Skip to content
AIBRAINWORX / OpenSourceCoder.comLOCAL-FIRST. BUILT IN THE OPEN.
Open Source

RAGFlow 1.0 rewrites its stack in Go and changes the upgrade path

The document and retrieval project adds agentic chat; its maintainers warn the 0.27.2 migration is irreversible.

Conceptual document search workstation
AI-generated conceptual editorial image; not a photograph of the product, company or event. · infiniflow public GitHub organization avatar; editorial identification only.
Event / context: RAGFlow v1.0.0 release: November 6, 2026.. Source check: Nov 9, 2026.

The RAGFlow project released version 1.0.0 on November 6, moving to a Go-based implementation and introducing an agentic chat mode for document-grounded questions. The project is available in its public Apache-2.0 repository.

What changed

The release notes describe iterative retrieval, tool calls and an evidence-sufficiency decision in the new chat flow. They also describe unified document parsers, NATS replacing Redis for the message queue, Kvrocks for cache and checkpoints, ARM64 Docker images and a rewritten CLI. Existing installations should check compatibility and the removed API aliases before upgrading.

The maintainers claim substantially smaller images, faster startup and higher concurrency. Those are project claims; workload details matter, and OpenSourceCoder.com has not reproduced the comparisons. An agent’s decision that evidence is sufficient is not independent proof that an answer is true.

Back up before changing versions

The release notes say data can migrate automatically from 0.27.2, but that migration is irreversible. They instruct operators to back up all data first and restore from the backup if migration fails. Older installations need the intermediate steps in the project’s migration guide; v1.0 release candidates have a different compatibility path.

In our view, the practical improvement to watch is not just a faster container. It is whether teams can trace answers back to reliable source documents, manage data safely and understand when retrieval has failed. Those checks remain necessary even when the orchestration is more sophisticated.

EXPLORE THE SOURCE

GitHub repositories

Original reporting and analysis based on the linked primary sources. We did not independently test the products or benchmark the models.

Follow your curiosity.

Pass the signal.