Repository files navigation

Theanthropic-tools library simplifies the usage of Anthropics’stool use feature. It abstracts away the complexity of parsing function signatures and docstrings by providing developers with a clean and intuitive interface. It's a near-clone of myopenai-functions library that does the same with OpenAI.

You can installanthropic-tools from PyPI using pip:

pip install anthropic-tools

1. Import the necessary modules and provide your API key:

importenumimportanthropicfromanthropic_toolsimportConversationclient=anthropic.Anthropic(api_key="<YOUR_API_KEY>",)

2. Create aConversation instance:

conversation=Conversation(client)

3. Define your tools using the@conversation.add_tool decorator:

classUnit(enum.Enum):FAHRENHEIT="fahrenheit"CELSIUS="celsius"@conversation.add_tool()defget_current_weather(location:str,unit:Unit=Unit.FAHRENHEIT)->dict:"""Get the current weather in a given location.    Args:        location (str): The city and state, e.g., San Francisco, CA        unit (Unit): The unit to use, e.g., fahrenheit or celsius    """return {"location":location,"temperature":"72","unit":unit.value,"forecast": ["sunny","windy"],    }

4. Ask the AI a question:

response=conversation.ask("What's the weather in San Francisco?")# Should return three messages, the last one's content being something like:# The current weather in San Francisco is 72 degrees Fahrenheit and it is sunny and windy.

You can read more about how to useConversationhere.

fromanthropic_toolsimportToolWrapperwrapper=ToolWrapper(get_current_weather)schema=wrapper.schemaresult=wrapper({"location":"San Francisco, CA"})

Or you could useskills.

anthropic-tools takes care of the following tasks:

• Parsing the function signatures (with type annotations) and docstrings.
• Sending the conversation and tool descriptions to Anthropic Claude.
• Deciding whether to call a tool based on the model's response.
• Calling the appropriate function with the provided arguments.
• Updating the conversation with the tool response.
• Repeating the process until the model generates a user-facing message.

This abstraction allows developers to focus on defining their functions and adding user messages without worrying about the details of tool use.

Please note thatanthropic-tools is an unofficial project not maintained by Anthropic. Use it at your discretion.