Ai

RagFlow v0.25.5 Accelerates Dataset Search with Major Performance Improvements

RagFlow v0.25.5 Accelerates Dataset Search with Major Performance Improvements

RagFlow v0.25.5, released May 20, 2026, brings a significant performance boost to dataset search operations. The team has stripped out expensive vector fetch and rerank similarity steps, slashing latency by 50 to 100 percent. Administrators also gain two new provider options: local and SSH. This release is a clear signal that RagFlow is doubling down on speed and flexibility for enterprise RAG workloads.

What Changed

The headline update is a reworked dataset search path. Previously, RagFlow would fetch vectors and rerank similarity for every search, a costly process. Now those steps are removed entirely. The result? Search latency drops between 50 and 100 percent, depending on dataset size and query complexity. This change lands via PR #14970.

Another engine-level improvement: metadata filters are now pushed down to Infinity, the underlying search engine. Instead of filtering after retrieval, filtering happens at the storage layer. That cuts another round of overhead. PR #14974 details the tweak.

On the admin side, RagFlow adds two new provider options for connecting infrastructure: local and SSH (PR #15039). Admins can now manage on-premise or remote deployments directly from the control panel, no extra tools required.

Why It Matters

Speed isn't just a nice-to-have in RAG — it's a competitive edge. When you're serving query results to users or feeding an automated pipeline, every millisecond adds up. A 50–100% reduction in search latency can transform user experience. It means instant answers instead of waiting. That's the difference between a product that feels sluggish and one that feels magical.

The metadata filter pushdown is equally critical. If you've ever tried to filter millions of documents by date, category, or source, you know how quickly traditional retrieval bogs down. By pushing filters down to Infinity, RagFlow avoids pulling unnecessary data into memory. The result: faster, more efficient queries, especially for large enterprise datasets.

And the new providers? They reduce friction. No more scripting custom connections for local or SSH infrastructure. It's a small change, but it lowers the barrier for teams that need to plug in diverse environments. RagFlow is clearly listening to what admins actually need.

This isn't a flashy release. It's a practical one. The kind that makes developers’ jobs easier and end users happier. For anyone running RagFlow in production, v0.25.5 is a must-upgrade.

Official Source: https://github.com/infiniflow/ragflow/releases/tag/v0.25.5