fix: add requests to pip install & tune max_turns/prompt

README.md

@@ -79,7 +79,7 @@ git clone https://github.com/lsdefine/GenericAgent.git
 cd GenericAgent
 
 # 2. Install minimal dependencies
-pip install streamlit pywebview
+pip install requests streamlit pywebview
 
 # 3. Configure API Key
 cp mykey_template.py mykey.py
@@ -283,7 +283,7 @@ git clone https://github.com/lsdefine/GenericAgent.git
 cd GenericAgent
 
 # 2. 安装最小依赖
-pip install streamlit pywebview
+pip install requests streamlit pywebview
 
 # 3. 配置 API Key
 cp mykey_template.py mykey.py

agentmain.py

@@ -141,7 +141,7 @@ class GeneraticAgent:
             if 'gpt' in self.get_llm_name(model=True): handler._done_hooks.append('请确定任务是否完成，如果完成请给出信息完整的简报回答，如未完成需要继续工具调用直到完成任务，确实需要问用户应使用ask_user工具')
             # although new handler, the **full** history is in llmclient, so it is full history!
             gen = agent_runner_loop(self.llmclient, sys_prompt, user_input, 
-                                handler, TOOLS_SCHEMA, max_turns=40, verbose=self.verbose)
+                                handler, TOOLS_SCHEMA, max_turns=70, verbose=self.verbose)
             try:
                 full_resp = ""; last_pos = 0
                 for chunk in gen:

ga.py

@@ -417,7 +417,7 @@ class GenericAgentHandler(BaseHandler):
     def _in_plan_mode(self): return self.working.get('in_plan_mode')
     def _exit_plan_mode(self): self.working.pop('in_plan_mode', None)
     def enter_plan_mode(self, plan_path): 
-        self.working['in_plan_mode'] = plan_path; self.max_turns = 80
+        self.working['in_plan_mode'] = plan_path; self.max_turns = 100
         print(f"[Info] Entered plan mode with plan file: {plan_path}"); return plan_path
     def _check_plan_completion(self):
         if not os.path.isfile(p:=self._in_plan_mode() or ''): return None
@@ -525,7 +525,7 @@ class GenericAgentHandler(BaseHandler):
             next_prompt += "\n[DANGER] 上一轮遗漏了<summary>，需要按协议在<summary>中输出极简单行摘要！" 
         summary = smart_format(summary, max_str_len=100)
         self.history_info.append(f'[Agent] {summary}')
-        if turn % 35 == 0 and 'plan' not in str(self.working.get('related_sop')):
+        if turn % 65 == 0 and 'plan' not in str(self.working.get('related_sop')):
             next_prompt += f"\n\n[DANGER] 已连续执行第 {turn} 轮。你必须总结情况进行ask_user，不允许继续重试。"
         elif turn % 7 == 0:
             next_prompt += f"\n\n[DANGER] 已连续执行第 {turn} 轮。禁止无效重试。若无有效进展，必须切换策略：1. 探测物理边界 2. 请求用户协助。如有需要，可调用 update_working_checkpoint 保存关键上下文。"
@@ -533,7 +533,7 @@ class GenericAgentHandler(BaseHandler):
 
         if (_plan := self._in_plan_mode()) and turn >= 10 and turn % 5 == 0:
             next_prompt = f"[Plan Hint] 你正在计划模式。必须 file_read({_plan}) 确认当前步骤，回复开头引用：📌 当前步骤：...\n\n" + next_prompt
-        if _plan and turn >= 70: next_prompt += f"\n\n[DANGER] Plan模式已运行 {turn} 轮，已达上限。必须 ask_user 汇报进度并确认是否继续。"
+        if _plan and turn >= 90: next_prompt += f"\n\n[DANGER] Plan模式已运行 {turn} 轮，已达上限。必须 ask_user 汇报进度并确认是否继续。"
 
         injkeyinfo = consume_file(self.parent.task_dir, '_keyinfo')
         injprompt = consume_file(self.parent.task_dir, '_intervene')

llmcore.py

@@ -315,8 +315,7 @@ def _stamp_oai_cache_markers(messages, model):
 
 def _openai_stream(api_base, api_key, messages, model, api_mode='chat_completions', *,
                    temperature=0.5, max_tokens=None, tools=None, reasoning_effort=None,
-                   max_retries=0, connect_timeout=10, read_timeout=300, proxies=None,
+                   max_retries=0, connect_timeout=10, read_timeout=300, proxies=None, stream=True):
-                   stream=True):
     """Shared OpenAI-compatible streaming request with retry. Yields text chunks, returns list[content_block]."""
     ml = model.lower()
     if 'kimi' in ml or 'moonshot' in ml: temperature = 1
@@ -912,19 +911,13 @@ class MixinSession:
 
 THINKING_PROMPT_ZH = """
 ### 行动规范（持续有效）
-每次回复请先在回复文字中包含：
+每次回复请先在回复文字中包含一个<summary></summary> 中输出极简单行（<30字）物理快照：上次结果新信息+本次意图。此内容进入长期工作记忆。
-1. 在 <thinking></thinking> 标签中先分析现状和策略
+\n**若用户需求未完成，必须进行工具调用！**
-2. 在 <summary></summary> 中输出极简单行（<30字）物理快照：上次结果新信息+本次意图。此内容进入长期工作记忆。
-再进行回答。
-\n**除了最后回答，必须进行工具调用！**
 """.strip()
 THINKING_PROMPT_EN = """
 ### Action Protocol (always in effect)
-The reply body should first include:
+The reply body should first include a minimal one-line (<30 words) physical snapshot in <summary></summary>: new info from last result + current intent. This goes into long-term working memory.
-1. Analyze the current situation and strategy inside <thinking></thinking>
+\n**If the user's request is not yet complete, tool calls are required!**
-2. Output a minimal one-line (<30 words) physical snapshot in <summary></summary>: new info from last result + current intent. This goes into long-term working memory.
-Then reply.
-\n**Tool calls are required for every turn except the final answer!**
 """.strip()
 
 class NativeToolClient: