[marecabo]

Overview

This addresses the two review comments of #21752:

• Adds honoring of NO_PROXY/no_proxy env vars and update readme
  • required since 2026-07-08 to ensure CLI loopback #24948 is not sent to proxy
• Adds logging for malformed proxy URLs

Additional information

On top of this branch proxy-env-vars-addons, I merged master back in branch proxy-env-vars-merged (cc576d1), see ggml-org/llama.cpp@master...marecabo:llama.cpp:proxy-env-vars-merged. There was just one small readme conflict.

I tested this merged branch locally on RHEL 9 with ZScaler successfully:

$ export proxy=http://... # your proxy here

$ http_proxy= https_proxy= no_proxy= llama download -hf ggml-org/Qwen3.5-0.8B-GGUF
0.00.039.245 E get_repo_commit: error: HTTPLIB failed: Could not establish connection
error: model download failed

$ http_proxy= https_proxy=nonsense no_proxy= llama serve -hf ggml-org/Qwen3.5-0.8B-GGUF
0.00.036.272 W common_http_client: ignoring malformed proxy URL from the environment: invalid URL: no scheme
0.00.044.224 E get_repo_commit: error: HTTPLIB failed: Could not establish connection
0.00.044.227 W get_repo_files: failed to resolve commit for ggml-org/Qwen3.5-0.8B-GGUF
...

$ http_proxy=$proxy https_proxy=$proxy no_proxy= llama download -hf ggml-org/Qwen3.5-0.8B-GGUF
Downloading Qwen3.5-0.8B-Q8_0.gguf ───────────────────────────────── 100%
/home/***/.cache/huggingface/hub/models--ggml-org--Qwen3.5-0.8B-GGUF/snapshots/8fea620810c4afa23dd6443f999a48574c1611a3/Qwen3.5-0.8B-Q8_0.gguf

$ http_proxy=http://127.0.0.1:1 https_proxy=$proxy no_proxy= llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF
Loading model...  # <-- hangs -> CTRL+C

$ http_proxy=http://127.0.0.1:1 https_proxy=$proxy no_proxy=127.0.0.1 llama cli -hf ggml-org/Qwen3.5-0.8B-GGUF
Loading model...  

▄▄ ▄▄
██ ██
██ ██  ▀▀█▄ ███▄███▄  ▀▀█▄    ▄████ ████▄ ████▄
██ ██ ▄█▀██ ██ ██ ██ ▄█▀██    ██    ██ ██ ██ ██
██ ██ ▀█▄██ ██ ██ ██ ▀█▄██ ██ ▀████ ████▀ ████▀
                                    ██    ██
                                    ▀▀    ▀▀

build      : b10824-cc576d166
model      : ggml-org/Qwen3.5-0.8B-GGUF
ftype      : Q8_0
modalities : text
...

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES, I used Claude Opus 5

[texasich]

LGTM, thanks