From ed42154c78d7d7a63c092684deaa47d22750d796 Mon Sep 17 00:00:00 2001
From: oobabooga <112222186+oobabooga@users.noreply.github.com>
Date: Sat, 19 Apr 2025 05:32:36 -0700
Subject: [PATCH] Revert "llama.cpp: close the connection immediately on
 'Stop'"

This reverts commit 5fdebc554b7ca46afb9695babf89397635e9f91d.
---
 modules/llama_cpp_server.py | 14 +++-----------
 1 file changed, 3 insertions(+), 11 deletions(-)

diff --git a/modules/llama_cpp_server.py b/modules/llama_cpp_server.py
index 3025aa7d..5071c40c 100644
--- a/modules/llama_cpp_server.py
+++ b/modules/llama_cpp_server.py
@@ -141,24 +141,16 @@ class LlamaServer:
             print()
 
         # Make a direct request with streaming enabled using a context manager
-        with self.session.post(url, json=payload, stream=True, timeout=(5, 0.1)) as response:
+        with self.session.post(url, json=payload, stream=True) as response:
             response.raise_for_status()  # Raise an exception for HTTP errors
 
             full_text = ""
-            iterator = response.iter_lines(decode_unicode=True)
 
-            while True:
+            # Process the streaming response
+            for line in response.iter_lines(decode_unicode=True):
                 if shared.stop_everything:
                     break
 
-                try:
-                    line = next(iterator)
-                except requests.exceptions.Timeout:
-                    # Check stop flag again on timeout
-                    continue
-                except StopIteration:
-                    break
-
                 if line:
                     try:
                         # Check if the line starts with "data: " and remove it