Daft
Multimodal data engine built for AI
Daft is an open-source multimodal data engine for AI workloads, built to take raw data through to training-ready datasets in a single pipeline. It processes video, images, audio, text, and sensor data alongside structured metadata in one dataframe, addressing the need to handle mixed-modality data at scale without separate tooling.
Daft offers a Python dataframe API with familiar operations such as filter, transform, aggregate, and write, plus SQL support. It runs CPU and GPU work in one pipeline, with scheduling and batching handled automatically. Native model operators cover embeddings, LLM extraction, and structured outputs, with support for models from OpenAI, Hugging Face, or custom providers. Multimodal column types, a managed UDF runtime with automatic batching and retries, and zero-copy execution powered by Apache Arrow are included. The core is written in Rust, and pipelines defined once can run locally or across distributed clusters. Integrations include Apache Hudi, Apache Iceberg, Delta Lake, AWS, Azure, Google Cloud, Unity Catalog, Ray, Pandas, and PyTorch.
Daft is released under the Apache 2.0 license and is installed via pip. It is suited to teams building AI search, data enrichment, and multimodal ETL pipelines. The pages do not specify a free plan, trial, or edition structure.
12 alternatives to Daft
Ranked by how well each tool replaces Daft: shared features, audience, price and popularity.
- 58 out of 100 match$30/mo
- 58 out of 100 match$39/mo
Time series database platform (InfluxDB) for real-time and operational data.
Covers 3 of 15 key features.
Free planOpen source58 out of 100 match$1,500/moA schema-less document database that creates and optimizes its own indexes, with AI and RQ
Covers 3 of 15 key features.
Free planOpen source57 out of 100 matchFreeInfrastructure for continual learning; the company behind Deeplake, a GPU database for...
Covers 7 of 15 key features.
Open source57 out of 100 matchContact salesMultimodal lakehouse for AI — unify data curation, feature engineering, retrieval, and.
Covers 0 of 15 key features.
Open source57 out of 100 matchContact salesDistributed SQL database for AI agents, transactions, and analytics
Covers 4 of 15 key features.
Free planOpen source57 out of 100 match$1,376/mo- 57 out of 100 matchUsage-based
- 57 out of 100 matchFree
Open source, native analytic database for open data and table formats.
Covers 1 of 15 key features.
Open source57 out of 100 match—