Skip to content

Latest commit

 

History

History
151 lines (107 loc) · 6.26 KB

tool_guide.md

File metadata and controls

151 lines (107 loc) · 6.26 KB
Raw

Custom Tools

Follow these instructions to create your own custom tools.

Custom tools will need to be built in the community folder. Make sure you've enabled the INSTALL_COMMUNITY_DEPS build arg in the docker-compose.yml file by setting it to true.

Step 1: Choose a Tool to Implement

You can take a tool implementation easily from:

Step 2: Select Your Tool Type

There are three types of tools:

  • Data Loader: This tool type retrieves data from a source. Examples include the LangChain Wikipedia retriever and Arxiv.
  • File Loader: This tool type loads and parses files. Examples include the LangChain Vector DB Retriever and LlamaIndex Upload PDF Retriever.
  • Function: This is a unique tool type that performs a specific action. Examples include the Python Interpreter and Calculator.

Step 3: Implement the Tool

Add your tool implementation here (please note that this link is subject to change).

If you need to install a new library to run your tool, execute the following command and run make dev again.

poetry add <MODULE> --group community

Implementing a Langchain Tool

Add the implementation inside a tool class that inherits from BaseTool. This class will need to implement the call() method, which should return a list of dictionary results.

Note: To enable citations, each result in the list should contain a "text" field.

For example, let's look at the community-implemented ArxivRetriever:

from typing import Any, Dict, List

from langchain_community.utilities import ArxivAPIWrapper

from community.tools import BaseTool


class ArxivRetriever(BaseTool):
    NAME = "arxiv"

    def __init__(self):
        self.client = ArxivAPIWrapper()

    @classmethod
    # If your tool requires any environment variables such as API keys,
    # you will need to assert that they're not None here
    def is_available(cls) -> bool:
        return True

    # Your tool needs to implement this call() method
    def call(self, parameters: str, **kwargs: Any) -> List[Dict[str, Any]]:
        result = self.client.run(parameters)

        return [{"text": result}] # <- Return list of results, in this case there is only one

Implementing a Custom Tool

In the tools folder, add a new file for your custom tool implementation. As a starting point, we recommend copy pasting the existing Calculator tool. Modify the class name and NAME attribute to suit your tool.

IMPORTANT: the call() method is where the tool call itself happens, i.e: this is where you want to perform your business logic (API call, data ETL, hardcoded response, etc). You can set custom parameters that you can use with the parameter_definitions key when specifying your tool config (see below). The return value of this tool MUST BE a list of dictionaries that have the text (required), url (optional), title (optional) fields as keys.

For example:

return [{"text": "The fox is blue", "url": "wikipedia.org/foxes", "title": "Color of foxes"}, {..}, {..}]

Next, add your tool class to the init file by locating it in src/community/tools/__init__.py. Import your tool here, then add it to the __all__ list.

To enable your tool, you will need to go to the configuration.yaml file and add your tool's name to the list of enabled_tools. This tool name will correspond to the one defined in the NAME attribute of your class.

Finally, you will need to add your tool definition to the config file. Locate it in src/community/config/tools.py, and import your tool at the top with from backend.tools import ...

In the ToolName enum, add your tool as an enum value. For example, My_Tool = MyTool.NAME.

In the ALL_TOOLS dictionary, add your tool definition. This should look like:

    ToolName.My_Tool: ManagedTool(  # THE TOOLNAME HERE CORRESPONDS TO THE ENUM YOU DEFINED EARLIER
        display_name="My Tool",
        implementation=MyTool, # THIS IS THE CLASS YOU IMPORTED AT THE TOP
        parameter_definitions={ # THESE ARE PARAMS THE MODEL WILL SEND TO YOUR TOOL, ADJUST AS NEEDED
            "query": {
                "description": "Query to search with",
                "type": "str",
                "required": True,
            }
        },
        is_visible=True, 
        is_available=MyTool.is_available(), 
        auth_implementation=None, # EMPTY IF NO AUTH NEEDED
        error_message="Something went wrong",
        category=Category.DataLoader,  # CHECK CATEGORY ENUM FOR POSSIBLE VALUES
        description="An example definition to get you started.",
    ),

Step 5: Test Your Tool!

Now, when you run the toolkit, all the visible tools, including the one you just added, should be available!

  • Run make dev
  • Open http://localhost:4000/
  • Open the side panel
  • Your tool should be there!
  • Select it and send a message that triggers it
  • Appreciate a grounded response with something ✨you created from scratch✨!

Remember, you can also access your tools via the API.

  • List tools:
curl --location --request GET 'http://localhost:8000/v1/tools' \
--header 'User-Id: me' \
--header 'Content-Type: application/json' \
--data '{}'
  • Chat turns with tools:
curl --location 'http://localhost:8000/v1/chat-stream' \
--header 'User-Id: me' \
--header 'Content-Type: application/json' \
--data '{
    "message": "Tell me about the aya model",
    "tools": [{"name": "Arxiv"}]
}
'

Step 6 (extra): Add Unit tests

If you would like to go above and beyond, it would be helpful to add some unit tests to ensure that your tool is working as expected. Create a file here and add a few test cases.