# pixeltable/pixeltable

Unified multimodal backend for AI data apps

Repository: https://github.com/pixeltable/pixeltable
Canonical: https://ross.abutalabs.com/products/pixeltable
Homepage: https://docs.pixeltable.com
Language: Python
License: Apache-2.0
License Family: permissive
Topics: llm, genai, computer-vision, ai, data-science, database, multimodal, vector-database, machine-learning, ml, feature-engineering, feature-store
Last push: 2026-08-25T07:23:32+00:00

## Health v2 (maintenance only)
Score: 92/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 86, longevity 86
- inputs: {"age_days": 1211, "days_push": 8, "days_rel": 19, "gap_med": 6.0, "n_releases_24m": 97}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1613, forks 219 (observed 2026-08-28T04:05:11.191638+00:00)

## What it is
Pixeltable is a Python library providing declarative, incremental data infrastructure for multimodal AI applications, unifying storage of images, video, audio, documents, and structured data with computed columns, embedding indexes, and model orchestration in one system. It replaces the typical glue of object storage, a vector database, and pipeline orchestration with a single table-based API featuring versioning, transactions, and built-in integrations with LLM and ML providers.

## Use cases
- store and index images video and audio for ai apps
- build rag pipelines over multimodal documents
- replace blob storage plus vector db plus orchestrator with one system
- run embeddings and llm calls automatically on new data
- version ai datasets and computed features
- build computer vision data pipelines in python
- manage feature engineering for machine learning apps
- serve ai app endpoints from stored media and models

## When to choose
- you are building a multimodal AI app and want storage, transformation, indexing, and serving in one declarative Python API
- you want incremental, versioned pipelines that recompute only on new data instead of brittle glue scripts
- you need built-in integrations with OpenAI, Anthropic, Gemini, Hugging Face, and other model providers
- you want to avoid stitching together S3, a vector database, and an orchestrator yourself

## When to avoid
- you need a battle-tested general-purpose database for high-throughput transactional workloads unrelated to AI
- you already have a mature MLOps stack and only need a single component like a standalone vector database
- your project requires a stable 1.0 release or long-term enterprise support guarantees
- you work outside Python or need non-Python client SDKs

## Facets
- artifact type: library
- maturity: active
- function: database, vector-database, rag, etl, machine-learning, data-science, workflow-automation, sdk, plugin-system, caching
- domain: artificial-intelligence, machine-learning, large-language-models, computer-vision, data-science, databases, media
- platform: python, windows, cross-platform
- tags: multimodal, feature-store, declarative-pipelines, computed-columns, data-versioning, embeddings, genai, mlops, media-storage, udf, data-engineering, retrieval-augmented-generation, linux, macos

## Member repositories
- pixeltable/pixeltable (main) score 92

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:11.191638+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:51:07.657719+00:00, confidence not recorded.
  - readme: https://github.com/pixeltable/pixeltable (fetched 2026-08-28T04:05:11.191638+00:00, sha 1d3e0cfcc496)
  - homepage: https://docs.pixeltable.com (fetched 2026-08-29T11:23:03.677962+00:00, sha 6a9067d410ee)
  - site_page: https://docs.pixeltable.com/ (fetched 2026-08-29T11:23:03.687240+00:00, sha 6a9067d410ee)
  - site_page: https://docs.pixeltable.com/integrations/frameworks (fetched 2026-08-29T11:23:03.689590+00:00, sha b7d93133d68b)
  - site_page: https://docs.pixeltable.com/changelog/changelog (fetched 2026-08-29T11:23:03.691487+00:00, sha 7a4b81ab8863)
  - site_page: https://docs.pixeltable.com/overview/quick-start (fetched 2026-08-29T11:23:03.694456+00:00, sha 46a298f3a677)
  - registry_pypi: https://pypi.org/pypi/pixeltable/json (fetched 2026-08-29T11:23:03.696291+00:00, sha 89a12f54f199)
- Data as of 2026-08-30T08:39:29.467469+00:00.