TL;DR

Ragflow now supports local and SSH data providers, significantly accelerating dataset searches and metadata filtering, improving overall performance and usability.

New

• Data Providers: Added local and SSH providers for connecting to data sources.
• Search Acceleration: Dataset search latency is reduced by 50-100% through optimized vector search and metadata filtering.
• Text-to-Speech (TTS): Added Redis caching and support for multiple TTS providers (MiniMax, Fishaudio, Openrouter).

Fixes Worth Knowing

• Resolved issues with file uploads/downloads in the agent, ensuring attachments are preserved.
• Corrected token usage reporting to Langfuse (observability platform).
• Fixed a bug where document counters were not updating atomically.
• Improved robustness of the reranking module.

Before You Upgrade

Ensure your Redis server is accessible if you plan to use the new TTS caching feature.

TL;DR

Ragflow now supports LangChain (a framework for building LLM applications) expression language, enabling more complex and flexible retrieval-augmented generation pipelines.

New

• LangChain Expression Language Support: Build sophisticated RAG pipelines using LangChain’s declarative syntax.

Fixes Worth Knowing

• Improved error messaging for invalid data sources.
• Resolved an issue where certain document loaders failed silently.

Before You Upgrade

• Update your LangChain version to at least 2.10 to take full advantage of the new expression language support.