Merge branch 'update_README' of github.com:ZiTao-Li/agentscope into z…

…t/update_readme

examples/conversation_with_langchain/READMD.md

@@ -0,0 +1,27 @@
+# Create an Agent with LangChain
+
+AgentScope is a highly flexible multi-agent platform. It allows developers
+to create agents with third-party libraries.
+
+In this example, we will show how to create an assistant agent with
+LangChain in AgentScope, and interact with user in a conversation.
+
+**Note** we use OpenAI API for LangChain in this example. Developers can
+modify it according to their own needs.
+
+## Install LangChain
+
+Before running the example, please install LangChain by the following command:
+```bash
+pip install langchain==0.1.11 langchain-openai==0.0.8
+```
+
+## Create Agent with LangChain
+
+In this example, the memory management, prompt engineering, and model
+invocation are all handled by LangChain.
+Specifically, we create an agent class named `LangChainAgent`.
+In its `reply` function, developers only need parse the input message and
+wrap the output message into `agentscope.message.Msg` class.
+After that, developers can build the conversation in AgentScope, and the
+`LangChainAgent` is the same as other agents in AgentScope.

examples/conversation_with_langchain/conversation_with_langchain.py

@@ -0,0 +1,85 @@
+# -*- coding: utf-8 -*-
+"""A simple example of using langchain to create an assistant agent in
+AgentScope."""
+import os
+from typing import Optional
+
+from langchain_openai import OpenAI
+from langchain.memory import ConversationBufferMemory
+from langchain.prompts import PromptTemplate
+from langchain.chains import LLMChain
+
+import agentscope
+from agentscope.agents import AgentBase
+from agentscope.agents import UserAgent
+from agentscope.message import Msg
+
+
+class LangChainAgent(AgentBase):
+    """An agent that implemented by langchain."""
+
+    def __init__(self, name: str) -> None:
+        """Initialize the agent."""
+
+        # Disable AgentScope memory and use langchain memory instead
+        super().__init__(name, use_memory=False)
+
+        # [START] BY LANGCHAIN
+        # Create a memory in langchain
+        memory = ConversationBufferMemory(memory_key="chat_history")
+
+        # Prepare prompt
+        template = """
+                You are a helpful assistant, and your goal is to help the user.
+
+                {chat_history}
+                Human: {human_input}
+                Assistant:"""
+
+        prompt = PromptTemplate(
+            input_variables=["chat_history", "human_input"],
+            template=template,
+        )
+
+        llm = OpenAI(openai_api_key=os.environ["OPENAI_API_KEY"])
+
+        # Prepare a chain and manage the memory by LLMChain in langchain
+        self.llm_chain = LLMChain(
+            llm=llm,
+            prompt=prompt,
+            verbose=False,
+            memory=memory,
+        )
+        # [END] BY LANGCHAIN
+
+    def reply(self, x: Optional[dict] = None) -> Msg:
+        # [START] BY LANGCHAIN
+
+        # Generate response
+        response_str = self.llm_chain.predict(human_input=x.content)
+
+        # [END] BY LANGCHAIN
+
+        # Wrap the response in a message object in AgentScope
+        return Msg(name=self.name, content=response_str)
+
+
+# Build a conversation between user and assistant agent
+
+# init AgentScope
+agentscope.init()
+
+# Create an instance of the langchain agent
+agent = LangChainAgent(name="Assistant")
+
+# Create a user agent from AgentScope
+user = UserAgent("User")
+
+msg = None
+while True:
+    # User input
+    msg = user(msg)
+    if msg.content == "exit":
+        break
+    # Agent speaks
+    msg = agent(msg)

src/agentscope/service/__init__.py

@@ -18,6 +18,7 @@
 from .sql_query.mongodb import query_mongodb
 from .web_search.search import bing_search, google_search
 from .service_response import ServiceResponse
+from .service_factory import ServiceFactory
 from .retrieval.similarity import cos_sim
 from .text_processing.summarization import summarization
 from .retrieval.retrieval_from_list import retrieve_from_list
@@ -48,6 +49,7 @@ def get_help() -> None:
     "query_sqlite",
     "query_mongodb",
     "ServiceResponse",
+    "ServiceFactory",
     "cos_sim",
     "summarization",
     "retrieve_from_list",