Pixeltable Documentation

The only open source Python library providing declarative data infrastructure for building multimodal AI applications, enabling incremental storage, transformation, indexing, retrieval, and orchestration of data.

Docs

Changelog: Release notes for Pixeltable covering new features, performance improvements, bug fixes, and breaking changes across SDK versions.
Agentic Patterns: Implement reflection, planning, tool use, and multi-agent collaboration patterns in Pixeltable to build cognitive agents on tabular pipelines.
Use tool calling and MCP servers with LLMs: Connect LLMs to Pixeltable UDFs, queries, and external MCP servers so agents can search data, call APIs, and execute typed Python tools.
Build an agent with memory: Give LLM agents persistent short-term and long-term memory in Pixeltable using conversation tables, summarization, and embedding-based recall.
Look up structured data with retrieval UDFs: Expose Pixeltable tables as retrieval UDFs so LLM agents can look up structured rows by key, run filters, and return typed results as tools.
Build a RAG pipeline: Build a complete RAG pipeline in Pixeltable with document ingestion, chunking, embeddings, semantic search, and LLM answer generation.
Use a table pipeline as a reusable function: Wrap an entire Pixeltable table pipeline as a reusable function that agents and other tables can call with typed inputs and outputs.
Extract audio from video: Extract audio tracks from video files in Pixeltable using FFmpeg-backed UDFs for transcription, music analysis, and audio ML pipelines.
Summarize podcasts and audio: Build a podcast summarization pipeline in Pixeltable that transcribes audio with Whisper and generates structured summaries with an LLM.
Convert text to speech: Generate speech audio from text columns in Pixeltable using OpenAI, ElevenLabs, and other TTS providers via declarative computed columns.
Transcribe audio files with Whisper: Transcribe audio and video files into searchable text columns in Pixeltable using Whisper, WhisperX, and other speech-to-text models.
Create custom aggregate functions (UDAs): Define user-defined aggregate functions (UDAs) in Pixeltable to compute custom group-by statistics over rows with init, update, and value steps.
Custom Iterators: Build custom ComponentIterators in Pixeltable to split documents, videos, audio, or other media into rows for view-based processing.
Split data into multiple rows with iterators: Split a single row into many derived rows in Pixeltable using built-in component iterators for chunks, frames, video segments, and tiled data.
Get fast feedback on transformations: Develop Pixeltable pipelines iteratively with versioned tables, computed columns, and time-travel queries to refine logic without losing data.
Your Backend for Multimodal AI Applications: Use Pixeltable as a unified backend for multimodal AI apps that store, transform, embed, and query images, video, audio, and documents.
Join tables to combine data: Join multiple Pixeltable tables on shared keys to combine metadata, embeddings, and computed columns into unified, queryable result sets.
Time Zones: Work with timezone-aware timestamps in Pixeltable: store UTC, convert between zones, and run accurate date queries across regions.
Track changes and revert to previous versions: Track every change to your Pixeltable tables, run time-travel queries against historical snapshots, and revert columns to previous versions.
Configure API keys for AI services: Manage API keys for OpenAI, Anthropic, and other providers in Pixeltable using environment variables and config files for safe pipelines.
Extract fields from LLM JSON responses: Extract structured fields from LLM JSON responses in Pixeltable using path expressions, validation, and computed columns for downstream queries.
Add unique identifiers to your tables: Add UUID identity columns to Pixeltable tables to give every row a stable, globally unique identifier for joins, exports, and external systems.
Export data for ML training: Export Pixeltable tables and views to PyTorch DataLoaders for training image, video, audio, and text models with streaming batches.
Upload media to S3 and other cloud storage: Upload media files generated by Pixeltable computed columns to Amazon S3 and other cloud storage providers for sharing and downstream use.
Export data to SQL databases: Export Pixeltable tables to PostgreSQL, SQLite, and other SQL databases for BI tools, dashboards, and downstream analytics workflows.
Import data from CSV files: Import CSV files into Pixeltable tables with automatic type inference, column mapping, and incremental loading for tabular datasets.
Import data from Excel files: Load XLSX and Excel spreadsheets into Pixeltable tables with sheet selection, header handling, and type inference for analysis pipelines.
Import data from Hugging Face datasets: Import Hugging Face datasets directly into Pixeltable tables for vision, text, and multimodal ML training and evaluation workflows.
Import data from JSON files: Load JSON and JSONL files into Pixeltable tables with nested object support, schema inference, and streaming ingestion of large datasets.
Import data from Parquet files: Ingest Apache Parquet files into Pixeltable tables for fast columnar loading of large analytics, ML training, and feature-store datasets.
Load media from S3 and other cloud storage: Load images, videos, audio, and documents from Amazon S3 and other cloud storage buckets into Pixeltable tables using URL references.
Sample data for training and testing: Create train, validation, and test splits in Pixeltable using reproducible row sampling, stratification, and seeded random shuffles.
Add watermarks to images: Add text or image watermarks to photos in Pixeltable using PIL-backed computed columns for branding, attribution, and rights protection.
Adjust image opacity: Adjust image transparency and alpha channels in Pixeltable with PIL operations for compositing layered images, overlays, and watermark effects.
Apply image filters: Apply blur, sharpen, edge detection, and other PIL image filters in Pixeltable using declarative computed columns to process datasets at scale.
Adjust image brightness and contrast: Tune image brightness, contrast, saturation, and sharpness in Pixeltable using PIL ImageEnhance operations exposed as computed columns at scale.
Detect objects in images: Detect objects in images at scale in Pixeltable using YOLOX, DETR, and other vision models with bounding box outputs and confidence scores.
Compare object detection and panoptic segmentation: Compare object detection bounding boxes with panoptic segmentation masks in Pixeltable to pick the right vision approach for your task.
Generate captions for images: Generate natural language captions for images in Pixeltable using BLIP, GPT-4 Vision, and other multimodal models with computed columns.
Transform images with AI-powered editing: Run image-to-image transformations in Pixeltable with diffusion models, style transfer, and AI editing APIs through computed columns.
Transform images with PIL operations: Resize, crop, rotate, and transform images at scale in Pixeltable using PIL operations exposed as built-in UDFs for computed columns.
Convert color images to grayscale: Convert RGB images to grayscale in Pixeltable using PIL mode conversion on computed columns for preprocessing, OCR, and machine learning pipelines.
Visualize object detections: Draw bounding boxes, labels, and segmentation masks over images in Pixeltable to visualize object detection and vision model outputs.
Analyze images in batch with AI vision: Run GPT-4 Vision, Claude, and Gemini over large image batches in Pixeltable with computed columns, retries, and structured outputs.
Extract structured data from images: Extract structured JSON from images in Pixeltable using vision LLMs with Pydantic schemas, validation, and typed computed columns.
Create text embeddings with OpenAI: Embed text columns with OpenAI embedding models in Pixeltable to build vector indices for semantic search and retrieval-augmented generation.
Build semantic search for text: Build semantic text search in Pixeltable with embedding indices, similarity queries, and top-k retrieval over documents and chunks.
Find similar images with CLIP: Find visually similar images in Pixeltable using CLIP and other vision embeddings with similarity search over indexed image columns.
Split documents into chunks for RAG: Split documents into RAG-ready chunks in Pixeltable using DocumentSplitter with overlap, token limits, and structural heading awareness.
Extract text from PowerPoint, Word, and Excel files: Extract text from PowerPoint, Word, and Excel office documents in Pixeltable for indexing, search, and downstream LLM RAG workflows.
Extract named entities from text: Extract named entities, relations, and structured fields from text in Pixeltable using LLMs with Pydantic schemas and typed outputs.
Summarize text with LLMs: Summarize long documents, articles, and transcripts in Pixeltable using LLMs with chunking, map-reduce, and structured output schemas.
Translate text between languages: Translate text columns between languages in Pixeltable using OpenAI, Anthropic, and other LLM providers through declarative computed columns.
Add text overlays to videos: Overlay text, captions, and timestamps onto videos in Pixeltable using FFmpeg-backed computed columns and frame-level transformations.
Extract frames from videos: Split videos into individual frames in Pixeltable with FrameIterator so you can run image models, vision LLMs, and analytics per frame.
Generate videos with AI: Generate AI video clips in Pixeltable from text or image prompts using Runway, Replicate, and other generative video provider integrations.
Generate thumbnails from videos: Generate thumbnail images and preview frames for videos in Pixeltable using FrameIterator views and FFmpeg-backed computed columns at scale.
Create a video slideshow from images: Build slideshow videos from sequences of images in Pixeltable using FFmpeg-backed UDFs with configurable timing, transitions, and audio tracks.
Detect scene changes in videos: Detect scene cuts and shot boundaries in videos using Pixeltable to split footage into segments for indexing, summarization, and editing.
Infrastructure Setup: Organize Pixeltable code, configure storage backends, and design infrastructure for production deployments of multimodal AI pipelines.
Monitoring & Performance: Monitor Pixeltable pipelines in production with structured logging, resource metrics, performance tuning, and provider rate-limit handling.
Production Operations: Operate Pixeltable in production with concurrency control, error handling, schema evolution, and zero-downtime deployment patterns.
Deployment Overview: Compare deployment options for Pixeltable applications across local, server, container, and managed cloud environments to pick the best fit.
Security & Backup: Secure Pixeltable deployments with backup strategies, disaster recovery procedures, access controls, and credential management best practices.
Serving Tables and Queries over HTTP: Serve Pixeltable tables and queries as HTTP API endpoints using TOML service definitions or Python code with FastAPIRouter integration.
Working with Anthropic in Pixeltable: Call Claude models from Pixeltable for chat completions, tool calling, and structured outputs as computed columns over text and image data.
Working with Bedrock in Pixeltable: Run Claude, Titan, and other AWS Bedrock foundation models from Pixeltable computed columns for enterprise LLM and embedding workflows.
Working with BFL FLUX in Pixeltable: Generate and edit images with Black Forest Labs FLUX models from Pixeltable using text prompts, image inputs, and computed columns.
Working with Deepseek in Pixeltable: Use DeepSeek chat and reasoning models from Pixeltable computed columns for code generation, math, structured outputs, and tool-using agents.
Working with Microsoft Fabric: Read from and write to Microsoft Fabric OneLake tables and datasets directly from Pixeltable for enterprise data and ML workflows.
Working with fal.ai in Pixeltable: Generate images, video, and audio and run open-source ML models from Pixeltable computed columns using the Fal serverless inference platform.
Working with Fireworks AI in Pixeltable: Call Llama, Mixtral, and other open-source LLMs hosted on Fireworks AI from Pixeltable computed columns for fast chat and text generation.
Working with Gemini in Pixeltable: Use Google Gemini multimodal models from Pixeltable for chat, vision, structured outputs, and embeddings over text, image, and video columns.
Working with Groq in Pixeltable: Run low-latency Llama, Mixtral, and other open LLMs hosted on Groq from Pixeltable computed columns for fast chat and streaming generation.
Working with Hugging Face: Run Hugging Face transformer, vision, and sentence-embedding models locally inside Pixeltable computed columns over text, image, and audio.
Working with Jina AI in Pixeltable: Use Jina AI embedding and reranker models from Pixeltable to build multilingual semantic search and RAG indexes over text and images.
Working with llama.cpp in Pixeltable: Run GGUF quantized LLMs locally with llama.cpp from Pixeltable computed columns for offline chat, generation, and embedding workflows.
Working with Mistral AI in Pixeltable: Use Mistral chat, embedding, and code models from Pixeltable computed columns for European data residency and open-weight LLM pipelines.
Working with Ollama in Pixeltable: Run Llama, Mistral, and other open LLMs locally with Ollama from Pixeltable computed columns for private, offline chat and generation.
Working with OpenAI in Pixeltable: Call GPT-4, embeddings, DALL-E, and Whisper from Pixeltable computed columns for chat, vision, image generation, and speech workflows.
Working with OpenRouter in Pixeltable: Access dozens of LLM providers through a single OpenRouter API from Pixeltable computed columns for chat, vision, and tool calling.
Working with Pydantic in Pixeltable: Define Pydantic schemas in Pixeltable to extract validated structured outputs from LLMs with type checking, defaults, and nested object models.
Working with Replicate in Pixeltable: Run thousands of open-source models hosted on Replicate from Pixeltable computed columns for image, video, audio, and language tasks.
Working with Reve in Pixeltable: Generate images with Reve from Pixeltable using text and reference prompts through declarative computed columns and batch inference.
Working with RunwayML in Pixeltable: Generate and edit AI videos with Runway from Pixeltable using text-to-video, image-to-video, and motion control through computed columns.
Working with Tigris in Pixeltable: Store and serve Pixeltable media files from Tigris globally distributed S3-compatible object storage for low-latency multimodal apps.
Working with Together AI in Pixeltable: Run Llama, Qwen, and other open LLMs hosted on Together AI from Pixeltable computed columns for chat, code, and embedding workflows.
Working with Twelve Labs in Pixeltable: Index and search videos by visual content, speech, and text using Twelve Labs models from Pixeltable computed columns and embeddings.
Working with Voyage AI in Pixeltable: Use Voyage AI embedding and reranker models from Pixeltable to build high-quality retrieval indices for RAG over documents and code.
Transcribing and Indexing Audio and Video in Pixeltable: End-to-end audio transcription pipeline in Pixeltable using Whisper, speaker diarization, and indexed transcripts for search and analysis.
Object Detection in Videos: Detect, track, and visualize objects across video frames in Pixeltable using YOLOX and other vision models with FrameIterator-backed views.
Document Indexing and RAG: Complete RAG demo in Pixeltable that ingests PDFs, chunks text, builds embeddings, runs semantic search, and generates grounded LLM answers.
RAG Operations in Pixeltable: Operate production RAG pipelines in Pixeltable with incremental indexing, versioning, evaluation, and observability over document collections.
Using Label Studio for Annotations with Pixeltable: Send Pixeltable video and image data to Label Studio for human annotation, then sync labeled tasks back into your tables for ML training.
Working with Voxel51 for Visualization in Pixeltable: Visualize Pixeltable images, detections, and embeddings interactively in the Voxel51 FiftyOne app for dataset exploration and quality review.
Cloud Storage: Connect Pixeltable to S3, Google Cloud Storage, Azure Blob, and other cloud storage backends to manage media files and external references.
Embedding Models: Plug custom embedding models into Pixeltable for vector indices, semantic search, and retrieval-augmented generation over your own data.
Ecosystem: Browse Pixeltable integrations with LangChain, FastAPI, PyTorch, Hugging Face, and other AI and ML frameworks for end-to-end pipelines.
Model Hub & Repositories: Browse pre-trained models built into Pixeltable for vision, language, speech, and embeddings across OpenAI, Anthropic, Hugging Face, and more.
Agent Frameworks: Map LangGraph, CrewAI, LangChain, and AutoGen agent concepts to Pixeltable tables, computed columns, and tool-calling UDFs for migration.
DIY Data Pipeline: Replace custom Python scripts, DVC, Airflow, and manual ETL with declarative Pixeltable tables, views, and incremental computed columns.
Hand-Written FastAPI Endpoints: Migrate hand-written FastAPI endpoints into declarative Pixeltable FastAPIRouter routes that serve tables and queries with typed schemas.
RDBMS & Vector DBs: Replace Postgres plus Pinecone plus LangChain stacks with a single Pixeltable system that handles structured data, embeddings, and RAG queries.
Building with LLMs: Build Pixeltable applications faster with AI coding tools like Cursor and Claude using bundled context, examples, and llms.txt resources.
What is Pixeltable?: Pixeltable is declarative AI data infrastructure that provides incremental computed columns, multimodal storage, and versioning in one Python API.
Quick Start: Install Pixeltable, create your first table, add a computed column powered by an LLM, and run your first query in just a few minutes of setup.
10-Minute Tour: Hands-on ten-minute walkthrough of Pixeltable that covers tables, computed columns, views, embedding indices, and multimodal pipelines.
CLI Reference: Serve Pixeltable tables and queries as HTTP endpoints directly from the terminal using the pxt command-line interface for fast iteration.
Configuration: Configure Pixeltable storage paths, providers, API keys, logging, and runtime options through environment variables and config files.
Local Dashboard: Browse Pixeltable tables, inspect pipelines, preview multimedia, and debug computed columns from a local web dashboard that starts automatically.
Data Sharing: Share Pixeltable tables, views, and snapshots across users and projects with read-only access, lineage tracking, and reproducible references.
Embedding Indices: Create and query embedding indices in Pixeltable to power semantic search, similarity lookup, and retrieval-augmented generation pipelines.
External Files: Reference media stored in Amazon S3, GCS, Azure, and other external locations from Pixeltable tables without copying files locally.
Iterators: Use Pixeltable iterators to split documents, video, audio, and images into row-level components for view-based downstream processing.
Multimodal Type System: Pixeltable type system covering scalars, JSON, arrays, image, video, audio, document, and embedding types for structured and ML pipelines.
UDFs in Pixeltable: Write Python user-defined functions in Pixeltable to extend tables with custom logic, model inference, and API calls as typed computed columns.
Version Control and Lineage: Pixeltable automatically versions every table and column, supports time-travel queries, and tracks full data lineage across pipeline changes.
Views: Create virtual derived tables in Pixeltable with views that filter, transform, or expand rows without copying underlying data or computations.
anthropic: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displa…
audio: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
bedrock: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
bfl: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'in…
CodecContextMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
ColumnMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ dis…
ColumnRef: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
ColumnSpec: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline',…
ContainerMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
date: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
deepseek: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
DirContents: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inlin…
DirectoryNode: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline',…
document: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
Expr: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inli…
fabric: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
fal: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'in…
FastAPIRouter: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
fireworks: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displa…
functions: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
gemini: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
groq: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
huggingface: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ disp…
image: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
IndexMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ di…
io: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inli…
jina: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
json: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
llama_cpp: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displa…
math: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
mistralai: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displa…
net: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'in…
ollama: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
openai: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
openrouter: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displ…
pixeltable: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline'…
Query: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline'…
replicate: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displa…
ResultCursor: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline'…
ResultSet: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline',…
reve: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
Row: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline'…
runwayml: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
serving: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
SqlExport: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline',…
StreamMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
string: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
Table: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
TableMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ di…
TableNode: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'inline',…
timestamp: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displa…
together: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
TreeNode
twelvelabs: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ displ…
UpdateStatus: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ dis…
uuid: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: 'i…
VersionMetadata: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ d…
video: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
vision: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
voyageai: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
whisper: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display:…
whisperx: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display…
yolox: <img src="https://img.shields.io/badge/View%20Source%20on%20Github-blue?logo=github&labelColor=gray" alt="View Source on GitHub" style={{ display: '…
Computed Columns: Learn how Pixeltable computed columns turn Python functions into incremental, cached transformations over tables, views, and media data.
Queries and Expressions: Tutorial on Pixeltable queries, expressions, filters, joins, and aggregations for retrieving structured and multimodal data with Python syntax.
Tables and Data Operations: Tutorial on creating Pixeltable tables, inserting rows, updating columns, and using views to build versioned multimodal data pipelines.
Agents & MCP: Build LLM agents in Pixeltable with declarative tool calling, persistent memory tables, and Model Context Protocol server integration.
Backend for AI Apps: Build multimodal AI applications on Pixeltable with declarative pipelines that combine images, video, audio, documents, and language data.
Get Started with Data Sharing: Get started with Pixeltable Cloud to explore, share, and collaborate on multimodal AI datasets and tabular pipelines in a hosted workspace.
Data Wrangling for ML: Wrangle video, audio, documents, and images into ML-ready datasets with Pixeltable computed columns, iterators, and embedding indices.
Cloud Offering: Share data, serve endpoints, and collaborate on multimodal AI workflows with Pixeltable Cloud services for teams and production deployments.

OpenAPI Specs

openapi