Merge pull request #408 from stellaHSR/minecraft

Minecraft: bug fix for code save and skill add.

Temp.md

@@ -31,6 +31,8 @@ ### 0926： 环境信息获取和更新 on_event()实际内容
     mc_player.set_port(2465) # Modify this to your LAN port
 ```
 
+*注意世界游戏模式为生存，这样人物才会有血量，饥饿和库存概念；*
+
 python minecraft_run.py
 
 <img src="docs/resources/workspace/minecraft_tests/on_event.jpeg" style="zoom:67%;" />
@@ -67,3 +69,6 @@ ### 0930：Curriculum agent 更新
 1. 若本地已克隆项目不好更改可尝试：删除 metagpt/mineflayer_env/mineflayer + 重新copy voyager/env/mineflayer到目录下 + （npm install...0926.B命令）
 2. 重新拉取最新提交+重新配置
 
+### 第一人称视角
+
+目前使用3007端口，可自行修改，代码运行以后，在浏览器输入127.0.0.1:3007访问

config/config.yaml

@@ -84,4 +84,7 @@ MODEL_FOR_RESEARCHER_REPORT: gpt-3.5-turbo-16k
 ### browser path for pyppeteer engine, support Chrome, Chromium,MS Edge
 #PYPPETEER_EXECUTABLE_PATH: "/usr/bin/google-chrome-stable"
 
-PROMPT_FORMAT: json #json or markdown
+PROMPT_FORMAT: json #json or markdown
+
+### Minecraft
+RESUME: false # If load json from ckpt dir(include chest_memory, skills, ...)

metagpt/actions/minecraft/design_curriculumn.py

@@ -91,44 +91,46 @@ class DesignCurriculum(Action):
 
     def __init__(self, name="", context=None, llm=None):
         super().__init__(name, context, llm)
-        # voyager vectordb using
 
-    @classmethod
-    def generate_qa(cls, events, chest_observation):
+    async def generate_qa(self, events, qa_cache, human_msg, system_msg):
         """
         Generate qa for DesignTask's HumanMessage
         """
-        questions_new, _ = cls.generate_qa_step1(
-            events=events, chest_observation=chest_observation
+        questions_new = await self.generate_qa_step1(
+            events=events, human_msg=human_msg, system_msg=system_msg
         )
+        logger.debug(f"Generate_qa_step1 result list is HERE: {questions_new}")
+        
         questions = []
         answers = []
         for question in questions_new:
-            if cls.qa_cache_questions_vectordb._collection.count() > 0:
+            if self.qa_cache_questions_vectordb._collection.count() > 0:
                 docs_and_scores = (
-                    cls.qa_cache_questions_vectordb.similarity_search_with_score(
+                    self.qa_cache_questions_vectordb.similarity_search_with_score(
                         question, k=1
                     )
                 )
                 if docs_and_scores and docs_and_scores[0][1] < 0.05:
                     question_cached = docs_and_scores[0][0].page_content
-                    assert question_cached in cls.qa_cache
-                    answer_cached = cls.qa_cache[question_cached]
+                    assert question_cached in qa_cache
+                    answer_cached = qa_cache[question_cached]
                     questions.append(question_cached)
                     answers.append(answer_cached)
                     continue
-            answer = cls.generate_qa_step2(question=question)
-            assert question not in cls.qa_cache
-            cls.qa_cache[question] = answer
-            cls.qa_cache_questions_vectordb.add_texts(
+            answer = await self.generate_qa_step2(question=question)
+            assert question not in qa_cache
+            qa_cache[question] = answer
+            self.qa_cache_questions_vectordb.add_texts(
                 texts=[question],
             )
             with open(f"{CKPT_DIR}/curriculum/qa_cache.json", "w") as f:
-                json.dump(cls.qa_cache, f)
-            cls.qa_cache_questions_vectordb.persist()
+                json.dump(qa_cache, f)
+            self.qa_cache_questions_vectordb.persist()
             questions.append(question)
             answers.append(answer)
         assert len(questions_new) == len(questions) == len(answers)
+        logger.info(f"Curriculum Agent generate_qa Questions: {questions}")
+        logger.info(f"Curriculum Agent generate_qa Answers: {answers}")
         return questions, answers
 
     async def generate_qa_step1(self, events, human_msg, system_msg):
@@ -157,7 +159,7 @@ class DesignCurriculum(Action):
 
     async def generate_qa_step2(self, question):
         # Implement the logic for another specific step in generating questions and answers.
-        logger.info(f"Curriculum Agent Question: {question}")
+        # logger.info(f"Curriculum Agent generate_qa_step2 Question: {question}")
         human_msg = HumanMessage(content=f"Question: {question}").content
         system_msg = [
             SystemMessage(
@@ -165,10 +167,10 @@ class DesignCurriculum(Action):
             ).content
         ]
         answer = await self._aask(prompt=human_msg, system_msgs=system_msg)
-        logger.info(f"Curriculum Agent {answer}")
+        # logger.info(f"Curriculum Agent generate_qa_step2 answer: {answer}")
         return answer
 
-    async def get_context_from_task(self, task):
+    async def get_context_from_task(self, task, qa_cache):
         """
         Args: task
         Returns: context: "Question: {question}\n{answer}"
@@ -179,21 +181,21 @@ class DesignCurriculum(Action):
             f"How to {task.replace('_', ' ').replace(' ore', '').replace(' ores', '').replace('.', '').strip().lower()}"
             f" in Minecraft?"
         )
-        if question in self.qa_cache:
-            answer = self.qa_cache[question]
+        if question in qa_cache:
+            answer = qa_cache[question]
         else:
             answer = await self.generate_qa_step2(question=question)
-            self.qa_cache[question] = answer
+            qa_cache[question] = answer
             self.qa_cache_questions_vectordb.add_texts(
                 texts=[question],
             )
             with open(f"{CKPT_DIR}/curriculum/qa_cache.json", "w") as f:
-                json.dump(self.qa_cache, f)
+                json.dump(qa_cache, f)
             self.qa_cache_questions_vectordb.persist()
         context = f"Question: {question}\n{answer}"
         return context
 
-    async def generate_context(self, task, max_retries=5):
+    async def generate_context(self, task, qa_cache, max_retries=5):
         """
         Refer to the code in the voyager/agents/curriculum.py propose_next_ai_task() for implementation details.
         Returns: context
@@ -204,21 +206,22 @@ class DesignCurriculum(Action):
             raise RuntimeError("Max retries reached, failed to propose context.")
         try:
             context = await self.get_context_from_task(
-                task=task
+                task=task, qa_cache=qa_cache
             )  # Curriculum Agent Question: How to craft 4 wooden planks in Minecraft? & Curriculum Agent Answer: ...
             return context
         except Exception as e:
             logger.info(f"Error parsing curriculum response: {e}. Trying again!")
             return await self.generate_context(
                 task=task,
+                qa_cache=qa_cache,
                 max_retries=max_retries - 1,
             )
 
-    async def run(self, task, human_msg, system_msg, *args, **kwargs):
+    async def run(self, task, qa_cache, human_msg, system_msg, *args, **kwargs):
         logger.info(f"run {self.__repr__()}")
         # Generate curriculum-related questions and answers.
         # curriculum_qustion = await self.generate_qa_step1(events, human_msg, system_msg)
-        curriculum_context = await self.generate_context(task)
+        curriculum_context = await self.generate_context(task, qa_cache)
 
         # Return the generated questions and answers.
         return curriculum_context

metagpt/actions/minecraft/generate_actions.py

@@ -30,19 +30,20 @@ class GenerateActionCode(Action):
 
         try:
             return (
+                parsed_result["program_code"],
                 parsed_result["program_code"] + "\n" + parsed_result["exec_code"],
                 parsed_result["program_name"],
             )
         except:
             logger.error(f"Failed to parse response: {parsed_result}")
-            return None, None
+            return "", "", ""
 
     async def run(self, human_msg, system_msg, *args, **kwargs):
         logger.info(f"run {self.__repr__()}")
         # Generate action code.
-        generated_code, program_name = await self.generate_code(
+        program_code, generated_code, program_name = await self.generate_code(
             human_msg=human_msg, system_msg=system_msg
         )
 
         # Return the generated code.
-        return generated_code, program_name
+        return program_code, generated_code, program_name

metagpt/actions/minecraft/manage_skills.py

@@ -5,6 +5,8 @@
 import os
 import json
 
+from traits.trait_types import self
+
 from metagpt.logs import logger
 from metagpt.actions.minecraft.player_action import PlayerActions as Action
 from metagpt.const import CKPT_DIR
@@ -15,11 +17,11 @@ class RetrieveSkills(Action):
     Action class for retrieving skills.
     Refer to the code in the voyager/agents/skill.py for implementation details.
     """
-
+    
     def __init__(self, name="", context=None, llm=None):
         super().__init__(name, context, llm)
         self.llm.model = "gpt-3.5-turbo"
-
+    
     async def run(self, query, skills, *args, **kwargs):
         # Implement the logic for retrieving skills here.
         k = min(self.vectordb._collection.count(), self.retrieval_top_k)
@@ -42,13 +44,13 @@ class AddNewSkills(Action):
     Action class for adding new skills.
     Refer to the code in the voyager/agents/skill.py for implementation details.
     """
-
+    
     def __init__(self, name="", context=None, llm=None):
         super().__init__(name, context, llm)
         self.llm.model = "gpt-3.5-turbo"
-
+    
     async def run(
-        self, task, program_name, program_code, skills, skill_desp, *args, **kwargs
+            self, task, program_name, program_code, skills, skill_desp, *args, **kwargs
     ):
         # Implement the logic for adding new skills here.
         # TODO: Fix this
@@ -75,12 +77,7 @@ class AddNewSkills(Action):
             ids=[program_name],
             metadatas=[{"name": program_name}],
         )
-
-        # FIXME
-        # assert self.vectordb._collection.count() == len(
-        #     skills
-        # ), "vectordb is not synced with skills.json"
-
+        
         with open(f"{CKPT_DIR}/skill/code/{dumped_program_name}.js", "w") as f:
             f.write(program_code)
         with open(f"{CKPT_DIR}/skill/description/{dumped_program_name}.txt", "w") as f:
@@ -99,13 +96,13 @@ class GenerateSkillDescription(Action):
     Action class for generating skill descriptions.
     Refer to the code in the voyager/agents/skill.py for implementation details.
     """
-
+    
     def __init__(self, name="", context=None, llm=None):
         super().__init__(name, context, llm)
         self.llm.model = "gpt-3.5-turbo"
-
+    
     async def run(self, program_name, human_message, system_message, *args, **kwargs):
         # Implement the logic for generating skill descriptions here.
         rsp = await self._aask(prompt=human_message, system_msgs=system_message)
-        skill_description = f"    // { rsp}"
+        skill_description = f"    // {rsp}"
         return f"async function {program_name}(bot) {{\n{skill_description}\n}}"

metagpt/actions/minecraft/player_action.py

@@ -2,58 +2,69 @@
 # @Date    : 2023/9/23 17:06
 # @Author  : stellahong (stellahong@fuzhi.ai)
 # @Desc    :
+import json
+
 from metagpt.actions import Action
 from langchain.vectorstores import Chroma
 from langchain.embeddings.openai import OpenAIEmbeddings
-from metagpt.document_store import FaissStore
 from metagpt.const import CKPT_DIR
+from metagpt.config import CONFIG
+
 
 class PlayerActions(Action):
     def __init__(self, name="", context=None, llm=None):
         super().__init__(name, context, llm)
-        self.skills = {}
-        self.qa_cache = {}
+        self.skills_check = {}  # for skills.json
+        self.qa_cache_check = {}
+        self.check_init = True
+
+        if (
+            CONFIG.resume
+        ):  # TODO: now for assert only, no update, cuz program using in step()
+            with open(f"{CKPT_DIR}/skill/skills.json", "r") as f:
+                self.skills_check = json.load(f)
+
+            with open(f"{CKPT_DIR}/curriculum/qa_cache.json", "r") as f:
+                self.qa_cache_check = json.load(f)
+
         self.retrieval_top_k = 5
+
         self.vectordb = Chroma(
             collection_name="skill_vectordb",
             embedding_function=OpenAIEmbeddings(),
             persist_directory=f"{CKPT_DIR}/skill/vectordb",
         )
-        
+
         self.qa_cache_questions_vectordb = Chroma(
             collection_name="qa_cache_questions_vectordb",
             embedding_function=OpenAIEmbeddings(),
             persist_directory=f"{CKPT_DIR}/curriculum/vectordb",
         )
-        # TODO: change to FaissStore
-        # self.qa_cache_questions_vectordb = FaissStore( {CKPT_DIR}/ 'curriculum/vectordb'
 
-    @classmethod
-    def set_skills(cls, skills):
-        cls.skills = skills
-        # Check if Skill Manager's vectordb right using
-        assert cls.vectordb._collection.count() == len(cls.skills), (
-            f"Skill Manager's vectordb is not synced with skills.json.\n"
-            f"There are {cls.vectordb._collection.count()} skills in vectordb but {len(cls.skills)} skills in skills.json.\n"
-            f"Did you set resume=False when initializing the manager?\n"
-            f"You may need to manually delete the vectordb directory for running from scratch."
-        )
+        # FIXME
+        if self.check_init:
+            # Check if Skill Manager's vectordb right using
+            assert self.vectordb._collection.count() >= len(self.skills_check), (
+                f"Skill Manager's vectordb is not synced with skills.json.\n"
+                f"There are {self.vectordb._collection.count()} skills in vectordb but {len(self.skills_check)} skills in skills.json.\n"
+                f"Did you set resume=False when initializing the manager?\n"
+                f"You may need to manually delete the vectordb directory for running from scratch."
+            )
+            # Check if Skill Manager's vectordb right using
+            assert self.qa_cache_questions_vectordb._collection.count() >= len(
+                self.qa_cache_check
+            ), (
+                f"Curriculum Agent's qa cache question vectordb is not synced with qa_cache.json.\n"
+                f"There are {self.qa_cache_questions_vectordb._collection.count()} questions in vectordb "
+                f"but {len(self.qa_cache_check)} questions in qa_cache.json.\n"
+                f"Did you set resume=False when initializing the agent?\n"
+                f"You may need to manually delete the qa cache question vectordb directory for running from scratch.\n"
+            )
+            self.check_init = False
+            # TODO: change to FaissStore
+            # self.qa_cache_questions_vectordb = FaissStore( {CKPT_DIR}/ 'curriculum/vectordb'
 
-    @classmethod
-    def set_qa_cache(cls, qa_cache):
-        cls.qa_cache = qa_cache
-        # Check if qa_cache right using
-        # Check if Skill Manager's vectordb right using
-        assert cls.qa_cache_questions_vectordb._collection.count() == len(
-            cls.qa_cache
-        ), (
-            f"Curriculum Agent's qa cache question vectordb is not synced with qa_cache.json.\n"
-            f"There are {cls.qa_cache_questions_vectordb._collection.count()} questions in vectordb "
-            f"but {len(cls.qa_cache)} questions in qa_cache.json.\n"
-            f"Did you set resume=False when initializing the agent?\n"
-            f"You may need to manually delete the qa cache question vectordb directory for running from scratch.\n"
-        )
-    
     """Minecraft player info without any implementation details"""
+
     async def run(self, *args, **kwargs):
-        raise NotImplementedError
+        raise NotImplementedError

metagpt/config.py

@@ -87,6 +87,7 @@ class Config(metaclass=Singleton):
         self.pyppeteer_executable_path = self._get("PYPPETEER_EXECUTABLE_PATH", "")
 
         self.prompt_format = self._get("PROMPT_FORMAT", "markdown")
+        self.resume = self._get("RESUME", False)
 
     def _init_with_config_files_and_env(self, configs: dict, yaml_file):
         """Load from config/key.yaml, config/config.yaml, and env in decreasing order of priority"""

metagpt/minecraft_team.py

@@ -7,6 +7,11 @@ from pydantic import BaseModel, Field
 import requests
 import json
 import re
+import time
+
+from langchain.vectorstores import Chroma
+from langchain.embeddings.openai import OpenAIEmbeddings
+from metagpt.document_store import FaissStore
 
 from metagpt.logs import logger
 from metagpt.roles import Role
@@ -18,6 +23,7 @@ from metagpt.roles.minecraft.minecraft_base import Minecraft
 from metagpt.environment import Environment
 from metagpt.mineflayer_environment import MineflayerEnv
 from metagpt.const import CKPT_DIR
+from metagpt.config import CONFIG
 from metagpt.actions.minecraft.control_primitives import load_skills_code
 
 
@@ -33,6 +39,7 @@ class GameEnvironment(BaseModel, arbitrary_types_allowed=True):
         default="You can mine one of oak, birch, spruce, jungle, acacia, dark oak, or mangrove logs."
     )
     code: str = Field(default="")
+    program_code: str = Field(default="") # write in skill/code/*.js
     program_name: str = Field(default="")
     critique: str = Field(default="")
     skills: dict = Field(default_factory=dict)  # for skills.json
@@ -76,12 +83,13 @@ class GameEnvironment(BaseModel, arbitrary_types_allowed=True):
     @property
     def core_inv_items_regex(self):
         return self.mf_instance.core_inv_items_regex
-    
+
     def set_mc_port(self, mc_port):
         self.mf_instance.set_mc_port(mc_port)
+        self.set_mc_resume()
     
-    def set_mc_resume(self, resume: bool = False):  # TODO: mv to config
-        if resume:
+    def set_mc_resume(self):
+        if CONFIG.resume:
             logger.info(f"Loading Action Developer from {CKPT_DIR}/action")
             with open(f"{CKPT_DIR}/action/chest_memory.json", "r") as f:
                 self.chest_memory = json.load(f)
@@ -91,12 +99,14 @@ class GameEnvironment(BaseModel, arbitrary_types_allowed=True):
                 self.completed_tasks = json.load(f)
             with open(f"{CKPT_DIR}/curriculum/failed_tasks.json", "r") as f:
                 self.failed_tasks = json.load(f)
-            with open(f"{CKPT_DIR}/curriculum/qa_cache.json", "r") as f:
-                self.qa_cache = json.load(f)
-            
+
             logger.info(f"Loading Skill Manager from {CKPT_DIR}/skill\033[0m")
             with open(f"{CKPT_DIR}/skill/skills.json", "r") as f:
                 self.skills = json.load(f)
+            
+            logger.info(f"Loading Qa Cache from {CKPT_DIR}/curriculum\033[0m")
+            with open(f"{CKPT_DIR}/curriculum/qa_cache.json", "r") as f:
+                self.qa_cache = json.load(f) 
     
     def register_roles(self, roles: Iterable[Minecraft]):
         for role in roles:
@@ -114,7 +124,10 @@ class GameEnvironment(BaseModel, arbitrary_types_allowed=True):
     
     def update_context(self, context: str):
         self.context = context
-    
+
+    def update_program_code(self, program_code: str):
+        self.program_code = program_code
+
     def update_code(self, code: str):
         self.code = code  # action_developer.gen_action_code to HERE
     
@@ -292,8 +305,19 @@ class GameEnvironment(BaseModel, arbitrary_types_allowed=True):
             self.update_event(events)
             return events
         except Exception as e:
+            time.sleep(3)  # wait for mineflayer to exit
+            # reset bot status here
+            events = self.mf_instance.reset(
+                options={
+                    "mode": "hard",
+                    "wait_ticks": 20,
+                    "inventory": self.event[-1][1]["inventory"],
+                    "equipment": self.event[-1][1]["status"]["equipment"],
+                    "position": self.event[-1][1]["status"]["position"],
+                }
+            )
+            self.update_event(events)
             logger.error(f"Failed to retrieve Minecraft events: {str(e)}")
-            raise {}
     
     async def on_event_execute(self, *args):
         """
@@ -316,8 +340,19 @@ class GameEnvironment(BaseModel, arbitrary_types_allowed=True):
             self.update_event(events)
             return events
         except Exception as e:
+            time.sleep(3)  # wait for mineflayer to exit
+            # reset bot status here
+            events = self.mf_instance.reset(
+                options={
+                    "mode": "hard",
+                    "wait_ticks": 20,
+                    "inventory": self.event[-1][1]["inventory"],
+                    "equipment": self.event[-1][1]["status"]["equipment"],
+                    "position": self.event[-1][1]["status"]["position"],
+                }
+            )
+            self.update_event(events)
             logger.error(f"Failed to execute Minecraft events: {str(e)}")
-            raise {}
 
 
 class MinecraftPlayer(SoftwareCompany):
@@ -335,9 +370,6 @@ class MinecraftPlayer(SoftwareCompany):
     def set_port(self, mc_port):
         self.game_memory.set_mc_port(mc_port)
     
-    def set_resume(self, resume: bool = False):
-        self.game_memory.set_mc_resume(resume=resume)
-    
     def check_complete_round(self):
         complete_round = []
         for role in self.environment.roles.values():
@@ -380,12 +412,23 @@ class MinecraftPlayer(SoftwareCompany):
     async def run(self, n_round=3):
         """Run company until target round or no money"""
         round_id = 0
-        self.game_memory.mf_instance.reset(
-            options={
-                "mode": "soft",
-                "wait_ticks": 20,
-            }
-        )
+        if CONFIG.resume:
+            # keep the inventory
+            self.game_memory.mf_instance.reset(
+                options={
+                    "mode": "soft",
+                    "wait_ticks": 20,
+                }
+            )
+        else:
+            # clear the inventory
+            self.game_memory.mf_instance.reset(
+                options={
+                    "mode": "hard",
+                    "wait_ticks": 20,
+                }
+            )
+
         events = self.game_memory.mf_instance.step(
             code="",
             programs="",

metagpt/mineflayer_env/mineflayer/index.js

@@ -14,6 +14,7 @@ const Inventory = require("./lib/observation/inventory");
 const OnSave = require("./lib/observation/onSave");
 const Chests = require("./lib/observation/chests");
 const { plugin: tool } = require("mineflayer-tool");
+const { mineflayer: mineflayerViewer } = require('prismarine-viewer')
 
 let bot = null;
 
@@ -50,6 +51,7 @@ app.post("/start", (req, res) => {
     });
 
     bot.once("spawn", async () => {
+        mineflayerViewer(bot, { port: 3007, firstPerson: false }) 
         bot.removeListener("error", onConnectionFailed);
         let itemTicks = 1;
         if (req.body.reset === "hard") {
@@ -222,7 +224,7 @@ app.post("/step", async (req, res) => {
         }
     }
 
-    bot.on("physicTick", onTick);
+    bot.on("physicsTick", onTick);
 
     // initialize fail count
     let _craftItemFailCount = 0;
@@ -248,7 +250,7 @@ app.post("/step", async (req, res) => {
         response_sent = true;
         res.json(bot.observe());
     }
-    bot.removeListener("physicTick", onTick);
+    bot.removeListener("physicsTick", onTick);
 
     async function evaluateCode(code, programs) {
         // Echo the code produced for players to see it. Don't echo when the bot code is already producing dialog or it will double echo

metagpt/roles/minecraft/action_developer.py

@@ -2,8 +2,6 @@
 # @Date    : 2023/9/23 12:45
 # @Author  : stellahong (stellahong@fuzhi.ai)
 # @Desc    :
-import copy
-
 from metagpt.logs import logger
 from metagpt.roles.minecraft.minecraft_base import Minecraft as Base
 from metagpt.schema import Message, HumanMessage, SystemMessage
@@ -40,14 +38,13 @@ class ActionDeveloper(Base):
         # Initialize actions specific to the Action role
         self._init_actions([GenerateActionCode])
         
-        
         # Set events or actions the ActionAgent should watch or be aware of
         # 需要根据events进行自己chest_observation的更新
         self._watch([RetrieveSkills])
         self.rollout_num_iter = 0
         self.task_max_retries = 4
         self.finish_state = len(self._actions)
-        self.critic_reviewer = None  # self._rc.env.roles["Task Reviewer"]
+        self.critic_reviewer = None  # self._rc.env.roles["Task Reviewer"]      
     
     def render_system_message(self, skills=[], *args, **kwargs):
         """
@@ -190,6 +187,8 @@ class ActionDeveloper(Base):
         return len(self._rc.news)
     
     async def run_step(self, human_msg, system_msg, *args, **kwargs):
+        await self._obtain_events()
+        logger.info("reset before step()!")
         while True:
             logger.info(f"self.rollout_num_iter {self.rollout_num_iter}")
             system_msg, human_msg, reward, done, info = await self.runcode_and_evaluate(human_msg, system_msg, *args,
@@ -208,6 +207,7 @@ class ActionDeveloper(Base):
     async def handle_add_new_skills(
             self, task, program_name, program_code, skills, *args, **kwargs
     ):
+        skills = self.game_memory.skills
         skill_desp = self.game_memory.skill_desp
         new_skills_info = await AddNewSkills().run(
             task, program_name, program_code, skills, skill_desp
@@ -231,11 +231,9 @@ class ActionDeveloper(Base):
         context = self.game_memory.context
         
         # 更新生成的代码和对应程序名称
-        code, program_name = await GenerateActionCode().run(
+        program_code, code, program_name = await GenerateActionCode().run(
             human_msg, system_msg, *args, **kwargs
         )
-        # logger.warning(type(code))
-        # logger.info(f"Code is Here:{code}")
         
         if code is not None:
             # fixme：若有独立的mc code执行入口函数，使用独立的函数
@@ -298,6 +296,8 @@ class ActionDeveloper(Base):
             "success": self.game_memory.runtime_status,
         }
         logger.info(f"info is {info}")
+        
+        self.perform_game_info_callback(program_code, self.game_memory.update_program_code)
         self.perform_game_info_callback(code, self.game_memory.update_code)
         self.perform_game_info_callback(
             program_name, self.game_memory.update_program_name
@@ -306,11 +306,10 @@ class ActionDeveloper(Base):
         return system_msg, human_msg, 0, done, info
     
     async def generate_action_code(self, human_msg, system_msg, *args, **kwargs):
-        code, program_name = await GenerateActionCode().run(
+        program_code, code, program_name = await GenerateActionCode().run(
             human_msg, system_msg, *args, **kwargs
         )
-        # logger.warning(type(code))
-        # logger.info(f"Code is Here:{code}")
+        self.perform_game_info_callback(program_code, self.game_memory.update_program_code)
         self.perform_game_info_callback(code, self.game_memory.update_code)
         self.perform_game_info_callback(
             program_name, self.game_memory.update_program_name

metagpt/roles/minecraft/critic_agent.py

@@ -132,7 +132,7 @@ class CriticReviewer(Base):
         }
 
     async def verify_task(self, human_msg, system_msg, *args, **kwargs):
-        success, critique = await VerifyTask().run(human_msg, system_msg, max_retries=5)
+        success, critique = await VerifyTask().run(human_msg, system_msg, max_retries=1)
         self.perform_game_info_callback(
             success, self.game_memory.update_exploration_progress
         )

metagpt/roles/minecraft/curriculum_agent.py

@@ -120,7 +120,7 @@ class CurriculumDesigner(Base):
         return observation
 
     # --------------------------------Design Task Prepare---------------------------------------
-    def render_design_task_human_message(
+    async def render_design_task_human_message(
         self, events, chest_observation, *args, **kwargs
     ):
         """
@@ -130,13 +130,20 @@ class CurriculumDesigner(Base):
 
         content = ""
         warm_up = self.game_memory.mf_instance.warm_up
+        qa_cache = self.game_memory.qa_cache
         observation = self.render_curriculum_observation(
             events=events, chest_observation=chest_observation
         )
         if self.game_memory.progress >= warm_up["context"]:
-            questions, answers = DesignCurriculum.generate_qa(
+        # if self.game_memory.progress >= 0: # TEST ONLY
+            human_msg = self.render_design_curriculum_human_message(
                 events=events, chest_observation=chest_observation
+            ).content
+            system_msg = [self.render_design_curriculum_system_message().content]
+            questions, answers = await DesignCurriculum().generate_qa(
+                events=events, qa_cache=qa_cache, human_msg=human_msg, system_msg=system_msg
             )
+            logger.debug(f"Generate_qa result is HERE: Ques: {questions}, Ans: {answers}")
             i = 1
             for question, answer in zip(questions, answers):
                 if "Answer: Unknown" in answer or "language model" in answer:
@@ -162,8 +169,8 @@ class CurriculumDesigner(Base):
     def render_design_task_system_message(self, *args, **kwargs):
         return SystemMessage(content=load_prompt("curriculum"))
 
-    def encapsule_design_task_message(self, events, chest_observation, *args, **kwargs):
-        human_msg = self.render_design_task_human_message(
+    async def encapsule_design_task_message(self, events, chest_observation, *args, **kwargs):
+        human_msg = await self.render_design_task_human_message(
             events=events, chest_observation=chest_observation, *args, **kwargs
         )
         system_msg = self.render_design_task_system_message(*args, **kwargs)
@@ -292,6 +299,7 @@ class CurriculumDesigner(Base):
         chest_observation = self.game_memory.chest_observation
         inventoryUsed = events[-1][1]["status"]["inventoryUsed"]
         task = self.game_memory.current_task
+        qa_cache = self.game_memory.qa_cache
 
         if self.game_memory.progress == 0:
             context = self.game_memory.context
@@ -301,7 +309,7 @@ class CurriculumDesigner(Base):
             )
         else:
             context = await DesignCurriculum().run(
-                task, human_msg, system_msg, *args, **kwargs
+                task, qa_cache, human_msg, system_msg, *args, **kwargs
             )
         self.perform_game_info_callback(context, self.game_memory.update_context)
         return Message(
@@ -313,18 +321,15 @@ class CurriculumDesigner(Base):
     async def _act(self) -> Message:
         todo = self._rc.todo
         logger.debug(f"Todo is {todo}")
-        self.maintain_actions(todo) 
+        self.maintain_actions(todo)
+
         # 获取最新的游戏周边环境信息
-        # events = await self._obtain_events()
-        events = self.game_memory.event
+        events = await self._obtain_events()
+        logger.info(f"Retrieve event is HERE: {events}, current task is {self.game_memory.current_task}")
+        # events = self.game_memory.event
         chest_observation = self.game_memory.chest_observation
 
-        # DesignCurriculum.set_qa_cache(self.game_memory.qa_cache)
-
-        # msg = self._rc.memory.get(k=1)[0]
-        # query = msg.content
-
-        design_task_message = self.encapsule_design_task_message(
+        design_task_message = await self.encapsule_design_task_message(
             events, chest_observation
         )
         design_curriculum_message = self.encapsule_design_curriculum_message(

metagpt/roles/minecraft/skill_manager.py

@@ -19,41 +19,40 @@ from metagpt.utils.minecraft import load_prompt
 @agent_registry.register("skill_manager")
 class SkillManager(Base):
     def __init__(
-        self,
-        name: str = "John",
-        profile: str = "Skills Management Specialist",
-        goal: str = "To oversee and optimize the acquisition, development, and utilization of skills within the organization, ensuring workforce competence and efficiency.",
-        constraints: str = "Resource allocation, training budgets, and alignment with organizational goals.",
+            self,
+            name: str = "John",
+            profile: str = "Skills Management Specialist",
+            goal: str = "To oversee and optimize the acquisition, development, and utilization of skills within the organization, ensuring workforce competence and efficiency.",
+            constraints: str = "Resource allocation, training budgets, and alignment with organizational goals.",
     ) -> None:
         super().__init__(name, profile, goal, constraints)
-
+        
         # Initialize actions specific to the SkillManager role
-        self._init_actions([RetrieveSkills, GenerateSkillDescription, AddNewSkills]) #AddNewSkills])#先去掉add
+        self._init_actions([RetrieveSkills, GenerateSkillDescription, AddNewSkills])
         
         # Set events or actions the SkillManager should watch or be aware of
         self._watch(
             [DesignCurriculum, GenerateActionCode, RetrieveSkills, GenerateSkillDescription]
         )
-        
         self.finish_state = len(self._actions)
-
+    
     def encapsule_message(self, program_code, program_name, *args, **kwargs):
         system_msg = self.render_system_message(load_prompt("skill"))
         human_msg = self.render_human_message(
             program_code + "\n\n" + f"The main function is `{program_name}`."
         )
         return {"system_msg": [system_msg.content], "human_msg": human_msg.content}
-
+    
     async def retrieve_skills(self, query, skills, *args, **kwargs):
         retrieve_skills = await RetrieveSkills().run(query, skills)
         logger.info(f"Render Action Agent system message with {len(retrieve_skills)} skills")
         self.perform_game_info_callback(retrieve_skills, self.game_memory.update_retrieve_skills)
-        return Message(content=f"{retrieve_skills}", instruct_content="retrieve_skills", 
+        return Message(content=f"{retrieve_skills}", instruct_content="retrieve_skills",
                        role=self.profile, send_to=agent_registry.entries["action_developer"]()._setting.name)
         # return Message(
         #     content=f"{skills}", instruct_content="retrieve_skills", role=self.profile
         # )  # Unit test only
-
+    
     async def generate_skill_descp(self, human_msg, system_msg, *args, **kwargs):
         program_name = self.game_memory.program_name
         desp = await GenerateSkillDescription().run(program_name, human_msg, system_msg)
@@ -63,16 +62,16 @@ class SkillManager(Base):
             instruct_content="generate_skill_descp",
             role=self.profile,
         )
-
+    
     async def handle_add_new_skills(
-        self, task, program_name, program_code, skills, *args, **kwargs
+            self, task, program_name, program_code, skills, *args, **kwargs
     ):
         if not self.game_memory.runtime_status:
             return Message(
-            content="",
-            instruct_content="handle_add_new_skills",
-            role=self.profile,
-        )
+                content="",
+                instruct_content="handle_add_new_skills",
+                role=self.profile,
+            )
         
         skill_desp = self.game_memory.skill_desp
         new_skills_info = await AddNewSkills().run(
@@ -84,43 +83,41 @@ class SkillManager(Base):
             instruct_content="handle_add_new_skills",
             role=self.profile,
         )
-
+    
     async def _act(self) -> Message:
         todo = self._rc.todo
         logger.debug(f"Todo is {todo}")
-        self.maintain_actions(todo)    
+        self.maintain_actions(todo)
         # 获取最新的游戏周边信息
+        
         context = self.game_memory.context
         task = self.game_memory.current_task
         
         code = self.game_memory.code
         self.perform_game_info_callback(self.game_memory.event, self.game_memory.summarize_chatlog)
         event_summary = self.game_memory.event_summary
-        try:
-            program_code = code["program_code"] # TODO: Handle code is None, cuz first round DesignCurriculum(code is None) trigger this 
-        except (KeyError, TypeError):
-            program_code = ""
-
+        program_code = self.game_memory.program_code
+        
         program_name = self.game_memory.program_name
         skills = self.game_memory.skills
-
+        
         # msg = self._rc.memory.get(k=1)[0]
-
+        
         retrieve_skills_message_step1 = {"query": context, "skills": skills}
         logger.info(f"check query {context}")
         logger.info(f"check event summary {event_summary}")
         
         retrieve_skills_message_step2 = {"query": context + "\n\n" + event_summary, "skills": skills}
-
+        
         generate_skill_message = self.encapsule_message(program_code, program_name)
-
+        
         add_new_skills_message = {
             "task": task,
             "program_name": program_name,
             "program_code": program_code,
             "skills": skills,
         }
-
+        
         handler_map = {
             DesignCurriculum: self.retrieve_skills,
             RetrieveSkills: self.retrieve_skills,
@@ -139,10 +136,10 @@ class SkillManager(Base):
                 msg = await handler(**generate_skill_message)
             else:
                 msg = await handler(**add_new_skills_message)
-
+            
             msg.cause_by = type(todo)
             msg.round_id = self.round_id
             self._publish_message(msg)
             return msg
-
+        
         raise ValueError(f"Unknown todo type: {type(todo)}")

metagpt/utils/minecraft/action_rsp_parser.py

@@ -38,10 +38,10 @@ def parse_action_response(msg: str):
 
     retry = 3
     error = None # 3 times failed return error
-    babel_generator = require("@babel/generator").default
     while retry > 0:
         try:
             parsed = parse_js_code(msg)
+            babel_generator = require("@babel/generator").default
             # Collect func list: check if func & async
             functions = []
             assert len(list(parsed.program.body)) > 0, "No functions found"

minecraft_run.py

@@ -11,10 +11,9 @@ from metagpt.roles.minecraft.critic_agent import CriticReviewer
 from metagpt.minecraft_team import MinecraftPlayer
 
 
-async def learn(task="Start", investment: float = 50.0, n_round: int = 3):
+async def learn(task="Start", investment: float = 50.0, n_round: int = 50):
     mc_player = MinecraftPlayer()
-    mc_player.set_port(33141) # Modify this to your Minecraft LAN port
-    # mc_player.set_resume(True) # If load json from ckpt dir(include chest_memory, skills, ...)
+    mc_player.set_port(30181) # Modify this to your Minecraft LAN port
     mc_player.hire(
         [
             CurriculumDesigner(),