From 8022f48d1a6d69c718598f08977ad2dddd41da69 Mon Sep 17 00:00:00 2001
From: JOBOYA <lena.boya26@gmail.com>
Date: Tue, 12 May 2026 10:56:42 +0200
Subject: [PATCH] Address Codex OAuth review feedback

---
 strix/llm/codex_oauth.py      |  2 +-
 strix/llm/llm.py              |  7 ++--
 tests/llm/test_codex_oauth.py | 65 +++++++++++++++++++++++++++++++++++
 3 files changed, 69 insertions(+), 5 deletions(-)

diff --git a/strix/llm/codex_oauth.py b/strix/llm/codex_oauth.py
index 2ecb0528..80e24e72 100644
--- a/strix/llm/codex_oauth.py
+++ b/strix/llm/codex_oauth.py
@@ -206,7 +206,7 @@ def _post_codex_responses(
         "Authorization": f"Bearer {credentials.access_token}",
         "Accept": "text/event-stream",
         "Content-Type": "application/json",
-        "version": "strix-codex-oauth",
+        "User-Agent": "strix-codex-oauth",
     }
     if credentials.account_id:
         headers["ChatGPT-Account-ID"] = credentials.account_id
diff --git a/strix/llm/llm.py b/strix/llm/llm.py
index ec35621e..4f70c013 100644
--- a/strix/llm/llm.py
+++ b/strix/llm/llm.py
@@ -10,7 +10,7 @@ from litellm import acompletion, completion_cost, stream_chunk_builder, supports
 from litellm.utils import supports_prompt_caching, supports_vision
 
 from strix.config import Config
-from strix.llm.codex_oauth import complete_codex_oauth
+from strix.llm.codex_oauth import CodexOAuthError, complete_codex_oauth
 from strix.llm.config import LLMConfig
 from strix.llm.memory_compressor import MemoryCompressor, get_message_tokens
 from strix.llm.utils import (
@@ -314,9 +314,6 @@ class LLM:
             self._total_stats.input_tokens += usage.get("input_tokens", 0)
             self._total_stats.output_tokens += usage.get("output_tokens", 0)
 
-        if content:
-            yield LLMResponse(content=content)
-
         content = _THINKING_BLOCK_RE.sub("", content)
         content = normalize_tool_format(content)
         content = fix_incomplete_tool_call(_truncate_to_first_function(content))
@@ -384,6 +381,8 @@ class LLM:
             return 0.0
 
     def _should_retry(self, e: Exception) -> bool:
+        if isinstance(e, CodexOAuthError):
+            return False
         code = getattr(e, "status_code", None) or getattr(
             getattr(e, "response", None), "status_code", None
         )
diff --git a/tests/llm/test_codex_oauth.py b/tests/llm/test_codex_oauth.py
index 234247cb..c848f22b 100644
--- a/tests/llm/test_codex_oauth.py
+++ b/tests/llm/test_codex_oauth.py
@@ -4,13 +4,16 @@ from pathlib import Path
 import pytest
 
 from strix.llm.codex_oauth import (
+    CodexOAuthCredentials,
     CodexOAuthError,
+    _post_codex_responses,
     build_codex_responses_payload,
     load_codex_oauth_credentials,
     parse_responses_sse_events,
     refresh_codex_oauth_credentials,
 )
 from strix.llm.config import LLMConfig
+from strix.llm.llm import LLM
 
 
 def test_llm_config_detects_codex_oauth_provider(monkeypatch: pytest.MonkeyPatch) -> None:
@@ -22,6 +25,41 @@ def test_llm_config_detects_codex_oauth_provider(monkeypatch: pytest.MonkeyPatch
     assert config.codex_model == "gpt-5.5"
 
 
+def test_llm_does_not_retry_codex_oauth_errors(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("STRIX_LLM", "codex/gpt-5.5")
+
+    llm = LLM(LLMConfig())
+
+    assert llm._should_retry(CodexOAuthError("Run `codex login` first.")) is False
+
+
+@pytest.mark.asyncio
+async def test_codex_oauth_stream_yields_single_processed_response(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    monkeypatch.setenv("STRIX_LLM", "codex/gpt-5.5")
+
+    def fake_complete_codex_oauth(*args, **kwargs):
+        return (
+            "<thinking>hidden</thinking><function=finish><parameter=summary>done</parameter></function>",
+            {"input_tokens": 1, "output_tokens": 2},
+        )
+
+    monkeypatch.setattr("strix.llm.llm.complete_codex_oauth", fake_complete_codex_oauth)
+
+    llm = LLM(LLMConfig())
+    responses = [
+        response
+        async for response in llm._stream_codex_oauth([{"role": "user", "content": "Hi"}])
+    ]
+
+    assert len(responses) == 1
+    assert "<thinking>" not in responses[0].content
+    assert responses[0].tool_invocations
+    assert llm._total_stats.input_tokens == 1
+    assert llm._total_stats.output_tokens == 2
+
+
 def test_load_codex_oauth_credentials_reads_codex_auth_json(tmp_path: Path) -> None:
     auth_file = tmp_path / "auth.json"
     auth_file.write_text(
@@ -95,6 +133,33 @@ def test_refresh_codex_oauth_credentials_persists_new_tokens(tmp_path: Path) ->
     assert calls[0][1]["json"]["grant_type"] == "refresh_token"
 
 
+def test_post_codex_responses_uses_user_agent_header(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    calls = []
+
+    class FakeResponse:
+        pass
+
+    def fake_post(url: str, **kwargs):
+        calls.append((url, kwargs))
+        return FakeResponse()
+
+    monkeypatch.setattr("strix.llm.codex_oauth.requests.post", fake_post)
+
+    response = _post_codex_responses(
+        "https://example.test/codex",
+        CodexOAuthCredentials(access_token="access-token", account_id="account-id"),
+        {"model": "gpt-5.5"},
+        60,
+    )
+
+    assert isinstance(response, FakeResponse)
+    headers = calls[0][1]["headers"]
+    assert headers["User-Agent"] == "strix-codex-oauth"
+    assert "version" not in headers
+
+
 def test_build_codex_responses_payload_converts_chat_messages() -> None:
     payload = build_codex_responses_payload(
         model="gpt-5.5",