n8n 2.48.3 fixes Pinecone search, Redis cache keys and expression speed

n8n published version 2.48.3 on November 13, 2026. The patch release contains bug fixes, a small set of feature changes and two performance improvements, according to the release notes.

On the core side, trusted-source cache keys now stay in one Redis cluster slot. AI agents can also turn off prompt caching for endpoints that reject the cache fields, and legacy-format data is read with the instance key when the stored key differs. The expression runtime does less setup work per item, and the replaceSpecialChars character map has moved out of the runtime bundle.

Other changes affect the Pinecone Vector Store node, which can now search indexes with a document schema through the documents API. In the editor, the JSON highlight color is visible on the light theme, and Gateway credits promotions appear based on the Gateway config.

Why it matters

Patch releases like this one often fix problems that appear only in production, such as cache handling with clustered Redis or searches against document-schema indexes. Teams that run agents against endpoints that reject prompt caching fields now have a way to turn that caching off. The expression runtime changes may help workflows that process many items, but the release notes do not quantify the gain.

What to do

If you run n8n with Redis cluster caching, Pinecone nodes or AI agents, test those workflows on 2.48.3 before upgrading production. Check the vendor’s site for upgrade steps and compatibility notes, because this summary does not cover them.