b10178

ggml-org/llama.cppb10178Jul 29, 2026by github-actions[bot]same day as b10176

What changed since b10176

low risk

b10178 adds trace-level logging for server prompt-cache slot similarity checks, on top of b10176's RPC tensor_memset command.

New features

  • Server now emits trace logs detailing slot similarity evaluation during prompt cache slot selection, including why candidates were skipped and how similarity was computed.

Improvements

  • Improved observability of prompt cache slot selection, making it easier to diagnose why a particular slot was or wasn't reused.

Upgrade notes

No configuration or API changes are required; the new diagnostics only appear when trace-level logging is enabled. Note that b10176's RPC tensor_memset command is also included in this build if you are jumping straight from an earlier revision.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Enhanced server logging by adding trace logging for slot similarity checking during prompt cache slot selection, including details on skip reasons and calculations.

Key Highlights

  • Added trace logging for slot similarity checking
  • Includes skip reasons and similarity calculation details

Release Notes (excerpt)

<details open> server : add trace logging for slot similarity checking (#26271) Adds trace logging in server-context.cpp for slot similarity checking during prompt cache slot selection, including skip reasons and similarity calculation details. Assisted-by: llama.cpp:Qwen3.6-27B </details> **Website:** - <https://llama.app> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **Linux:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories