feat(codex): 支持compact端点非流式请求，更新请求头与字段清理

- Codex适配器支持compact端点：非流式请求使用application/json Accept头，
  stream_policy根据compact上下文返回FORCE_NON_STREAM
- 更新Codex请求头格式：添加Version/Connection头，header key首字母大写
- 简化include列表处理：normalizer和request_patching统一强制为固定列表
- 清理Codex不支持的字段：truncation、context_management、user
- 默认instructions改为空字符串
- 前端用量页面：用户页面使用前端筛选后总数，避免不必要的后端分页请求

tests/services/test_codex_request.py

@@ -12,8 +12,8 @@ def test_patch_openai_cli_request_for_codex_sets_store_and_instructions() -> Non
 
     assert out is not req
     assert out["store"] is False
-    assert "instructions" in out
-    assert isinstance(out["instructions"], str)
+    assert out["stream"] is True
+    assert out["instructions"] == ""
 
 
 def test_patch_openai_cli_request_for_codex_strips_rejected_params() -> None:
@@ -22,20 +22,24 @@ def test_patch_openai_cli_request_for_codex_strips_rejected_params() -> None:
         "input": [],
         "max_output_tokens": 123,
         "max_completion_tokens": 456,
-        "max_tokens": 789,
         "temperature": 0.5,
         "top_p": 0.9,
         "service_tier": "default",
+        "truncation": "auto",
+        "context_management": {"compaction": {"type": "summary"}},
+        "user": "u_123",
     }
     out = patch_openai_cli_request_for_codex(req)
 
     for key in (
         "max_output_tokens",
         "max_completion_tokens",
-        "max_tokens",
         "temperature",
         "top_p",
         "service_tier",
+        "truncation",
+        "context_management",
+        "user",
     ):
         assert key not in out
 
@@ -68,8 +72,28 @@ def test_patch_openai_cli_request_for_codex_adds_required_include_item() -> None
     req = {"model": "gpt-test", "input": []}
     out = patch_openai_cli_request_for_codex(req)
 
-    assert "include" in out
-    assert "reasoning.encrypted_content" in out["include"]
+    assert out["include"] == ["reasoning.encrypted_content"]
+
+
+def test_patch_openai_cli_request_for_codex_overrides_include() -> None:
+    req = {
+        "model": "gpt-test",
+        "input": [],
+        "include": ["foo", "bar"],
+    }
+    out = patch_openai_cli_request_for_codex(req)
+    assert out["include"] == ["reasoning.encrypted_content"]
+
+
+def test_patch_openai_cli_request_for_codex_compact_drops_stream() -> None:
+    req = {
+        "model": "gpt-test",
+        "input": [],
+        "_aether_compact": True,
+        "stream": True,
+    }
+    out = patch_openai_cli_request_for_codex(req)
+    assert "stream" not in out
 
 
 def test_maybe_patch_request_for_codex_is_noop_for_non_codex() -> None:
@@ -110,7 +134,34 @@ def test_codex_envelope_extra_headers_includes_sse_accept_and_session() -> None:
 
     headers = codex_oauth_envelope.extra_headers() or {}
     assert headers.get("Accept") == "text/event-stream"
-    assert headers.get("x-oai-web-search-eligible") == "true"
-    assert headers.get("originator") == "codex_cli_rs"
-    assert isinstance(headers.get("session_id"), str)
-    assert headers.get("session_id")
+    assert headers.get("Originator") == "codex_cli_rs"
+    assert headers.get("Version") == "0.101.0"
+    assert headers.get("Connection") == "Keep-Alive"
+    assert isinstance(headers.get("Session_id"), str)
+    assert headers.get("Session_id")
+
+
+def test_codex_envelope_extra_headers_compact_uses_json_accept() -> None:
+    from src.services.provider.adapters.codex.context import (
+        CodexRequestContext,
+        set_codex_request_context,
+    )
+    from src.services.provider.adapters.codex.envelope import codex_oauth_envelope
+
+    set_codex_request_context(CodexRequestContext(is_compact=True))
+    headers = codex_oauth_envelope.extra_headers() or {}
+    assert headers.get("Accept") == "application/json"
+    set_codex_request_context(None)
+
+
+def test_codex_envelope_extra_headers_uses_account_id_header() -> None:
+    from src.services.provider.adapters.codex.context import (
+        CodexRequestContext,
+        set_codex_request_context,
+    )
+    from src.services.provider.adapters.codex.envelope import codex_oauth_envelope
+
+    set_codex_request_context(CodexRequestContext(account_id="acc_123"))
+    headers = codex_oauth_envelope.extra_headers() or {}
+    assert headers.get("Chatgpt-Account-Id") == "acc_123"
+    set_codex_request_context(None)

tests/services/test_provider_transport_codex.py

@@ -3,6 +3,10 @@ from __future__ import annotations
 from dataclasses import dataclass
 from types import SimpleNamespace
 
+from src.services.provider.adapters.codex.context import (
+    CodexRequestContext,
+    set_codex_request_context,
+)
 from src.services.provider.transport import build_provider_url
 
 
@@ -44,3 +48,19 @@ def test_codex_openai_cli_does_not_duplicate_responses_suffix() -> None:
     )
 
     assert url == "https://chatgpt.com/backend-api/codex/responses"
+
+
+def test_codex_openai_cli_uses_compact_suffix_when_context_marked_compact() -> None:
+    endpoint = _DummyEndpoint(
+        base_url="https://chatgpt.com/backend-api/codex",
+        api_format="openai:cli",
+        provider=SimpleNamespace(provider_type="codex"),
+    )
+    set_codex_request_context(CodexRequestContext(is_compact=True))
+    url = build_provider_url(
+        endpoint,  # type: ignore[arg-type]
+        path_params={"model": "ignored"},
+        is_stream=False,
+    )
+    assert url == "https://chatgpt.com/backend-api/codex/responses/compact"
+    set_codex_request_context(None)

tests/services/test_upstream_stream_policy.py

@@ -3,6 +3,10 @@ from __future__ import annotations
 from dataclasses import dataclass
 from types import SimpleNamespace
 
+from src.services.provider.adapters.codex.context import (
+    CodexRequestContext,
+    set_codex_request_context,
+)
 from src.services.provider.stream_policy import (
     UpstreamStreamPolicy,
     enforce_stream_mode_for_upstream,
@@ -42,6 +46,19 @@ def test_get_upstream_stream_policy_codex_ignores_force_non_stream_config() -> N
     assert get_upstream_stream_policy(ep) == UpstreamStreamPolicy.FORCE_STREAM
 
 
+def test_get_upstream_stream_policy_codex_compact_forces_non_stream() -> None:
+    ep = _DummyEndpoint(
+        api_format="openai:cli",
+        config=None,
+        provider=SimpleNamespace(provider_type="codex"),
+    )
+    try:
+        set_codex_request_context(CodexRequestContext(is_compact=True))
+        assert get_upstream_stream_policy(ep) == UpstreamStreamPolicy.FORCE_NON_STREAM
+    finally:
+        set_codex_request_context(None)
+
+
 def test_enforce_stream_mode_for_upstream_openai_chat_sets_stream_options_usage() -> None:
     body = {"stream": False}
     out = enforce_stream_mode_for_upstream(