github infiniflow/ragflow v1.0.0-rc1

6 hours ago

RAGFlow v1.0.0-rc1 is a preview release of v1.0.0. This release marks a comprehensive rewrite of RAGFlow in Go, delivering a substantial reduction in system resource consumption and replacing the previous heavyweight deployment model with a lightweight architecture.

Architecture changes

  • Message queue: Migrated from Redis to NATS.
  • Cache and checkpoints: Kvrocks now serves as the backing store for both cache and checkpoints.

Data upgrade

Upon container startup, RAGFlow automatically initiates the data upgrade process, migrating existing data from v0.27.2 to v1.0.0-rc1. Note that this upgrade is irreversible — once completed, data cannot be rolled back to v0.27.2.

Upgrade prerequisites and precautions

  1. Back up all data prior to upgrading.
  2. Ensure your existing deployment is already on v0.27.2 before proceeding.
  3. Should the upgrade fail, restore your system from the backup and submit an issue detailing the failure.

Compatibility

  1. RAGFlow CLI: Reimplemented in Go, replacing the previous Python version.
  2. API compatibility: Deprecated APIs are no longer supported by the Go implementation.
  3. SDK compatibility: The Python SDK remains unaffected.

Known issues

  1. DeepDoc currently supports CPU environments only.
  2. The local sandbox is no longer supported.
  3. The Team/Me permission model is not yet supported.

What's Changed

Don't miss a new ragflow release

NewReleases is sending notifications on new releases.