diff --git a/rust/crates/api/tests/openai_compat_integration.rs b/rust/crates/api/tests/openai_compat_integration.rs
index 0cbe8732..fa3789fd 100644
--- a/rust/crates/api/tests/openai_compat_integration.rs
+++ b/rust/crates/api/tests/openai_compat_integration.rs
@@ -2,6 +2,7 @@ use std::collections::HashMap;
 use std::ffi::OsString;
 use std::sync::Arc;
 use std::sync::{Mutex as StdMutex, OnceLock};
+use std::time::Duration;
 
 use api::{
     ApiError, ContentBlockDelta, ContentBlockDeltaEvent, ContentBlockStartEvent,
@@ -66,6 +67,56 @@ async fn send_message_uses_openai_compatible_endpoint_and_auth() {
     assert_eq!(body["tools"][0]["type"], json!("function"));
 }
 
+#[tokio::test]
+async fn send_message_passes_optional_openai_compatible_parameters_on_wire() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let body = concat!(
+        "{",
+        "\"id\":\"chatcmpl_params\",",
+        "\"model\":\"gpt-4o\",",
+        "\"choices\":[{",
+        "\"message\":{\"role\":\"assistant\",\"content\":\"Parameters preserved\",\"tool_calls\":[]},",
+        "\"finish_reason\":\"stop\"",
+        "}],",
+        "\"usage\":{\"prompt_tokens\":3,\"completion_tokens\":2}",
+        "}"
+    );
+    let server = spawn_server(
+        state.clone(),
+        vec![http_response("200 OK", "application/json", body)],
+    )
+    .await;
+
+    let client = OpenAiCompatClient::new("openai-test-key", OpenAiCompatConfig::openai())
+        .with_base_url(server.base_url());
+    let response = client
+        .send_message(&MessageRequest {
+            model: "gpt-4o".to_string(),
+            temperature: Some(0.2),
+            top_p: Some(0.8),
+            frequency_penalty: Some(0.15),
+            presence_penalty: Some(0.25),
+            stop: Some(vec!["END".to_string()]),
+            reasoning_effort: Some("low".to_string()),
+            ..sample_request(false)
+        })
+        .await
+        .expect("request should succeed");
+
+    assert_eq!(response.total_tokens(), 5);
+
+    let captured = state.lock().await;
+    let request = captured.first().expect("server should capture request");
+    let body: serde_json::Value = serde_json::from_str(&request.body).expect("json body");
+    assert_eq!(body["model"], json!("gpt-4o"));
+    assert_eq!(body["temperature"], json!(0.2));
+    assert_eq!(body["top_p"], json!(0.8));
+    assert_eq!(body["frequency_penalty"], json!(0.15));
+    assert_eq!(body["presence_penalty"], json!(0.25));
+    assert_eq!(body["stop"], json!(["END"]));
+    assert_eq!(body["reasoning_effort"], json!("low"));
+}
+
 #[tokio::test]
 async fn send_message_preserves_deepseek_reasoning_content_before_text() {
     let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
@@ -280,6 +331,65 @@ async fn stream_message_normalizes_text_and_multiple_tool_calls() {
     assert!(request.body.contains("\"stream\":true"));
 }
 
+#[allow(clippy::await_holding_lock)]
+#[tokio::test]
+async fn stream_message_retries_retryable_sse_handshake_failures() {
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let sse = concat!(
+        "data: {\"id\":\"chatcmpl_stream_retry\",\"model\":\"gpt-4o\",\"choices\":[{\"delta\":{\"content\":\"Recovered\"}}]}\n\n",
+        "data: {\"id\":\"chatcmpl_stream_retry\",\"choices\":[{\"delta\":{},\"finish_reason\":\"stop\"}]}\n\n",
+        "data: [DONE]\n\n"
+    );
+    let server = spawn_server(
+        state.clone(),
+        vec![
+            http_response(
+                "500 Internal Server Error",
+                "application/json",
+                "{\"error\":{\"message\":\"try again\",\"type\":\"server_error\",\"code\":500}}",
+            ),
+            http_response_with_headers(
+                "200 OK",
+                "text/event-stream",
+                sse,
+                &[("x-request-id", "req_stream_retry")],
+            ),
+        ],
+    )
+    .await;
+
+    let client = OpenAiCompatClient::new("openai-test-key", OpenAiCompatConfig::openai())
+        .with_base_url(server.base_url())
+        .with_retry_policy(1, Duration::ZERO, Duration::ZERO);
+    let mut stream = client
+        .stream_message(&MessageRequest {
+            model: "gpt-4o".to_string(),
+            ..sample_request(false)
+        })
+        .await
+        .expect("stream should retry once then start");
+
+    assert_eq!(stream.request_id(), Some("req_stream_retry"));
+    let mut events = Vec::new();
+    while let Some(event) = stream.next_event().await.expect("event should parse") {
+        events.push(event);
+    }
+    assert!(events.iter().any(|event| matches!(
+        event,
+        StreamEvent::ContentBlockDelta(ContentBlockDeltaEvent {
+            delta: ContentBlockDelta::TextDelta { text },
+            ..
+        }) if text == "Recovered"
+    )));
+
+    let captured = state.lock().await;
+    assert_eq!(captured.len(), 2, "one original request plus one retry");
+    for request in captured.iter() {
+        let body: serde_json::Value = serde_json::from_str(&request.body).expect("json body");
+        assert_eq!(body["stream"], json!(true));
+    }
+}
+
 #[allow(clippy::await_holding_lock)]
 #[tokio::test]
 async fn openai_streaming_requests_opt_into_usage_chunks() {
@@ -358,6 +468,47 @@ async fn openai_streaming_requests_opt_into_usage_chunks() {
     assert_eq!(body["stream_options"], json!({"include_usage": true}));
 }
 
+#[allow(clippy::await_holding_lock)]
+#[tokio::test]
+async fn openai_compatible_client_honors_http_proxy_for_requests() {
+    let _lock = env_lock();
+    let state = Arc::new(Mutex::new(Vec::<CapturedRequest>::new()));
+    let proxy = spawn_server(
+        state.clone(),
+        vec![http_response(
+            "200 OK",
+            "application/json",
+            "{\"id\":\"chatcmpl_proxy\",\"model\":\"gpt-4o\",\"choices\":[{\"message\":{\"role\":\"assistant\",\"content\":\"Via proxy\",\"tool_calls\":[]},\"finish_reason\":\"stop\"}],\"usage\":{\"prompt_tokens\":4,\"completion_tokens\":3}}",
+        )],
+    )
+    .await;
+    let _http_proxy = ScopedEnvVar::set("HTTP_PROXY", proxy.base_url());
+    let _https_proxy = ScopedEnvVar::unset("HTTPS_PROXY");
+    let _no_proxy = ScopedEnvVar::unset("NO_PROXY");
+    let _http_proxy_lower = ScopedEnvVar::unset("http_proxy");
+    let _https_proxy_lower = ScopedEnvVar::unset("https_proxy");
+    let _no_proxy_lower = ScopedEnvVar::unset("no_proxy");
+
+    let client = OpenAiCompatClient::new("openai-test-key", OpenAiCompatConfig::openai())
+        .with_base_url("http://origin.invalid/v1");
+    let response = client
+        .send_message(&MessageRequest {
+            model: "gpt-4o".to_string(),
+            ..sample_request(false)
+        })
+        .await
+        .expect("proxy should return the OpenAI-compatible response");
+
+    assert_eq!(response.total_tokens(), 7);
+    let captured = state.lock().await;
+    let request = captured.first().expect("proxy should capture request");
+    assert_eq!(request.path, "http://origin.invalid/v1/chat/completions");
+    assert_eq!(
+        request.headers.get("authorization").map(String::as_str),
+        Some("Bearer openai-test-key")
+    );
+}
+
 #[allow(clippy::await_holding_lock)]
 #[tokio::test]
 async fn provider_client_dispatches_xai_requests_from_env() {
@@ -568,6 +719,12 @@ impl ScopedEnvVar {
         std::env::set_var(key, value);
         Self { key, previous }
     }
+
+    fn unset(key: &'static str) -> Self {
+        let previous = std::env::var_os(key);
+        std::env::remove_var(key);
+        Self { key, previous }
+    }
 }
 
 impl Drop for ScopedEnvVar {