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.

artemis/agents/video_analyzer/universal_engine.py

@@ -35,6 +35,8 @@ from langchain_core.messages import (
     ToolMessage,
 )
 
+from artemis.llm.structured import content_to_text
+
 from artemis.agents.video_analyzer import video_analyzer as _va
 from artemis.agents.video_analyzer.reliability import (
     SubAgentAnswerExhausted,
@@ -241,7 +243,7 @@ def _extract_submit_payload(response) -> tuple[str, str, list]:
         return final_summary, final_analysis, list(args.get("timeline_events") or [])
 
     raw_content = getattr(response, "content", None)
-    text_content = str(raw_content).strip() if raw_content else ""
+    text_content = content_to_text(raw_content).strip()
     if text_content:
         return text_content[:200].replace("\n", " "), text_content, []
 
@@ -435,7 +437,7 @@ async def run_universal(
             messages.append(response)
 
             if not getattr(response, "tool_calls", None):
-                agent_outcome = str(response.content)
+                agent_outcome = content_to_text(response.content)
                 break
 
             for tc in response.tool_calls: