Key Capabilities
Persistent Storage
All data and computed results are automatically stored and versioned.
Incremental Updates
Data transformations run automatically on new data—no orchestration code needed.
Multimodal-Native
Images, video, audio, and documents integrate seamlessly with structured data.
AI Integration
Built-in support for OpenAI, Anthropic, Gemini, Hugging Face, and dozens more.
Get started
Quick Start
Install Pixeltable and run your first pipeline in 5 minutes.
10-Minute Tour
See Pixeltable in action with a hands-on image workflow.
Core Concepts
Learn about tables, computed columns, views, and the type system.
SDK Reference
Complete API reference for the Pixeltable Python SDK.
Core Primitives
Pixeltable provides a small set of primitives that compose into any multimodal AI workflow, including but not limited to:| Primitive | What It Enables |
|---|---|
pxt.create_table() with pxt.Image, pxt.Video, pxt.Audio, pxt.Document | Store any multimodal data natively |
pxt.create_view() with iterators | Extract frames from video, chunk documents, split audio |
add_computed_column() | Run any AI model or transformation—incrementally |
add_embedding_index() | Semantic search on any column |
@pxt.udf / @pxt.query | Extend with your own Python code |
select(), where(), order_by() | SQL-like querying with Python syntax |
history(), revert() | Time travel and version control |
pxt.replicate(), pxt.publish() | Share and replicate datasets via Pixeltable Cloud |
What can you build?
Declarative Pipelines
Replace complex orchestration with simple computed columns. Define transformations once—they run automatically on all data.
Multimodal Workloads
Production RAG with automatic embedding indexing. Find relevant scenes in video. Semantic search across text, images, and audio.
Version Control and Lineage
Automatic versioning on every change. Time travel queries to any point. Full data lineage for reproducibility.
AI Agents & MCP
Build tool-calling agents with persistent memory, MCP server integration, and automatic conversation history.
ML Feature Engineering
Curate, augment, and export data to PyTorch, Parquet, COCO format, LanceDB, and pandas for training and analytics.
Explore by use case
- AI & LLMs
- Media Processing
- Data Management
- RAG Pipeline — Document retrieval & generation
- Vision Analysis — GPT-4 Vision on images
- Audio Transcription — Speech-to-text with Whisper
- Document Chunking — Split docs for RAG