Merge pull request #858 from Fwind43/fix/claude-stream-thinking

feat(claude): 新增 show_thinking 选项，可关闭 redact-thinking 让思考过程流式显示

llmcore.py

@@ -808,6 +808,7 @@ class NativeClaudeSession(BaseSession):
         self.tools = None
         if self.user_agent == self.default_ua: self.user_agent = self.native_ua
         self.api_key_header = str(cfg.get('api_key_header', 'auto')).strip().lower()
+        self.show_thinking = bool(cfg.get('show_thinking'))  # True: stream thinking text (like CC showThinkingSummaries)
     def raw_ask(self, messages):
         if self.max_tokens is None: self.max_tokens = 8192
         model = self.model
@@ -815,6 +816,7 @@ class NativeClaudeSession(BaseSession):
         if 'claude' in model.lower(): messages = _drop_unsigned_thinking(messages)
         messages = _ensure_thinking_blocks(messages, self.model)
         beta_parts = ["claude-code-20250219", "interleaved-thinking-2025-05-14", "redact-thinking-2026-02-12", "thinking-token-count-2026-05-13", "context-management-2025-06-27", "prompt-caching-scope-2026-01-05", "mid-conversation-system-2026-04-07", "effort-2025-11-24", "fallback-credit-2026-06-01"]
+        if self.show_thinking: beta_parts.remove("redact-thinking-2026-02-12")
         if "[1m]" in model.lower():
             beta_parts.insert(1, "context-1m-2025-08-07"); model = model.replace("[1m]", "").replace("[1M]", "")
         headers = {"Content-Type": "application/json", "anthropic-version": "2023-06-01",

mykey_template.py

@@ -35,6 +35,7 @@
 #     'thinking_type': 'adaptive',      # Claude 思考：adaptive / enabled / disabled
 #     'thinking_budget_tokens': 32768,  # thinking_type='enabled' 时需要
 #     'claude_compat': 'half',          # 需要 Claude Code 请求格式/指纹校验的透传渠道（simple=仅身份行，half=含 CC 主提示）
+#     'show_thinking': True,            # Claude：流式显示思考过程（默认按 Claude Code 隐藏）
 #     'api_key_header': 'x-api-key',    # Claude 鉴权头不匹配时指定；也支持 bearer
 #     'api_mode': 'responses',          # OAI：默认 chat_completions，按渠道要求修改
 # 运行中也可临时调整，例如：/session.reasoning_effort=high

mykey_template_en.py

@@ -35,6 +35,7 @@
 #     'thinking_type': 'adaptive',      # Claude thinking: adaptive / enabled / disabled
 #     'thinking_budget_tokens': 32768,  # Required when thinking_type='enabled'
 #     'claude_compat': 'half',          # Relays requiring CC request format/fingerprint (simple=identity only, half=+CC main prompt)
+#     'show_thinking': True,            # Claude: stream thinking text (hidden by default, as in Claude Code)
 #     'api_key_header': 'x-api-key',    # Override Claude auth header; also supports bearer
 #     'api_mode': 'responses',          # OAI: default is chat_completions; follow provider docs
 # Temporary runtime overrides also work, e.g. /session.reasoning_effort=high