Share AI Tables Seamlessly

Collaborate on multimodal AI datasets with complete reproducibility. Share tables between local and cloud with read-only replicas, complete versioning, and full data lineage.

Documentation

MULTIMODAL

Handle all media types

AI-READY

Support AI workloads

VERSIONED

Track versions, stay synced

REPRODUCIBLE

Same data, same results

Key Features

Complete Multimodal API

Work with video, audio, images, and documents using one API

Publishers share multimodal data in simple tables
Recipients access all media types through one interface

Complete Sharing System

Eliminate manual file transfers and disconnected script sharing

Publishers bundle data and all generated outputs in tables
Recipients get embeddings and model outputs automatically

Complete Control & Flexibility

Maintain complete reproducibility, versioning, and lineage

Publishers push updates and control access with public or private sharing
Recipients maintain local development workflows with on-demand sync

How It Works

Publish

Publish your local table to Pixeltable Cloud with a single command. Control access with public or private visibility.

Replicate

Team members replicate the shared table to their local environment as a read-only replica. All data, schema, and computed columns are copied.

Sync

Push and pull updates seamlessly. Changes are automatically tracked with full versioning and lineage.

Run a recipe on Cloud

pxt.publish / pxt.replicate retired June 2026. Start from a starter-kit recipe instead — hosted API + catalog directory on Community main.

Code Examples

Three simple workflows for complete data sharing and collaboration

Note: Dataset publish/replicate retired June 2026. Use the recipe gallery to run a hosted API, or pxt deploy for your own service.

publish_example.py

1import pixeltable as pxt
2from pixeltable.functions import openai
3import os
4
5# Set your Pixeltable Cloud API key
6os.environ['PIXELTABLE_API_KEY'] = '<YOUR_API_KEY>'
7
8# Create and populate your table
9t = pxt.create_table('images', {'input_image': pxt.Image})
10
11# Add AI-powered computed columns
12t.add_computed_column(
13    description=openai.chat_completions(
14        messages=[{
15            'role': 'user',
16            'content': [
17                {'type': 'text', 'text': "Describe this image in detail."},
18                {'type': 'image_url', 'image_url': {'url': t.input_image}},
19            ],
20        }],
21        model='gpt-4o-mini',
22    ).choices[0].message.content
23)
24
25# Insert data
26t.insert(input_image='https://example.com/photo.jpg')
27
28# Publish to cloud (make it shareable)
29pxt.publish(
30    source=t,
31    destination_uri='pxt://org_name/shared_images',
32    access='private'  # or 'public' for open access
33)

Use Cases

Perfect for teams collaborating on AI and ML projects

Team Collaboration

Share training datasets, models, and results across DS/AI/ML teams with complete versioning and reproducibility.

Data Science Projects

Collaborate on data analysis with shared cleaned datasets, computed embeddings, and version-tracked reproducibility.

Production Deployment

Develop locally, then publish to cloud for production. Push updates when ready.

Dataset Distribution

Share public datasets and benchmark results with the research community for reproducible science.

Start sharing today

Join developers and researchers collaborating on multimodal AI datasets.

Documentation