feat: Full code review, bug fixes, and philosophy book generation

This commit includes:
- A full code review and bug fixes for language drift, package loading, and CLI crashes.
- The generated 15,000-word philosophy manuscript.
- CODE_REVIEW.md and CHANGELOG.md documenting the process.

opus_orchestrator/utils/llm.py

@@ -83,6 +83,9 @@ class LLMClient:
             "Authorization": f"Bearer {self.api_key}",
             "Content-Type": "application/json",
         }
+
+        # STRICT ENGLISH ENFORCEMENT
+        system_prompt += "\n\nIMPORTANT: You must respond ONLY in English. Do not use Chinese characters or any other language under any circumstances."
         
         if self.provider == "minimax":
             return self._complete_minimax_sync(
@@ -110,6 +113,10 @@ class LLMClient:
                 "Content-Type": "application/json",
             }
             
+            # STRICT ENGLISH ENFORCEMENT
+            nonlocal system_prompt
+            system_prompt += "\n\nIMPORTANT: You must respond ONLY in English. Do not use Chinese characters or any other language under any circumstances."
+            
             if self.provider == "minimax":
                 return await self._complete_minimax_async(
                     system_prompt, user_prompt, temperature, max_tokens, headers
@@ -140,8 +147,8 @@ class LLMClient:
         # Anthropic-compatible format
         payload = {
             "model": self.minimax_model,
+            "system": system_prompt,
             "messages": [
-                {"role": "system", "content": system_prompt},
                 {"role": "user", "content": user_prompt},
             ],
             "temperature": temperature,
@@ -167,8 +174,15 @@ class LLMClient:
         
         # Handle Anthropic-compatible response format
         if "content" in data:
-            # Return the text content
             if isinstance(data["content"], list) and len(data["content"]) > 0:
+                # Look for text content, skip thinking
+                text_parts = []
+                for item in data["content"]:
+                    if item.get("type") == "text":
+                        text_parts.append(item.get("text", ""))
+                if text_parts:
+                    return "".join(text_parts)
+                # If no text found, return first item's text or the item itself
                 return data["content"][0].get("text", str(data["content"][0]))
             return str(data["content"])
         else:
@@ -224,8 +238,8 @@ class LLMClient:
         """Call MiniMax API (sync) using Anthropic-compatible endpoint."""
         payload = {
             "model": self.minimax_model,
+            "system": system_prompt,
             "messages": [
-                {"role": "system", "content": system_prompt},
                 {"role": "user", "content": user_prompt},
             ],
             "temperature": temperature,