MultiOn client toolkit update 2.0 (#8750)

- Updated to use newer better function interaction
 - Previous version had only one callback
 - @hinthornw @hwchase17  Can you look into this
 -  Shout out to @MultiON_AI @DivGarg9 on twitter

---------

Co-authored-by: Naman Garg <ngarg3@binghamton.edu>
Co-authored-by: Harrison Chase <hw.chase.17@gmail.com>

docs/extras/integrations/toolkits/multion.ipynb

@@ -5,7 +5,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# Multion Toolkit\n",
+    "# MultiOn Toolkit\n",
     "\n",
     "This notebook walks you through connecting LangChain to the MultiOn Client in your browser\n",
     "\n",
@@ -18,7 +18,32 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install --upgrade multion > /dev/null"
+    "!pip install --upgrade multion langchain -q"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langchain.agents.agent_toolkits import MultionToolkit\n",
+    "import os\n",
+    "\n",
+    "\n",
+    "toolkit = MultionToolkit()\n",
+    "\n",
+    "toolkit"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "tools = toolkit.get_tools()\n",
+    "tools"
    ]
   },
   {
@@ -38,8 +63,9 @@
    "outputs": [],
    "source": [
     "# Authorize connection to your Browser extention\n",
-    "import multion \n",
+    "import multion\n",
-    "multion.login()\n"
+    "multion.login()\n",
+    "\n"
    ]
   },
   {
@@ -57,38 +83,18 @@
    },
    "outputs": [],
    "source": [
-    "from langchain.agents.agent_toolkits import create_multion_agent\n",
+    "from langchain import OpenAI\n",
-    "from langchain.tools.multion.tool import MultionClientTool\n",
+    "from langchain.agents import initialize_agent, AgentType\n",
-    "from langchain.agents.agent_types import AgentType\n",
+    "llm = OpenAI(temperature=0)\n",
-    "from langchain.chat_models import ChatOpenAI"
+    "from langchain.agents.agent_toolkits import MultionToolkit\n",
-   ]
+    "toolkit = MultionToolkit()\n",
-  },
+    "tools=toolkit.get_tools()\n",
-  {
+    "agent = initialize_agent(\n",
-   "cell_type": "code",
+    "    tools=toolkit.get_tools(),\n",
-   "execution_count": null,
+    "    llm=llm,\n",
-   "metadata": {
+    "    agent=AgentType.STRUCTURED_CHAT_ZERO_SHOT_REACT_DESCRIPTION,\n",
-    "tags": []
+    "    verbose = True\n",
-   },
+    ")"
-   "outputs": [],
-   "source": [
-    "\n",
-    "agent_executor = create_multion_agent(\n",
-    "    llm=ChatOpenAI(temperature=0),\n",
-    "    tool=MultionClientTool(),\n",
-    "    agent_type=AgentType.OPENAI_FUNCTIONS,\n",
-    "    verbose=True\n",
-    ")\n"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {
-    "tags": []
-   },
-   "outputs": [],
-   "source": [
-    "agent.run(\"show me the weather today\")"
    ]
   },
   {
@@ -100,7 +106,7 @@
    "outputs": [],
    "source": [
     "agent.run(\n",
-    "    \"Tweet about Elon Musk\"\n",
+    "    \"Tweet 'Hi from MultiOn'\"\n",
     ")"
    ]
   }

libs/langchain/langchain/agents/agent_toolkits/__init__.py

@@ -17,7 +17,7 @@ from langchain.agents.agent_toolkits.gmail.toolkit import GmailToolkit
 from langchain.agents.agent_toolkits.jira.toolkit import JiraToolkit
 from langchain.agents.agent_toolkits.json.base import create_json_agent
 from langchain.agents.agent_toolkits.json.toolkit import JsonToolkit
-from langchain.agents.agent_toolkits.multion.base import create_multion_agent
+from langchain.agents.agent_toolkits.multion.toolkit import MultionToolkit
 from langchain.agents.agent_toolkits.nla.toolkit import NLAToolkit
 from langchain.agents.agent_toolkits.office365.toolkit import O365Toolkit
 from langchain.agents.agent_toolkits.openapi.base import create_openapi_agent
@@ -52,6 +52,7 @@ __all__ = [
     "GmailToolkit",
     "JiraToolkit",
     "JsonToolkit",
+    "MultionToolkit",
     "NLAToolkit",
     "O365Toolkit",
     "OpenAPIToolkit",
@@ -65,7 +66,6 @@ __all__ = [
     "ZapierToolkit",
     "create_csv_agent",
     "create_json_agent",
-    "create_multion_agent",
     "create_openapi_agent",
     "create_pandas_dataframe_agent",
     "create_pbi_agent",

libs/langchain/langchain/agents/agent_toolkits/multion/base.py

@@ -1,58 +0,0 @@
-"""MultiOn agent."""
-
-from typing import Any, Dict, Optional
-
-from langchain.agents.agent import AgentExecutor, BaseSingleActionAgent
-from langchain.agents.agent_toolkits.python.prompt import PREFIX
-from langchain.agents.mrkl.base import ZeroShotAgent
-from langchain.agents.openai_functions_agent.base import OpenAIFunctionsAgent
-from langchain.agents.types import AgentType
-from langchain.base_language import BaseLanguageModel
-from langchain.callbacks.base import BaseCallbackManager
-from langchain.chains.llm import LLMChain
-from langchain.schema import SystemMessage
-from langchain.tools.multion.tool import MultionClientTool
-
-
-def create_multion_agent(
-    llm: BaseLanguageModel,
-    tool: MultionClientTool,
-    agent_type: AgentType = AgentType.ZERO_SHOT_REACT_DESCRIPTION,
-    callback_manager: Optional[BaseCallbackManager] = None,
-    verbose: bool = False,
-    prefix: str = PREFIX,
-    agent_executor_kwargs: Optional[Dict[str, Any]] = None,
-    **kwargs: Dict[str, Any],
-) -> AgentExecutor:
-    """Construct a multion agent from an LLM and tool."""
-    tools = [tool]
-    agent: BaseSingleActionAgent
-
-    if agent_type == AgentType.ZERO_SHOT_REACT_DESCRIPTION:
-        prompt = ZeroShotAgent.create_prompt(tools, prefix=prefix)
-        llm_chain = LLMChain(
-            llm=llm,
-            prompt=prompt,
-            callback_manager=callback_manager,
-        )
-        tool_names = [tool.name for tool in tools]
-        agent = ZeroShotAgent(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs)
-    elif agent_type == AgentType.OPENAI_FUNCTIONS:
-        system_message = SystemMessage(content=prefix)
-        _prompt = OpenAIFunctionsAgent.create_prompt(system_message=system_message)
-        agent = OpenAIFunctionsAgent(
-            llm=llm,
-            prompt=_prompt,
-            tools=[tool],
-            callback_manager=callback_manager,
-            **kwargs,
-        )
-    else:
-        raise ValueError(f"Agent type {agent_type} not supported at the moment.")
-    return AgentExecutor.from_agent_and_tools(
-        agent=agent,
-        tools=tools,
-        callback_manager=callback_manager,
-        verbose=verbose,
-        **(agent_executor_kwargs or {}),
-    )

libs/langchain/langchain/agents/agent_toolkits/multion/toolkit.py

@@ -0,0 +1,22 @@
+"""MultiOn agent."""
+from __future__ import annotations
+
+from typing import List
+
+from langchain.agents.agent_toolkits.base import BaseToolkit
+from langchain.tools import BaseTool
+from langchain.tools.multion.create_session import MultionCreateSession
+from langchain.tools.multion.update_session import MultionUpdateSession
+
+
+class MultionToolkit(BaseToolkit):
+    """Toolkit for interacting with the Browser Agent"""
+
+    class Config:
+        """Pydantic config."""
+
+        arbitrary_types_allowed = True
+
+    def get_tools(self) -> List[BaseTool]:
+        """Get the tools in the toolkit."""
+        return [MultionCreateSession(), MultionUpdateSession()]

libs/langchain/langchain/tools/multion/__init__.py

@@ -1 +1,6 @@
-"""MutliOn Client API toolkit."""
+"""MutliOn Client API tools."""
+
+from langchain.tools.multion.create_session import MultionCreateSession
+from langchain.tools.multion.update_session import MultionUpdateSession
+
+__all__ = ["MultionCreateSession", "MultionUpdateSession"]

libs/langchain/langchain/tools/multion/create_session.py

@@ -0,0 +1,50 @@
+from typing import TYPE_CHECKING, Optional, Type
+
+from pydantic import BaseModel, Field
+
+from langchain.callbacks.manager import CallbackManagerForToolRun
+from langchain.tools.base import BaseTool
+
+if TYPE_CHECKING:
+    # This is for linting and IDE typehints
+    import multion
+else:
+    try:
+        # We do this so pydantic can resolve the types when instantiating
+        import multion
+    except ImportError:
+        pass
+
+
+class CreateSessionSchema(BaseModel):
+    """Input for CreateSessionTool."""
+
+    query: str = Field(
+        ...,
+        description="The query to run in multion agent.",
+    )
+    url: str = Field(
+        "https://www.google.com/",
+        description="""The Url to run the agent at. Note: accepts only secure \
+            links having https://""",
+    )
+
+
+class MultionCreateSession(BaseTool):
+    name: str = "create_multion_session"
+    description: str = """Use this tool to create a new Multion Browser Window \
+        with provided fields.Always the first step to run \
+            any activities that can be done using browser."""
+    args_schema: Type[CreateSessionSchema] = CreateSessionSchema
+
+    def _run(
+        self,
+        query: str,
+        url: Optional[str] = "https://www.google.com/",
+        run_manager: Optional[CallbackManagerForToolRun] = None,
+    ) -> dict:
+        try:
+            response = multion.new_session({"input": query, "url": url})
+            return {"tabId": response["tabId"], "Response": response["message"]}
+        except Exception as e:
+            raise Exception(f"An error occurred: {e}")

libs/langchain/langchain/tools/multion/tool.py

@@ -1,37 +0,0 @@
-"""Tool for MultiOn Extension API"""
-from typing import Any, Optional
-
-from pydantic import Field
-
-from langchain.callbacks.manager import CallbackManagerForToolRun
-from langchain.tools.base import BaseTool
-from langchain.utilities.multion import MultionClientAPIWrapper
-
-
-def _get_default_multion_client() -> MultionClientAPIWrapper:
-    return MultionClientAPIWrapper()
-
-
-class MultionClientTool(BaseTool):
-    """Simulates a Browser interacting agent."""
-
-    name = "Multion_Client"
-    description = (
-        "A api to communicate with browser extension multion "
-        "Useful for automating tasks and actions in the browser "
-        "Input should be a task and a url."
-        "The result is text form of action that was executed in the given url."
-    )
-    api_wrapper: MultionClientAPIWrapper = Field(
-        default_factory=_get_default_multion_client
-    )
-
-    def _run(
-        self,
-        task: str,
-        url: str = "https://www.google.com/",
-        tabId: Optional[Any] = None,
-        run_manager: Optional[CallbackManagerForToolRun] = None,
-    ) -> str:
-        """Use the tool."""
-        return self.api_wrapper.run(task, url, tabId)

libs/langchain/langchain/tools/multion/update_session.py

@@ -0,0 +1,63 @@
+from typing import TYPE_CHECKING, Optional, Type
+
+from pydantic import BaseModel, Field
+
+from langchain.callbacks.manager import CallbackManagerForToolRun
+from langchain.tools.base import BaseTool
+
+if TYPE_CHECKING:
+    # This is for linting and IDE typehints
+    import multion
+else:
+    try:
+        # We do this so pydantic can resolve the types when instantiating
+        import multion
+    except ImportError:
+        pass
+
+
+class UpdateSessionSchema(BaseModel):
+    """Input for UpdateSessionTool."""
+
+    tabId: str = Field(
+        ..., description="The tabID, received from one of the createSessions run before"
+    )
+    query: str = Field(
+        ...,
+        description="The query to run in multion agent.",
+    )
+    url: str = Field(
+        "https://www.google.com/",
+        description="""The Url to run the agent at. \
+        Note: accepts only secure links having https://""",
+    )
+
+
+class MultionUpdateSession(BaseTool):
+    name: str = "update_multion_session"
+    description: str = """Use this tool to update \
+        a existing corresponding \
+    Multion Browser Window with provided fields. \
+        Note:TabId is got from one of the previous Browser window creation."""
+    args_schema: Type[UpdateSessionSchema] = UpdateSessionSchema
+
+    def _run(
+        self,
+        tabId: str,
+        query: str,
+        url: Optional[str] = "https://www.google.com/",
+        run_manager: Optional[CallbackManagerForToolRun] = None,
+    ) -> dict:
+        try:
+            try:
+                response = multion.update_session(tabId, {"input": query, "url": url})
+                content = {"tabId": tabId, "Response": response["message"]}
+                self.tabId = tabId
+                return content
+            except Exception as e:
+                print(f"{e}, creating a new session")
+                response = multion.new_session({"input": query, "url": url})
+                self.tabID = response["tabId"]
+                return {"tabId": response["tabId"], "Response": response["message"]}
+        except Exception as e:
+            raise Exception(f"An error occurred: {e}")

libs/langchain/langchain/utilities/multion.py

@@ -1,75 +0,0 @@
-"""Util that calls MultiOn Client.
-
-In order to set this up, follow instructions at:
-https://multion.notion.site/Download-MultiOn-ddddcfe719f94ab182107ca2612c07a5
-"""
-from typing import Any, Optional
-
-from pydantic import BaseModel
-
-
-class MultionAPI:
-    def __init__(self) -> None:
-        self.tabId = None
-        self.new_session_count = 0
-
-    def create_session(self, query: str, url: str) -> str:
-        """Always the first step to run any activities that can be done using browser.
-
-        Args:
-            'query': the query that you need to perform in the given url.
-                If there is no 'query' set it as open.
-            'url': the base url of a site.
-        """
-        import multion
-
-        # Only create new session once and continue using update session
-        if self.new_session_count < 2:
-            response = multion.new_session({"input": query, "url": url})
-            self.new_session_count += 1
-            self.tabId = response["tabId"]
-            return response["message"]
-        else:
-            return "Continue using update session"
-
-    def update_session(self, query: str, url: str) -> str:
-        """Updates the existing browser session.
-
-        Updates with given action and url, used consequently to handle browser
-        activities after creating one session of browser.
-
-        Args:
-            'query': the query that you need to perform in the given url.
-                If there is no 'query' set it as open.
-            'url': the base url of a site.
-        """
-        import multion
-
-        response = multion.update_session(self.tabId, {"input": query, "url": url})
-        return response["message"]
-
-
-class MultionClientAPIWrapper(BaseModel):
-    """Wrapper for Multion Client API.
-
-    In order to set this up, follow instructions at:
-    NEED TO ADD
-    """
-
-    client: Any = MultionAPI()
-
-    def run(self, task: str, url: str, tabId: Optional[Any]) -> str:
-        """Run body through Multion Client and respond with action.
-
-        Args:
-            task:
-            url:
-            tabId:
-        """
-        if self.client.tabId is None or tabId is None:
-            self.client = MultionAPI()
-            message = self.client.create_session(task, url)
-        else:
-            message = self.client.update_session(task, url)
-
-        return message