feat: add ModelScope quick setup, dynamic summary trigger, and remove max_tokens

- Add 魔搭快捷配置 flow: one-click setup with 12 preset models via API key
- Replace hardcoded summary trigger (170k) with dynamic calculation based on
  model context window size (90% threshold)
- Remove deprecated max_tokens parameter in favor of max_completion_tokens
- Fix edit_current_model not persisting updated config to disk
- Improve error message display with full exception details on retry failure
- Enhance /summary JSON parsing with bracket-matching fallback for LLM output
- Add new model context window sizes (deepseek-v3.2, deepseek-r1-0528, etc.)
- Add psutil and pytest-timeout dependencies
- Update tests to reflect config flow changes

Author: ScarletMercy

README.md

@@ -34,7 +34,7 @@ https://github.com/ScarletMercy/chcode/blob/main/assets/test.mp4
 - Compatible with **all OpenAI-compatible APIs** (OpenAI, DeepSeek, Qwen, GLM, Claude via proxy, etc.)
 - First-run wizard with **env auto-detection** (scans `OPENAI_API_KEY`, `DEEPSEEK_API_KEY`, `ZHIPU_API_KEY`, etc.)
 - Create / edit / switch models at runtime
-- Per-model hyperparameter tuning (temperature, top_p, top_k, max_tokens, stop_sequences, etc.)
+- Per-model hyperparameter tuning (temperature, top_p, top_k, max_completion_tokens, stop_sequences, etc.)
 - Automatic **retry with exponential backoff** (3/10/30/60s) and fallback model switching on persistent failure
 
 ### Session & History

chcode/agent_setup.py

@@ -136,13 +136,13 @@ async def model_retry_with_backoff(
                 if fallback:
                     console.print(f"[yellow]主模型重试{retry_count}次失败，切换到备用模型...[/yellow]")
                     raise ModelSwitchError("切换到备用模型")
-                console.print(f"[red]请求失败，无备用模型可用，放弃请求[/red]")
+                console.print(f"[red]请求失败，无备用模型可用，放弃请求\n  {e}[/red]")
                 raise
 
             delay_idx = min(retry_count - 1, len(RETRY_DELAYS) - 1)
             delay = RETRY_DELAYS[delay_idx]
 
-            console.print(f"[yellow]请求失败 ({retry_count}/{max_retries}), {delay}秒后重试... ({str(e)[:50]})[/yellow]")
+            console.print(f"[yellow]请求失败 ({retry_count}/{max_retries}), {delay}秒后重试...\n  {e}[/yellow]")
 
             await asyncio.sleep(delay)
 
@@ -266,13 +266,18 @@ def build_agent(
     _summarization_model = EnhancedChatOpenAI(**cfg)
 
     # 加载 fallback 模型配置
-    from chcode.config import load_model_json
+    from chcode.config import load_model_json, get_context_window_size
 
     data = load_model_json()
     fallback = data.get("fallback", {})
     if fallback:
         set_fallback_models(list(fallback.values()))
 
+    # 摘要触发阈值 = 上下文窗口的 90%
+    model_name = cfg.get("model", "")
+    ctx_window = get_context_window_size(model_name)
+    summary_trigger = int(ctx_window * 0.9)
+
     agent = create_agent(
         model,
         _get_all_tools() + (mcp_tools or []),
@@ -295,7 +300,7 @@ def build_agent(
             ),
             SummarizationMiddleware(
                 model=_summarization_model,
-                trigger=("tokens", 170_000),
+                trigger=("tokens", summary_trigger),
                 keep=("messages", 20),
             ),
             _hitl_middleware,

chcode/chat.py

@@ -434,7 +434,6 @@ def _tab_toggle_mode(event):
                 from chcode.agent_setup import update_hitl_config
 
                 update_hitl_config(self.yolo)
-                mode = "Yolo" if self.yolo else "Common"
                 event.app.renderer._last_rendered_width = 0  # 强制刷新 toolbar
 
             _last_width = 0
@@ -738,9 +737,30 @@ async def _cmd_compress(self, _arg: str) -> None:
                 import re
 
                 content = raw_resp.content.strip()
+                # 去除 markdown 代码块包裹
                 if content.startswith("```"):
                     content = re.sub(r"^```(?:json)?\s*\n?", "", content)
                     content = re.sub(r"\n?```\s*$", "", content)
+                # 提取包含 "summary" 的 JSON 对象（模型可能在 JSON 前输出思考内容）
+                json_match = re.search(r'\{[^{}]*"summary"[^{}]*\}', content)
+                if json_match:
+                    content = json_match.group()
+                else:
+                    # 可能 summary 值中包含嵌套对象，用逐字符括号匹配兜底
+                    depth = 0
+                    start = -1
+                    for i, ch in enumerate(content):
+                        if ch == '{':
+                            if depth == 0:
+                                start = i
+                            depth += 1
+                        elif ch == '}':
+                            depth -= 1
+                            if depth == 0 and start >= 0:
+                                candidate = content[start:i+1]
+                                if '"summary"' in candidate:
+                                    content = candidate
+                                    break
                 data = json.loads(content)
                 ai_content = data.get("summary", "")
                 if not ai_content:

chcode/config.py

@@ -122,6 +122,7 @@ async def first_run_configure() -> dict | None:
 
     if detected:
         choices = [f"{d['name']} (检测到 {d['env_var']})" for d in detected]
+        choices.append("魔搭快捷配置...")
         choices.append("手动配置...")
         choices.append("退出")
 
@@ -135,6 +136,9 @@ async def first_run_configure() -> dict | None:
         if "手动" in result:
             return await configure_new_model()
 
+        if "魔搭" in result:
+            return await _configure_modelscope_with_test()
+
         idx = choices.index(result)
         chosen = detected[idx]
 
@@ -169,19 +173,26 @@ async def first_run_configure() -> dict | None:
         return config
     else:
         console.print("[yellow]未检测到环境变量中的 API Key[/yellow]")
-        choices = ["手动配置...", "退出"]
+        choices = ["魔搭快捷配置...", "手动配置...", "退出"]
         result = await select("选择:", choices)
         if result is None or "退出" in result:
             console.print("[dim]提示: 在环境变量中设置 API Key 后重新运行，例如:[/dim]")
             console.print("[dim]  set BIGMODEL_API_KEY=your_key[/dim]")
             console.print("[dim]或执行 chcode config new 手动配置[/dim]")
             return None
+        if "魔搭" in result:
+            return await _configure_modelscope_with_test()
         return await configure_new_model()
 
 
 async def configure_new_model() -> dict | None:
     """新建模型配置（交互式表单）"""
     ensure_config_dir()
+    result = await select("配置方式:", ["魔搭快捷配置...", "手动配置..."])
+    if result is None:
+        return None
+    if "魔搭" in result:
+        return await _configure_modelscope_with_test()
     config = await model_config_form()
     if config is None:
         return None
@@ -222,6 +233,56 @@ async def configure_new_model() -> dict | None:
     return config
 
 
+async def _configure_modelscope_with_test() -> dict | None:
+    """魔搭快捷配置：收集 API Key → 测试连接 → 保存 12 个预定义模型。"""
+    from chcode.prompts import configure_modelscope
+
+    ms_config = await configure_modelscope()
+    if ms_config is None:
+        return None
+
+    default = ms_config["default"]
+
+    # 测试连接
+    console.print("[yellow]测试连接中...[/yellow]")
+    try:
+        from chcode.utils.enhanced_chat_openai import EnhancedChatOpenAI
+
+        model_inst = EnhancedChatOpenAI(**default)
+        await asyncio.to_thread(model_inst.invoke, "你好")
+    except Exception as e:
+        import traceback
+
+        err_msg = str(e)
+        if "null value for 'choices'" not in err_msg:
+            console.print(f"[red]连接测试失败: {err_msg}[/red]")
+            console.print(f"[dim]{traceback.format_exc()}[/dim]")
+            return None
+
+    # 合并到已有配置，保留非魔搭的已有模型
+    data = load_model_json()
+    old_default = data.get("default")
+    existing_fallback = data.get("fallback", {})
+
+    if not old_default:
+        # 首次配置 — 魔搭直接作为完整配置
+        save_model_json(ms_config)
+    else:
+        # 已有配置 — 旧的 default 移入 fallback，魔搭作为新 default，合并 fallback
+        if old_default["model"] not in ms_config["fallback"]:
+            existing_fallback[old_default["model"]] = old_default
+        existing_fallback.update(ms_config["fallback"])
+        data["default"] = ms_config["default"]
+        data["fallback"] = existing_fallback
+        save_model_json(data)
+    fallback_names = ", ".join(ms_config["fallback"].keys())
+    console.print(f"[green]配置完成: {default['model']} (默认)[/green]")
+    console.print(f"[dim]备用模型 ({len(ms_config['fallback'])} 个): {fallback_names}[/dim]")
+
+    await configure_tavily()
+    return default
+
+
 async def edit_current_model() -> dict | None:
     """编辑当前默认模型"""
     data = load_model_json()
@@ -249,6 +310,7 @@ async def edit_current_model() -> dict | None:
             console.print(f"[red]连接测试失败: {err_msg}[/red]")
             console.print(f"[dim]{traceback.format_exc()}[/dim]")
             return None
+    data["default"] = config
     save_model_json(data)
     console.print(f"[green]模型配置已更新: {config['model']}[/green]")
     return config
@@ -357,11 +419,15 @@ def save_tavily_api_key(api_key: str) -> None:
     "gpt-4o-mini": 128000,
     "claude-sonnet-4-20250514": 200000,
     "deepseek-chat": 65536,
+    "deepseek-v3.2": 128000,
+    "deepseek-r1-0528": 65536,
     "glm-5.1": 200000,
     "glm-5": 200000,
     "glm-4.7": 200000,
     "minimax-m2": 204800,
-    "kimi-k2": 262144,
+    "minimax-m2.5": 200000,
+    "kimi-k2": 256000,
+    "mimo-v2-flash": 256000,
     "qwen3.5-plus": 1000000,
     "qwen3.6-plus": 1000000,
     "qwen": 256000,