Ecosystem - Pixeltable Documentation

Each AI provider is a function you assign as a computed column on a TableModel. Create the table with pxt schema update. In notebooks you can still call add_computed_column(). Each provider page lists its API key and optional extras. If a provider is missing, wrap it in a @pxt.udf or ask on GitHub Discussions.

import pixeltable as pxt
import pixeltable.functions as pxtf

TableModel = pxt.model_base()


class Articles(TableModel, name='articles'):
    body: pxt.String
    response = pxtf.openai.chat_completions(
        messages=[{'role': 'user', 'content': body}],
        model='gpt-4o-mini',
    )

Cloud LLM providers

Anthropic Claude

Claude models for language and multimodal generation

Google Gemini

Gemini via Google AI Studio or Vertex AI

OpenAI

GPT models for text, embeddings, and image analysis

Azure OpenAI

OpenAI models via Azure

Mistral AI

Mistral language models

DeepSeek

DeepSeek language and code models

Groq

Groq-hosted models

Model hubs

Platforms that host or route many models from a shared catalog.

Hugging Face Hub

CLIP, DETR, sentence-transformers, ViT, Speech2Text, and more

Replicate

Run models on Replicate

Together AI

Open-source models via Together

Fireworks

Fireworks inference

OpenRouter

One API across many LLM providers

AWS Bedrock

Models through Bedrock

Nebius

Nebius Token Factory via an OpenAI-compatible API

Embeddings and reranking

Voyage AI

Embeddings and reranking for text, images, and video

Jina AI

Embeddings and reranking for search

Twelve Labs

Multimodal embeddings via the TwelveLabs Embed API

Media generation

BFL (FLUX)

FLUX image generation and editing from Black Forest Labs

fal.ai

Flux, Stable Diffusion, and other image models

RunwayML

Runway video generation

Local LLM runtimes

Llama.cpp

Local LLMs on CPU and GPU

Ollama

Local open-source models via Ollama

vLLM

High-throughput local inference

Computer vision

YOLOX

YOLOX object detection on images and video frames

Voxel51

Image and video datasets with FiftyOne

Audio

WhisperX

Transcription, speaker diarization, and search

Whisper

Speech recognition with Whisper

Enterprise

Microsoft Fabric

Azure OpenAI through Microsoft Fabric

DataFrames

Pandas

Import and export pandas DataFrames. Do not use pandas as the database that holds your app data.

Agent frameworks

Already on Agno? Keep it. The cookbook stores the agent’s knowledge through the langchain-pixeltable adapter, which creates the table and its index on first insert. Each run drops agno_rag/docs first, so any data already in that table is deleted. A new Pixeltable app declares its embedding index on a TableModel instead (embedding indexes).

Agno

Local RAG: an Agno agent searches a Pixeltable embedding index through langchain-pixeltable

Observability

Langfuse

Trace operations, UDF calls, and LLM generations

Arize Phoenix

Trace operations, UDF calls, and LLM spans in Phoenix

Pydantic Logfire

Trace operations, UDF calls, and OpenAI calls in Logfire

SigNoz

Trace operations, UDF calls, and LLM spans in SigNoz

Opik

Trace operations, UDF calls, and LLM spans in Opik

OpenTelemetry

Export spans and metrics from the OpenTelemetry bridge

Provider notebooks: docs/release/howto/providers.

​Cloud LLM providers