Fix story maker bug

rootfs/agents/fairy_tale_writer/agent.toml

@@ -1,6 +1,7 @@
 instance_id = "fairy_tale_writer"
 fullname = "tracy wang"
 llm_model_name = "gpt-3.5-turbo-16k-0613"
+enable_function = []
 
 [[prompt]]
 role = "system"

rootfs/agents/manager/agent.toml

@@ -1,5 +1,7 @@
 instance_id = "agent:xxxxxxabcde"
 fullname = "musk"
+enable_function = []
+
 [[prompt]]
 role = "system"
 content = "你有丰富的管理技能，擅长将复杂工作拆解成简单的任务，让团队成员高效协作。"

rootfs/agents/studio_director/agent.toml

@@ -1,7 +1,7 @@
 instance_id = "studio_director"
 fullname = "tracy wang"
-llm_model_name = "gpt-3.5-turbo-16k-0613"
+enable_function = ["text_to_speech"]
 
 [[prompt]]
 role = "system"
-content = "你是一个演播导演，请将下面故事改编成朗读剧本，提取旁白和角色台词，每个角色需要有性别、年龄、以及每句台词的语气。并调用text_to_speech function生成音频数据。"
+content = "你是一个故事播音员，请将下面故事改编成播音剧本，提取旁白和角色台词，以及每个角色需要有性别、年龄、以及每句台词的语气，最后生成音频文件。"

rootfs/workflows/story_maker/workflow.toml

@@ -8,10 +8,12 @@ name = "story_maker"
 name = "manager"
 fullname = "总导演"
 agent="manager"
+enable_function = []
+
 [[roles.manager.prompt]]
 role="system"
 content="""
-你是一个语音故事制作总导演,与客户对接并向团队下达指令。你的团队分为下面几个成员:writer,studio_director。一个故事制作分成两个阶段：让writer写出故事，再交由studio_director演播故事。你的基本工作模式是：
+你是一个语音故事制作总导演,与客户对接并向团队下达指令。你的团队分为下面几个成员:writer,studio_director。一个故事制作分成两个阶段：让writer写出故事，再交由studio_director演播故事生成音频文件。你的基本工作模式是：
 1. 收到客户的明确的指令后,让writer写出故事
 2. 将writer写出的故事交给studio_director演播
 3. 当你决定要和成员通信时，请使用下面形式输出需要通信的消息
@@ -25,6 +27,7 @@ content="""
 name = "writer"
 agent = "fairy_tale_writer"
 fullname = "作家"
+enable_function = []
 [[roles.writer.prompt]]
 role="system"
 content=""
@@ -32,6 +35,7 @@ content=""
 [roles.studio_director]
 name = "studio_director"
 agent = "studio_director"
+enable_function = ["text_to_speech"]
 [[roles.studio_director.prompt]]
 role="system"
 content=""

src/aios_kernel/agent.py

@@ -288,12 +288,13 @@ class AIAgent:
         result_len = 0
         for inner_func in all_inner_function:
             func_name = inner_func.get_name()
-            if self.enable_function_list:
+            if self.enable_function_list is not None:
                 if len(self.enable_function_list) > 0:
                     if func_name not in self.enable_function_list:
                         logger.debug(f"ageint {self.agent_id} ignore inner func:{func_name}")
                         continue
-
+                else:
+                    continue
             this_func = {}
             this_func["name"] = func_name
             this_func["description"] = inner_func.get_description()
@@ -322,6 +323,7 @@ class AIAgent:
             logger.error(f"llm execute inner func:{func_name} error:{e}")
 
 
+        logger.info("llm execute inner func result:" + result_str)
         inner_functions,inner_function_len = self._get_inner_functions()
         prompt.messages.append({"role":"function","content":result_str,"name":func_name})
         task_result:ComputeTaskResult = await ComputeKernel.get_instance().do_llm_completion(prompt,self.llm_model_name,self.max_token_size,inner_functions)

src/aios_kernel/bus.py

@@ -87,7 +87,7 @@ class AIBus:
 
             await asyncio.sleep(0.2)
             retry_times += 1
-            if retry_times > 5*120: # default timeout is 120 sec
+            if retry_times > 5*240: # default timeout is 240 sec
                 msg.status = AgentMsgStatus.ERROR
                 return None
 
@@ -107,12 +107,14 @@ class AIBus:
             # Wait for a message
             message = await handler.queue.get()
 
-            #try:
+            try:
                 # Try to handle the message
-            await handler.handle_message(message)
+                await handler.handle_message(message)
-            #except Exception as e:
+            except Exception as e:
                 # If an error occurs, put the message back into the queue
-            #    logger.error(f"handle message {message.msg_id} failed! {e}")
+                logger.error(f"handle message {message.msg_id} failed! {e}")
+                logger.exception(e)
+                raise e
                 #self.queues[name].put_nowait(message)
 
         return

src/aios_kernel/compute_kernel.py

@@ -118,7 +118,7 @@ class ComputeKernel:
                 if task_req.state == ComputeTaskState.ERROR:
                     break
 
-                if check_times >= 20:
+                if check_times >= 120:
                     task_req.state = ComputeTaskState.ERROR
                     break
 
@@ -129,7 +129,7 @@ class ComputeKernel:
         if task_req.state == ComputeTaskState.DONE:
             return task_req.result
 
-        return "error!"
+        raise Exception("error!")
 
     
     def text_embedding(self,input:str,model_name:Optional[str] = None):

src/aios_kernel/open_ai_node.py

@@ -100,16 +100,16 @@ class OpenAI_ComputeNode(ComputeNode):
                 if max_token_size is None:
                     max_token_size = 4000
 
-                result_token = int(max_token_size * 0.4)
+                result_token = max_token_size
-                    
-                logger.info(f"call openai {mode_name} prompts: {prompts}")
 
                 if llm_inner_functions is None:
+                    logger.info(f"call openai {mode_name} prompts: {prompts}")
                     resp = openai.ChatCompletion.create(model=mode_name,
                                                     messages=prompts,
                                                     max_tokens=result_token,
                                                     temperature=0.7)
                 else:
+                    logger.info(f"call openai {mode_name} prompts: {prompts} functions: {json.dumps(llm_inner_functions)}")
                     resp = openai.ChatCompletion.create(model=mode_name,
                                                         messages=prompts,
                                                         functions=llm_inner_functions,
@@ -139,6 +139,7 @@ class OpenAI_ComputeNode(ComputeNode):
                 result.result_message = resp["choices"][0]["message"]
                 if token_usage:
                     result.result_refers["token_usage"] = token_usage
+                logger.info(f"openai success response: {result.result_str}")
                 return result
             case _:
                 task.state = ComputeTaskState.ERROR

src/aios_kernel/text_to_speech_function.py

@@ -83,9 +83,9 @@ class TextToSpeechFunction(AIFunction):
                     continue
 
         if audio is not None:
-            path = os.path.join(os.curdir, "{}.mp3".format(random.sample('zyxwvutsrqponmlkjihgfedcba', 10)))
+            path = os.path.join(os.path.realpath(os.curdir), "{}.mp3".format(''.join(random.sample('zyxwvutsrqponmlkjihgfedcba', 10))))
             audio.export(path, format="mp3")
-            return "complete.file path:{}".format(path)
+            return "已经生成音频文件, 文件路径为{}".format(path)
         else:
             return "failed"
 

src/aios_kernel/workflow.py

@@ -376,11 +376,13 @@ class Workflow:
         result_func = []
         for inner_func in all_inner_function:
             func_name = inner_func.get_name()
-            if the_role.enable_function_list:
+            if the_role.enable_function_list is not None:
                 if len(the_role.enable_function_list) > 0:
                     if func_name not in the_role.enable_function_list:
-                        logger.debug(f"ageint {self.agent_id} ignore inner func:{func_name}")
+                        logger.debug(f"agent {self.agent_id} ignore inner func:{func_name}")
                         continue
+                else:
+                    continue
             this_func = {}
             this_func["name"] = func_name
             this_func["description"] = inner_func.get_description()
@@ -404,7 +406,7 @@ class Workflow:
 
         result_str:str = await func_node.execute(**arguments)
 
-        inner_functions = self._get_inner_functions()
+        inner_functions = self._get_inner_functions(the_role)
         prompt.messages.append({"role":"function","content":result_str,"name":func_name})
         task_result:ComputeTaskResult = await ComputeKernel.get_instance().do_llm_completion(prompt,
                                                                                 the_role.agent.llm_model_name,the_role.agent.max_token_size,