fix: keep loopback and LAN traffic off an ambient proxy

Pre-existing on main, not introduced by the driver work, but it makes a local
RAG backend unusable so it belongs with this change.

build_client only called set_proxy when a client had configured one of its own.
With nothing configured, reqwest's own detection applied, which sends every
request through a *_PROXY variable including ones bound for 127.0.0.1 or a LAN
address. A proxy cannot usefully forward those, and anything that intercepts
proxied traffic answers on behalf of a service that is running perfectly well,
so the error names the proxy rather than the store and reads as a Coyote fault.

Concretely, an installed Socket Firewall exports HTTP_PROXY to the processes it
wraps and rejects hosts outside its allow list. That turned a healthy Ollama on
the LAN into 'error decoding response body: expected value at line 2 column 1' —
its HTML refusal page parsed as JSON — and a loopback Qdrant into an HTTP 405.

Proxy handling is now always applied and always exempts loopback and private
ranges, with NO_PROXY merged in since replacing reqwest's detection also
replaces its handling of that variable. HTTP_PROXY and HTTPS_PROXY are kept
separate because they are allowed to differ. An explicitly configured proxy
still wins, and '-' still means none.

This also supersedes the unconditional no_proxy() added to the Qdrant client in
af9622d: that made it the only client to ignore a proxy outright, on a
justification I got wrong. It now shares this path, so a remote store behind a
real proxy keeps working.
This commit is contained in:
2026-08-12 10:33:03 -06:00
parent 4dd6e794b2
commit 54685be9a2
3 changed files with 115 additions and 12 deletions
+1 -3
View File
@@ -56,9 +56,7 @@ pub trait Client: Sync + Send {
let mut builder = ReqwestClient::builder(); let mut builder = ReqwestClient::builder();
let extra = self.extra_config(); let extra = self.extra_config();
let timeout = extra.and_then(|v| v.connect_timeout).unwrap_or(10); let timeout = extra.and_then(|v| v.connect_timeout).unwrap_or(10);
if let Some(proxy) = extra.and_then(|v| v.proxy.as_deref()) { builder = apply_proxy(builder, extra.and_then(|v| v.proxy.as_deref()))?;
builder = set_proxy(builder, proxy)?;
}
if let Some(user_agent) = self.app_config().user_agent.as_ref() { if let Some(user_agent) = self.app_config().user_agent.as_ref() {
builder = builder.user_agent(user_agent); builder = builder.user_agent(user_agent);
} }
+2 -3
View File
@@ -1,5 +1,6 @@
use crate::rag::provider::RagProvider; use crate::rag::provider::RagProvider;
use crate::rag::{DocumentId, RagData}; use crate::rag::{DocumentId, RagData};
use crate::utils::apply_proxy;
use anyhow::{Context, Result, bail}; use anyhow::{Context, Result, bail};
use async_trait::async_trait; use async_trait::async_trait;
@@ -191,9 +192,7 @@ impl QdrantProvider {
value.set_sensitive(true); value.set_sensitive(true);
headers.insert("api-key", value); headers.insert("api-key", value);
} }
Client::builder() apply_proxy(Client::builder().default_headers(headers), None)?
.default_headers(headers)
.no_proxy()
.build() .build()
.context("Failed to build reqwest client") .context("Failed to build reqwest client")
} }
+112 -6
View File
@@ -295,18 +295,83 @@ pub fn is_url(path: &str) -> bool {
path.starts_with("http://") || path.starts_with("https://") path.starts_with("http://") || path.starts_with("https://")
} }
pub fn set_proxy( /// 127.0.0.1 means something different to a proxy than it does to us, so a tool
/// that intercepts proxied traffic answers for a local service that is running
/// fine, and the failure reads as a fault in Coyote or in that service.
const LOCAL_NO_PROXY: &str =
"localhost,127.0.0.0/8,::1,10.0.0.0/8,172.16.0.0/12,192.168.0.0/16,.local";
/// Applies proxy settings, keeping local traffic direct. `configured` is a
/// client's own `extra.proxy`, where `"-"` means no proxy at all; with nothing
/// configured an ambient `*_PROXY` is still honoured for public hosts.
pub fn apply_proxy(
mut builder: reqwest::ClientBuilder, mut builder: reqwest::ClientBuilder,
proxy: &str, configured: Option<&str>,
) -> Result<reqwest::ClientBuilder> { ) -> Result<reqwest::ClientBuilder> {
// reqwest offers no way to add rules to the proxies it auto-detects, so
// detection is disabled and redone below.
builder = builder.no_proxy(); builder = builder.no_proxy();
if !proxy.is_empty() && proxy != "-" {
builder = builder let configured = configured.map(str::trim).filter(|p| !p.is_empty());
.proxy(reqwest::Proxy::all(proxy).with_context(|| format!("Invalid proxy `{proxy}`"))?); if configured == Some("-") {
}; return Ok(builder);
}
let exempt = no_proxy_rules();
if let Some(url) = configured {
let proxy = reqwest::Proxy::all(url)
.with_context(|| format!("Invalid proxy `{url}`"))?
.no_proxy(reqwest::NoProxy::from_string(&exempt));
return Ok(builder.proxy(proxy));
}
// Split per scheme, because HTTP_PROXY and HTTPS_PROXY are allowed to differ.
for (is_https, keys) in [
(
true,
["HTTPS_PROXY", "https_proxy", "ALL_PROXY", "all_proxy"],
),
(
false,
["HTTP_PROXY", "http_proxy", "ALL_PROXY", "all_proxy"],
),
] {
let Some(url) = first_env(&keys) else {
continue;
};
let proxy = if is_https {
reqwest::Proxy::https(&url)
} else {
reqwest::Proxy::http(&url)
};
let proxy = proxy
.with_context(|| format!("Invalid proxy `{url}`"))?
.no_proxy(reqwest::NoProxy::from_string(&exempt));
builder = builder.proxy(proxy);
}
Ok(builder) Ok(builder)
} }
fn first_env(keys: &[&str]) -> Option<String> {
keys.iter()
.find_map(|key| env::var(key).ok())
.map(|value| value.trim().to_string())
.filter(|value| !value.is_empty())
}
/// Replacing reqwest's auto-detection loses its `NO_PROXY` handling, so that is
/// merged back in here.
fn no_proxy_rules() -> String {
let mut rules = LOCAL_NO_PROXY.to_string();
let extra = env::var("NO_PROXY")
.or_else(|_| env::var("no_proxy"))
.unwrap_or_default();
if !extra.trim().is_empty() {
rules.push(',');
rules.push_str(extra.trim());
}
rules
}
pub fn decode_bin<T: serde::de::DeserializeOwned>(data: &[u8]) -> Result<T> { pub fn decode_bin<T: serde::de::DeserializeOwned>(data: &[u8]) -> Result<T> {
let (v, _) = bincode::serde::decode_from_slice(data, bincode::config::legacy())?; let (v, _) = bincode::serde::decode_from_slice(data, bincode::config::legacy())?;
Ok(v) Ok(v)
@@ -316,6 +381,47 @@ pub fn decode_bin<T: serde::de::DeserializeOwned>(data: &[u8]) -> Result<T> {
mod tests { mod tests {
use super::*; use super::*;
/// Invalid rule syntax makes `from_string` return `None`, which silently drops
/// every exemption and sends local traffic back through the proxy.
#[test]
fn the_local_no_proxy_rules_are_valid() {
assert!(
reqwest::NoProxy::from_string(LOCAL_NO_PROXY).is_some(),
"reqwest rejected LOCAL_NO_PROXY, so nothing would be exempt"
);
}
#[test]
fn local_rules_cover_loopback_and_private_ranges() {
for host in [
"localhost",
"127.0.0.0/8",
"10.0.0.0/8",
"172.16.0.0/12",
"192.168.0.0/16",
] {
assert!(
LOCAL_NO_PROXY.contains(host),
"{host} must stay exempt from proxying"
);
}
}
#[test]
fn a_dash_means_no_proxy_at_all() {
assert!(apply_proxy(reqwest::ClientBuilder::new(), Some("-")).is_ok());
assert!(apply_proxy(reqwest::ClientBuilder::new(), Some(" - ")).is_ok());
}
#[test]
fn an_unparseable_proxy_is_reported() {
let err = apply_proxy(reqwest::ClientBuilder::new(), Some("not a url"))
.unwrap_err()
.to_string();
assert!(err.contains("Invalid proxy"), "got: {err}");
}
#[test] #[test]
#[cfg(not(target_os = "windows"))] #[cfg(not(target_os = "windows"))]
fn test_safe_join_path() { fn test_safe_join_path() {