add muagen sdk v0.1.0

.github/workflows/docker-image-pull.yml

@@ -11,8 +11,8 @@ jobs:
         architecture: [amd64, arm64]
         os: [linux]
         service:
-          - name: runtime:0.1.0
-          - name: muagent:0.1.0
+          - name: runtime:0.1.1
+          - name: muagent:0.1.1
           - name: ekgfrontend:0.1.0
 
     steps:

.github/workflows/docker-image.yml

@@ -12,7 +12,7 @@ jobs:
           - name: runtime
             context: ./runtime
             dockerfile: ./runtime/Dockerfile.no-package
-            tag: ghcr.io/codefuse-ai/runtime:0.1.0
+            tag: ghcr.io/codefuse-ai/runtime:0.1.1
             tag_latest: ghcr.io/codefuse-ai/runtime:latest
           - name: ekgfrontend
             context: .
@@ -22,7 +22,7 @@ jobs:
           - name: ekgservice
             context: .
             dockerfile: ./Dockerfile_gh
-            tag: ghcr.io/codefuse-ai/muagent:0.1.0
+            tag: ghcr.io/codefuse-ai/muagent:0.1.1
             tag_latest: ghcr.io/codefuse-ai/muagent:latest
 
     steps:

docker-compose.yaml

@@ -190,7 +190,7 @@ services:
       context: .
       dockerfile: Dockerfile
     container_name: ekgservice
-    image: muagent:0.1.0
+    image: muagent:0.1.1
     environment:
       USER: root
       TZ:   "${TZ}"

docker_pull_images.sh

@@ -17,11 +17,11 @@ docker pull redis/redis-stack:7.4.0-v0
 docker pull ollama/ollama:0.3.6
 
 # pull images from github ghcr.io by nju
-docker pull ghcr.nju.edu.cn/runtime:0.1.0
-docker pull ghcr.nju.edu.cn/muagent:0.1.0
+docker pull ghcr.nju.edu.cn/runtime:0.1.1
+docker pull ghcr.nju.edu.cn/muagent:0.1.1
 docker pull ghcr.nju.edu.cn/ekgfrontend:0.1.0
 
 # # pull images from github ghcr.io
-# docker pull ghcr.io/runtime:0.1.0
-# docker pull ghcr.io/muagent:0.1.0
+# docker pull ghcr.io/runtime:0.1.1
+# docker pull ghcr.io/muagent:0.1.1
 # docker pull ghcr.io/ekgfrontend:0.1.0

examples/ekg_examples/start.py

@@ -37,6 +37,7 @@
     import test_config
 
 from muagent.schemas.db import *
+from muagent.schemas.apis.ekg_api_schema import LLMFCRequest
 from muagent.db_handler import *
 from muagent.llm_models.llm_config import EmbedConfig, LLMConfig
 from muagent.service.ekg_construct.ekg_construct_base import EKGConstructService
@@ -46,7 +47,8 @@
 
 from pydantic import BaseModel
 
-
+from muagent.schemas.models import ModelConfig
+from muagent.models import get_model
 
 
 cur_dir = os.path.dirname(__file__)
@@ -92,56 +94,75 @@ def update_params(self, **kwargs):
 
     def _llm_type(self, *args):
         return ""
-
-    def predict(self, prompt: str, stop = None) -> str:
-        return self._call(prompt, stop)
-
-    def _call(self, prompt: str,
-              stop = None) -> str:
+
+    def _get_model(self):
         """_call
         """
-        return_str = ""
-        stop = stop or self.stop
-
-        if self.model_type == "ollama":
-            stream = ollama.chat(
-                model=self.model_name,
-                messages=[{'role': 'user', 'content': prompt}],
-                stream=True,
-            )
-            answer = ""
-            for chunk in stream:
-                answer += chunk['message']['content']
-
-            return answer
-        elif self.model_type == "openai":
+        if self.model_type in [
+            "ollama", "qwen", "openai", "lingyiwanwu",
+            "kimi", "moonshot",
+        ]:
             from muagent.llm_models.openai_model import getChatModelFromConfig
             llm_config = LLMConfig(
                 model_name=self.model_name,
-                model_engine="openai",
+                model_engine=self.model_type,
                 api_key=self.api_key,
                 api_base_url=self.url,
                 temperature=self.temperature,
                 stop=self.stop
             )
             model = getChatModelFromConfig(llm_config)
-            return model.predict(prompt, stop=self.stop)
-        elif self.model_type in ["lingyiwanwu", "kimi", "moonshot", "qwen"]:
-            from muagent.llm_models.openai_model import getChatModelFromConfig
-            llm_config = LLMConfig(
+        else:
+            model_config = ModelConfig(
+                model_type=self.model_type,
                 model_name=self.model_name,
-                model_engine=self.model_type,
                 api_key=self.api_key,
-                api_base_url=self.url,
+                api_url=self.url,
                 temperature=self.temperature,
-                stop=self.stop
             )
-            model = getChatModelFromConfig(llm_config)
-            return model.predict(prompt, stop=self.stop)
-        else:
-            pass
+            model = get_model(model_config)
+        return model
+
+    def predict(self, prompt: str, stop = None) -> str:
+        return self._call(prompt, stop)
 
-        return return_str
+    def fc(self, request: LLMFCRequest) -> str:
+        """_function_call
+        """
+        if self.model_type not in [
+            "openai", "ollama", "lingyiwanwu", "kimi", "moonshot", "qwen"
+        ]:
+            return f"{self.model_type} not in valid model range"
+
+        model = self._get_model()
+        return model.fc(
+            messages=request.messages, 
+            tools=request.tools, 
+            tool_choice=request.tool_choice, 
+            parallel_tool_calls=request.parallel_tool_calls, 
+        )
+
+    def _call(self, prompt: str,
+              stop = None) -> str:
+        """_call
+        """
+        return_str = ""
+        stop = stop or self.stop
+        if self.model_type not in [
+            "openai", "ollama", "lingyiwanwu", "kimi", "moonshot", "qwen"
+        ]:
+            pass
+        elif self.model_type not in [
+            "dashscope_chat", "moonshot_chat", "ollama_chat",
+            "openai_chat", "qwen_chat", "yi_chat",
+            "dashscope_text_embedding", "ollama_embedding", "openai_embedding", "qwen_text_embedding"
+        ]:
+            pass
+        else:
+            return f"{self.model_type} not in valid model range"
+
+        model = self._get_model()
+        return model.predict(prompt, stop=self.stop)
 
 
 class CustomEmbeddings(Embeddings):
@@ -185,6 +206,17 @@ def _get_sentence_emb(self, sentence: str) -> dict:
             )
             text2vector_dict = get_embedding("openai", [sentence], embed_config=embed_config)
             return text2vector_dict[sentence]
+        elif self.embedding_type in [
+            "dashscope_text_embedding", "ollama_embedding", "openai_embedding", "qwen_text_embedding"
+        ]:
+            model_config = ModelConfig(
+                model_type=self.embedding_type,
+                model_name=self.model_name,
+                api_key=self.api_key,
+                api_url=self.url,
+            )
+            model = get_model(model_config)
+            return model.embed_query(sentence)
         else:
             pass
 
@@ -280,6 +312,7 @@ def embed_query(self, text: str) -> List[float]:
     llm_config=llm_config,
     tb_config=tb_config,
     gb_config=gb_config,
+    initialize_space=True,
     clear_history_data=clear_history_data
 )
 

examples/muagent_examples/docchat_example.py

@@ -60,7 +60,8 @@
 # create your knowledge base
 from muagent.service.kb_api import create_kb, upload_files2kb
 from muagent.utils.server_utils import run_async
-from muagent.orm import create_tables
+# from muagent.orm import create_tables
+from muagent.db_handler import create_tables
 
 
 # use to test, don't create some directory

examples/test_config.py.example

@@ -1,6 +1,8 @@
 import os, openai, base64
 from loguru import logger
 
+os.environ["DM_llm_name"] = 'Qwen2_72B_Instruct_OpsGPT'  #or  gpt_4
+
 # 兜底大模型配置
 OPENAI_API_BASE = "https://api.openai.com/v1"
 os.environ["API_BASE_URL"] = OPENAI_API_BASE
@@ -19,6 +21,78 @@ os.environ["gpt4-llm_temperature"] = "0.0"
 
 
 
+MODEL_CONFIGS = {
+    # old llm config
+    "default": {
+        "model_name": "gpt-3.5-turbo",
+        "model_engine": "qwen",
+        "temperature": "0", 
+        "api_key": "",
+        "api_base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+    },
+    "codefuser":{
+        "model_name": "gpt-4",
+        "model_engine": "openai",
+        "temperature": "0", 
+        "api_key": "",
+        "api_base_url": OPENAI_API_BASE,
+    },
+    # new llm config
+    "dashscope_chat": {
+        "model_type": "dashscope_chat",
+        "model_name": "qwen2.5-72b-instruct" ,
+        "api_key": "",
+    },
+    "moonshot_chat": {
+        "model_type": "moonshot_chat",
+        "model_name": "moonshot-v1-8k" ,
+        "api_key": "",
+    },
+    "ollama_chat": {
+        "model_type": "ollama_chat",
+        "model_name": "qwen2.5-0.5b",
+        "api_key": "",
+    },
+    "openai_chat": {
+        "model_type": "openai_chat",
+        "model_name": "gpt-4",
+        "api_key": "",
+    },
+    "qwen_chat": {
+        "model_type": "qwen_chat",
+        "model_name": "qwen2.5-72b-instruct",
+        "api_key": "",
+    },
+    "yi_chat": {
+        "model_type": "yi_chat",
+        "model_name": "yi-lightning" ,
+        "api_key": "",
+    },
+    # embedding configs
+    "dashscope_text_embedding": {
+        "model_type": "dashscope_text_embedding",
+        "model_name": "text-embedding-v3",
+        "api_key": "",
+    },
+    "ollama_embedding": {
+        "model_type": "ollama_embedding",
+        "model_name": "qwen2.5-0.5b",
+        "api_key": "",
+    },
+    "openai_embedding": {
+        "model_type": "openai_embedding",
+        "model_name": "text-embedding-ada-002",
+        "api_key": "",
+    },
+    "qwen_text_embedding": {
+        "model_type": "dashscope_text_embedding",
+        "model_name": "text-embedding-v3",
+        "api_key": "",
+    },
+}
+
+os.environ["MODEL_CONFIGS"] = json.dumps(MODEL_CONFIGS)
+
 ####   NebulaHandler   ####
 os.environ['nb_host'] = 'graphd'
 os.environ['nb_port'] = '9669'
@@ -41,8 +115,36 @@ os.environ["tb_index_name"] = "ekg_migration_new"
 os.environ['tb_definition_value'] = 'message_test_new'
 os.environ['tb_expire_time'] = '604800' #86400*7
 
-# clear history data in tb and gb
-os.environ['clear_history_data'] = 'True'
+
+#################
+##  DB_CONFIGS ##
+#################
+DB_CONFIGS = {
+    "gb_config": {
+        "gb_type": "NebulaHandler",
+        "extra_kwargs": {
+            'host':'graphd',
+            'port': '9669',
+            'username': os.environ['nb_username'],
+            'password': os.environ['nb_password'],
+            'space': "client"
+        }
+    },
+    "tb_config": {
+        "tb_type": 'TBaseHandler',
+        "index_name": "opsgptkg",
+        "host": 'redis-stack',
+        "port": '6379',
+        "username": os.environ['tb_username'],
+        "password": os.environ['tb_password'],
+        "extra_kwargs": {
+            "definition_value": "opsgptkg",
+            "memory_definition_value": "opsgptkg_message"
+        }
+    }
+}
+os.environ["DB_CONFIGS"] = json.dumps(DB_CONFIGS)
+
 
 
 ########################################

muagent/__init__.py

@@ -1,7 +1,11 @@
-# encoding: utf-8
-'''
-@author: 温进
-@file: __init__.py.py
-@time: 2023/11/9 下午4:01
-@desc:
-'''
+from .ekg_project import EKG, get_ekg_project_config_from_env
+from .project_manager import get_project_config_from_env
+from .models import get_model
+from .agents import get_agent
+from .tools import get_tool
+
+__all__ = [
+    "EKG", "get_model", "get_agent", "get_tool",
+    "get_ekg_project_config_from_env", 
+    "get_project_config_from_env"
+]

muagent/agents/__init__.py

@@ -0,0 +1,30 @@
+from .base_agent import BaseAgent
+from .single_agent import SingleAgent
+from .react_agent import ReactAgent
+from .task_agent import TaskAgent
+from .group_agent import GroupAgent
+from .user_agent import UserAgent
+from .functioncall_agent import FunctioncallAgent
+from ..schemas import AgentConfig
+
+__all__ = [
+    "BaseAgent",
+    "SingleAgent",
+    "ReactAgent",
+    "TaskAgent",
+    "GroupAgent",
+    "UserAgent",
+    "FunctioncallAgent"
+]
+
+
+def get_agent(agent_config: AgentConfig) -> BaseAgent:
+    """Get the agent by agent config
+
+    Args:
+        agent_config (`AgentConfig`): The agent config
+
+    Returns:
+        `BaseAgent`: The specific agent
+    """
+    return BaseAgent.init_from_project_config(agent_config)