatically and build your own custom labeling tools. Under the hood the dataset will be indexed into DuckDB within a Workspace to make it fast to query and update the data before fully committing it back to your repository.

​

Indexing a Data Frame

When you instantiate a DataFrame object, the data will automatically get indexed into an Oxen Workspace in an uncommitted state. This gives you fast read/write access to the data frame without committing it.

from oxen import DataFrame

# Connect to the data frame with the repository name and file path
df = DataFrame("my-username/spam-or-ham", "data.tsv")

​

Fetching Rows

To get the full size of the data frame you can use the size() method.

width, height = df.size()

To get an individual row index, pass in the offset to the get_row method.

# Get the Nth row of the data frame
row = df.get_row(42)

​

Insert Rows

To add a row to the end of the dataset, use the insert_row function with a dictionary that contains the column names as keys, and the row data as values. This will return an internal _oxen_id that can be used to access the same row again.

# Add a row
row_id = df.insert_row({"category": "spam", "message": "CLICK HERE TO WIN INSTANTLY."})

The _oxen_id can be used to fetch the same row again.

# Get a row by the internal '_oxen_id'
row = df.get_row_by_id(row_id)

​

Update Rows

To update a row, pass in the _oxen_id as well as the columns you want to update as a dictionary. Only the keys that are present will be updated, leaving the rest unchanged.

row = df.update_row(row_id, {"category": "new_category"})

​

Delete Rows

To delete a row, pass in the _oxen_id to the delete_row method.

df.delete_row(row_id)

​

Committing Changes

Finally, when you are happy with the dataset, and have reviewed the changes, you can commit this dataset to the repositories history with commit and a message.

df.commit("Update label")

To see the full set of APIs, check out the Python Documentation.