From d7387b09b70c1f80453d590c99d0d0f6662cc69e Mon Sep 17 00:00:00 2001
From: "Rodrigo Rodriguez (Pragmatismo)" <me@rodrigorodriguez.com>
Date: Wed, 12 Nov 2025 17:22:33 -0300
Subject: [PATCH] feat(llm): add n_ctx_size parameter to LLM server startup

Added support for configuring the context window size (n_ctx_size) when starting the local LLM server. The parameter is read from config with a default value of 4096 if not specified. This allows for better control over the model's memory usage and performance characteristics.
---
 src/llm/local.rs | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/src/llm/local.rs b/src/llm/local.rs
index 75ea41ea2..4c13d78d2 100644
--- a/src/llm/local.rs
+++ b/src/llm/local.rs
@@ -228,7 +228,13 @@ pub async fn start_llm_server(
     let n_predict = config_manager
         .get_config(&default_bot_id, "llm-server-n-predict", None)
         .unwrap_or("50".to_string());
-    let mut args = format!(
+
+        let n_ctx_size = config_manager
+        .get_config(&default_bot_id, "llm-server-n_ctx_size", None)
+        .unwrap_or("4096".to_string());
+
+
+        let mut args = format!(
         "-m {} --host 0.0.0.0 --port {} --top_p 0.95 --temp 0.6 --repeat-penalty 1.2 --n-gpu-layers {}",
         model_path, port,  gpu_layers
     );
@@ -253,6 +259,8 @@ pub async fn start_llm_server(
     if n_predict != "0" {
         args.push_str(&format!(" --n-predict {}", n_predict));
     }
+        args.push_str(&format!(" --n-ctx-size {}", n_ctx_size));
+    
     if cfg!(windows) {
         let mut cmd = tokio::process::Command::new("cmd");
         cmd.arg("/C").arg(format!(