AI Potluck
Product / UX / Orchestration & agents

SyftHub

OpenMined

Federated RAG framework from OpenMined. Rather than centralizing documents, a query is broadcast across a network of peers; each peer searches its own private index and returns only the most relevant snippets, which are aggregated, ranked and synthesized into a grounded answer. Its SDK loads data sources from the network, selects a model and runs the pipeline, built on OpenMined's Syft privacy stack.

Early-stage, created August 2025, and part of OpenMined's wider federated-RAG work. Verified 2026-08-13 via the OpenMined/syft-hub-sdk repository and the SyftHub project page.

Openness

5 high confidence
5.0
license
Apache-2.0(OSI)
source
public(github.com/OpenMined/syft-hub-sdk)
self-host
yes
service
none

Fully open source: Apache-2.0 SDK, self-hostable, no proprietary tier. Built on OpenMined's open Syft privacy stack.

  • https://github.com/OpenMined/syft-hub-sdk recorded 2026-08-13

    Apache-2.0 license on the repo, full SDK source published, README "Build Federated RAG and tap into distributed data sources without centralizing"; no paid tier or hosted service; 2 stars

Adoption

1 low confidence
1.0

2 stars across the one declared repo, which bands at <1K stars, level 1 on the stars scale - pre-adoption. Part of OpenMined's federated-RAG effort alongside local-rag and local-rag-router, and included as an emerging entry rather than on adoption merit.

Capability

2 low confidence
2.0

Implements a full federated RAG pipeline — distributed private indexing, federated query broadcast, local top-k retrieval, global aggregation/ranking, and LLM synthesis. A distinctive privacy-preserving approach (nothing else on the map does federated RAG), but early-stage and narrow.

  • https://openmined.org/syfthub/ recorded 2026-08-13

    federated RAG architecture - distributed indexing, federated query broadcast, aggregation and ranking, LLM synthesis; no benchmark figure

Verified 2026-08-13