feat(core): Add stream data return and reception

1. add file: utils/steam_pipe.py
2. add demo: samples/flask_web_api.py
3. Other core code modifications, Add and use the StreamPipe class at night
4. Add flask library to requirements

metagpt/actions/action.py

@@ -23,6 +23,7 @@ from metagpt.schema import (
     TestingContext,
 )
 from metagpt.utils.project_repo import ProjectRepo
+from metagpt.utils.stream_pipe import StreamPipe
 
 
 class Action(SerializationMixin, ContextMixin, BaseModel):
@@ -35,6 +36,7 @@ class Action(SerializationMixin, ContextMixin, BaseModel):
     prefix: str = ""  # aask*时会加上prefix，作为system_message
     desc: str = ""  # for skill manager
     node: ActionNode = Field(default=None, exclude=True)
+    stream_pipe: Optional[StreamPipe] = None
 
     @property
     def repo(self) -> ProjectRepo:
@@ -90,6 +92,8 @@ class Action(SerializationMixin, ContextMixin, BaseModel):
 
     async def _aask(self, prompt: str, system_msgs: Optional[list[str]] = None) -> str:
         """Append default prefix"""
+        if self.stream_pipe and not self.llm.stream_pipe:
+            self.llm.stream_pipe = self.stream_pipe
         return await self.llm.aask(prompt, system_msgs)
 
     async def _run_action_node(self, *args, **kwargs):

metagpt/provider/anthropic_api.py

@@ -62,6 +62,9 @@ class AnthropicLLM(BaseLLM):
             elif event_type == "content_block_delta":
                 content = event.delta.text
                 log_llm_stream(content)
+                if self.stream_pipe:
+                    self.stream_pipe.set_message(content)
+
                 collected_content.append(content)
             elif event_type == "message_delta":
                 usage.output_tokens = event.usage.output_tokens  # update final output_tokens

metagpt/provider/base_llm.py

@@ -28,6 +28,7 @@ from metagpt.logs import logger
 from metagpt.schema import Message
 from metagpt.utils.common import log_and_reraise
 from metagpt.utils.cost_manager import CostManager, Costs
+from metagpt.utils.stream_pipe import StreamPipe
 
 
 class BaseLLM(ABC):
@@ -42,6 +43,7 @@ class BaseLLM(ABC):
     cost_manager: Optional[CostManager] = None
     model: Optional[str] = None  # deprecated
     pricing_plan: Optional[str] = None
+    stream_pipe: Optional[StreamPipe] = None
 
     @abstractmethod
     def __init__(self, config: LLMConfig):

metagpt/provider/dashscope_api.py

@@ -221,6 +221,8 @@ class DashScopeLLM(BaseLLM):
             content = chunk.output.choices[0]["message"]["content"]
             usage = dict(chunk.usage)  # each chunk has usage
             log_llm_stream(content)
+            if self.stream_pipe:
+                self.stream_pipe.set_message(content)
             collected_content.append(content)
         log_llm_stream("\n")
         self._update_costs(usage)

metagpt/provider/google_gemini_api.py

@@ -149,6 +149,8 @@ class GeminiLLM(BaseLLM):
                 logger.warning(f"messages: {messages}\nerrors: {e}\n{BlockedPromptException(str(chunk))}")
                 raise BlockedPromptException(str(chunk))
             log_llm_stream(content)
+            if self.stream_pipe:
+                self.stream_pipe.set_message(content)
             collected_content.append(content)
         log_llm_stream("\n")
 

metagpt/provider/ollama_api.py

@@ -83,6 +83,8 @@ class OllamaLLM(BaseLLM):
                 content = self.get_choice_text(chunk)
                 collected_content.append(content)
                 log_llm_stream(content)
+                if self.stream_pipe:
+                    self.stream_pipe.set_message(content)
             else:
                 # stream finished
                 usage = self.get_usage(chunk)

metagpt/provider/openai_api.py

@@ -87,6 +87,9 @@ class OpenAILLM(BaseLLM):
                 chunk.choices[0].finish_reason if chunk.choices and hasattr(chunk.choices[0], "finish_reason") else None
             )
             log_llm_stream(chunk_message)
+            if self.stream_pipe:
+                self.stream_pipe.set_message(chunk_message)
+
             collected_messages.append(chunk_message)
             if finish_reason:
                 if hasattr(chunk, "usage"):

metagpt/provider/qianfan_api.py

@@ -124,6 +124,8 @@ class QianFanLLM(BaseLLM):
             content = chunk.body.get("result", "")
             usage = chunk.body.get("usage", {})
             log_llm_stream(content)
+            if self.stream_pipe:
+                self.stream_pipe.set_message(content)
             collected_content.append(content)
         log_llm_stream("\n")
 

metagpt/provider/zhipuai_api.py

@@ -73,6 +73,8 @@ class ZhiPuAILLM(BaseLLM):
                 content = self.get_choice_delta_text(chunk)
                 collected_content.append(content)
                 log_llm_stream(content)
+                if self.stream_pipe:
+                    self.stream_pipe.set_message(content)
 
         log_llm_stream("\n")
 

metagpt/roles/role.py

@@ -39,6 +39,8 @@ from metagpt.strategy.planner import Planner
 from metagpt.utils.common import any_to_name, any_to_str, role_raise_decorator
 from metagpt.utils.project_repo import ProjectRepo
 from metagpt.utils.repair_llm_raw_output import extract_state_value_from_output
+from metagpt.utils.stream_pipe import StreamPipe
+
 
 if TYPE_CHECKING:
     from metagpt.environment import Environment  # noqa: F401
@@ -139,6 +141,8 @@ class Role(SerializationMixin, ContextMixin, BaseModel):
     role_id: str = ""
     states: list[str] = []
 
+    stream_pipe: Optional[StreamPipe] = None
+
     # scenarios to set action system_prompt:
     #   1. `__init__` while using Role(actions=[...])
     #   2. add action to role while using `role.set_action(action)`

metagpt/roles/tutorial_assistant.py

@@ -10,7 +10,7 @@ from datetime import datetime
 from typing import Dict
 
 from metagpt.actions.write_tutorial import WriteContent, WriteDirectory
-from metagpt.const import TUTORIAL_PATH
+from metagpt.const import TUTORIAL_PATH, METAGPT_ROOT
 from metagpt.logs import logger
 from metagpt.roles.role import Role, RoleReactMode
 from metagpt.schema import Message
@@ -40,7 +40,7 @@ class TutorialAssistant(Role):
 
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
-        self.set_actions([WriteDirectory(language=self.language)])
+        self.set_actions([WriteDirectory(language=self.language, stream_pipe=self.stream_pipe)])
         self._set_react_mode(react_mode=RoleReactMode.BY_ORDER.value)
 
     async def _handle_directory(self, titles: Dict) -> Message:
@@ -58,7 +58,7 @@ class TutorialAssistant(Role):
         self.total_content += f"# {self.main_title}"
         actions = list()
         for first_dir in titles.get("directory"):
-            actions.append(WriteContent(language=self.language, directory=first_dir))
+            actions.append(WriteContent(language=self.language, directory=first_dir, stream_pipe=self.stream_pipe))
             key = list(first_dir.keys())[0]
             directory += f"- {key}\n"
             for second_dir in first_dir[key]:
@@ -91,4 +91,8 @@ class TutorialAssistant(Role):
         root_path = TUTORIAL_PATH / datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
         await File.write(root_path, f"{self.main_title}.md", self.total_content.encode("utf-8"))
         msg.content = str(root_path / f"{self.main_title}.md")
+
+        if self.stream_pipe:
+            self.stream_pipe.set_k_message("file_name", msg.content.replace(str(METAGPT_ROOT), ""))
+            self.stream_pipe.with_finish()
         return msg

metagpt/utils/stream_pipe.py

@@ -0,0 +1,70 @@
+# -*- coding: utf-8 -*-
+# @Time    : 2024/3/27 10:00
+# @Author  : leiwu30 
+# @File    : stream_pipe.py
+# @Version : None
+# @Description : None
+
+import time
+import json
+from multiprocessing import Pipe
+
+
+class StreamPipe:
+    parent_conn, child_conn = Pipe()
+
+    variable: list = {}
+    finish: bool = False
+
+    format_data = {
+        "id": "chatcmpl-96bVnBOOyPFZZxEoTIGbdpFcVEnur",
+        "object": "chat.completion.chunk",
+        "created": 1711361191,
+        "model": "gpt-3.5-turbo-0125",
+        "system_fingerprint": "fp_3bc1b5746c",
+        "choices": [
+            {
+                "index": 0,
+                "delta":
+                    {
+                        "role": "assistant",
+                        "content": "content"
+                    },
+                "logprobs": None,
+             "finish_reason": None
+            }
+        ]
+    }
+
+    def set_message(self, msg):
+        self.parent_conn.send(msg)
+
+    def wait(self):
+        pass
+
+    def get_message(self):
+        if self.child_conn.poll(timeout=3):
+            return self.child_conn.recv()
+        else:
+            return None
+
+    def set_k_message(self, k, msg):
+        self.variable[k] = msg
+
+    def get_k_message(self, k):
+        return self.variable[k]
+
+    def msg2stream(self, msg):
+        self.format_data['created'] = int(time.time())
+        self.format_data['choices'][0]['delta']['content'] = msg
+        return f"data: {json.dumps(self.format_data, ensure_ascii=False)}\n".encode("utf-8")
+
+    def with_finish(self, timeout: int = 3):
+        """
+        Args:
+            timeout: /s
+        """
+        # Pipe is not empty waiting for pipe condition
+        # while self.child_conn.poll(timeout=timeout):
+        #     time.sleep(0.5)
+        self.finish = True