Product / UX / Orchestration & agents

OpenRAG

Linagora's lightweight, modular, and extensible retrieval-augmented generation (RAG) framework — a batteries-included, self-hostable application stack rather than a bare library. It ingests multiple document formats (PDFs, documents, audio, and images with AI captioning), organizes knowledge into partitions, performs hybrid semantic + keyword search with multilingual reranking, and exposes indexed content through a bundled web chat interface, an admin UI, and an OpenAI-compatible REST API. Backend in Python with a TypeScript/JavaScript frontend; scales across machines and GPUs via Ray and deploys via Docker Compose. AGPL-3.0, fully open source with no vendor lock-in.

OpenRAG by Linagora, AGPL-3.0. Full application stack (admin UI + chat UI + OpenAI-compatible REST API), Python backend / TS-JS frontend, Ray-based distributed processing, GPU acceleration, Docker Compose deployment. ~237 GitHub stars (Aug 2026). No first-party PyPI/npm package; self-host is the primary distribution. Verified live on GitHub, August 2026.

Openness

5 high confidence

5.0

license: AGPL-3.0(OSI)
source: public
self-host: primary(Docker Compose)
core-gated: ungated
commercial: none

Fully open AGPL-3.0 RAG application stack — the published source is the whole product, self-hosted via Docker Compose, with no feature-gated enterprise tier and no vendor lock-in. AGPL is OSI-approved and sits beside Apache on the software ladder, so osi + public + ungated lands at 5.

https://github.com/linagora/openrag recorded 2026-08-07
AGPL-3.0 license; full source for backend (Python) + frontend (TS/JS); self-hostable via Docker Compose; admin UI, chat UI, and OpenAI-compatible REST API bundled; no paid/enterprise tier

Adoption

1 medium confidence

1.0

~237 GitHub stars, 55 forks (Aug 2026) — early-stage. No download metric available (Docker-distributed, no first-party PyPI/npm), so adoption falls back to stars, which the ladder caps at 3; 237 stars places it in the lowest band.

https://github.com/linagora/openrag recorded 2026-08-07
~237 stars, 55 forks, 88 open issues (as of 2026-08-07)

Capability

3 medium confidence

3.0

Feature-complete, batteries-included general RAG stack with a broad ingestion + retrieval surface and a bundled UI/API. A notch below RAGFlow (capability 4), whose distinctive layout-aware deep document understanding and larger, more mature community set the ceiling for this category.

https://github.com/linagora/openrag recorded 2026-08-07
feature set: multimodal ingestion, hybrid search + multilingual reranking, chat/admin UI, OpenAI-compatible API, Ray distribution, GPU acceleration

Unchanged since 2026-08-07 (last edited, not re-checked)