From 308e726e11887c504569e48ac93f43bfdd6b8368 Mon Sep 17 00:00:00 2001
From: mamei16 <marcel.1710@live.de>
Date: Mon, 13 Oct 2025 04:00:11 +0200
Subject: [PATCH] log error when llama-server request exceeds context size
 (#7263)

---
 modules/llama_cpp_server.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/modules/llama_cpp_server.py b/modules/llama_cpp_server.py
index 38589cf2..27890d8c 100644
--- a/modules/llama_cpp_server.py
+++ b/modules/llama_cpp_server.py
@@ -200,7 +200,10 @@ class LlamaServer:
         # Make the generation request
         response = self.session.post(url, json=payload, stream=True)
         try:
-            response.raise_for_status()  # Raise an exception for HTTP errors
+            if response.status_code == 400 and response.json()["error"]["type"] == "exceed_context_size_error":
+                logger.error("The request exceeds the available context size, try increasing it")
+            else:
+                response.raise_for_status()  # Raise an exception for HTTP errors
 
             full_text = ""