Daft

What it does

AI operations

Built-in AI functions include prompts, text and image embeddings, and zero-shot text and image classification.checked 7 Oct 2026

AI providers

Documented AI providers include OpenAI, Hugging Face Transformers, Google, and LM Studio.checked 7 Oct 2026

Dashboard security limit

The Daft Dashboard has no built-in authentication, and its documentation advises against exposing it directly to the public internet.checked 7 Oct 2026

Data types

Daft processes images, audio, video, text, and structured data through a unified Python API.checked 7 Oct 2026

Execution

Daft runs locally by default and can scale across multiple machines using Ray or Kubernetes.checked 7 Oct 2026

Installation requirement

Daft requires Python 3.10 or higher and can be installed with pip install -U daft.checked 7 Oct 2026

Intended users

Eventual describes Daft as an open-source engine for engineers processing petabytes of multimodal data.checked 7 Oct 2026

Purpose

Daft is a high-performance data engine for processing multimodal and structured data in AI workloads.checked 7 Oct 2026

Security and privacy

Daft's usage telemetry collects non-identifiable system metadata, not proprietary code or data, and can be disabled with DO_NOT_TRACK=true.checked 7 Oct 2026

Storage integrations

Documented object storage connectors include AWS S3, Azure Blob Storage, Google Cloud Storage, Tencent Cloud COS, and GooseFS.checked 7 Oct 2026

Support

The documentation directs users to the Distributed Data Slack Community and the #daft-dev channel for contributor guidance.checked 7 Oct 2026

Table and catalog integrations

Daft documents support for table formats including Hudi, Iceberg, and Delta Lake, and catalogs including AWS Glue and Unity Catalog.checked 7 Oct 2026