feat: add Chinese UI, Responses models, and mirrored Windows setup

Localize the web console and persist custom model and capability settings.
Use OpenAI Responses exclusively and move runtime limits into environment configuration.
Bootstrap uv and the virtual environment on Windows, support optional China mirrors,
and package source distributions without environments, build output, or credentials.

Update regression tests and repository quality checks for the new behavior.

tests/unit/memory/test_history_chunking.py

@@ -1005,7 +1005,7 @@ async def test_capsule_outage_without_fallback_exhausts_to_failed():
     assert service.get_summary("chunk:1-2") is None
 
 
-def test_capsule_fallback_model_resolution_google_only_and_not_primary():
+def test_capsule_fallback_model_resolution_openai_only_and_not_primary():
     mgr = HistoryChunkManager(
         capsule_service=StubCapsuleService(),
         chunking_config=SimpleNamespace(
@@ -1024,13 +1024,13 @@ def test_capsule_fallback_model_resolution_google_only_and_not_primary():
         )
 
     assert (
-        mgr._resolve_capsule_fallback_model(ctx_with("google", "gemini-3.6-flash"))
+        mgr._resolve_capsule_fallback_model(ctx_with("openai", "gemini-3.6-flash"))
         == "gemini-3.6-flash"
     )
-    # Non-google fallbacks cannot ride the raw google model path.
-    assert mgr._resolve_capsule_fallback_model(ctx_with("openai", "gpt-4o-mini")) is None
+    # Non-Responses providers are rejected.
+    assert mgr._resolve_capsule_fallback_model(ctx_with("google", "gpt-4o-mini")) is None
     # A fallback identical to the primary adds nothing.
-    assert mgr._resolve_capsule_fallback_model(ctx_with("google", "gemini-3.7-flash")) is None
+    assert mgr._resolve_capsule_fallback_model(ctx_with("openai", "gemini-3.7-flash")) is None
 
 
 @pytest.mark.asyncio