From 968c43891abb1def4db76f25713b558629fb92b1 Mon Sep 17 00:00:00 2001 From: Xuan-Son Nguyen Date: Thu, 18 Jun 2026 12:15:46 +0200 Subject: [PATCH] server: fix router args not being forwarded to child instances (#24760) --- tools/server/server-models.cpp | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/tools/server/server-models.cpp b/tools/server/server-models.cpp index ff9a0df12f..cc7184165d 100644 --- a/tools/server/server-models.cpp +++ b/tools/server/server-models.cpp @@ -351,6 +351,12 @@ void server_models::load_models() { source_map[name] = SERVER_MODEL_SOURCE_PRESET; } + // overlay router's own CLI args on top of every model preset so that + // e.g. `llama-server --temp 0` is honoured by all child processes + for (auto & [name, preset] : final_presets) { + preset.merge(base_preset); + } + auto get_source = [&](const std::string & name) { return source_map.count(name) ? source_map.at(name) : SERVER_MODEL_SOURCE_PRESET; };