Skip to content

Latest commit

 

History

History
165 lines (101 loc) · 4.89 KB

addToolMessage.md

File metadata and controls

165 lines (101 loc) · 4.89 KB

addToolMessage

Add tool message to message history

updatedMessages = addToolMessage(messages,toolCallID,name,content)

Description

Add tool messages to the message history to pass the return of a function call to a large language model. For more information on function calling, see openAIFunction.

updatedMessages = addToolMessage(messages,toolCallID,name,content) adds a tool message to the messageHistory object messages and specifies the tool call ID, the name of the speaker, and the content of the message.

Examples

Compute Sine Using OpenAI Function Call

First, specify the OpenAI® API key as an environment variable and save it to a file called ".env". Next, load the environment file using the loadenv function.

loadenv(".env")

Create an openAIFunction object that represents the sind function. The sind function has a single input argument, x, representing the input angle in degrees.

f = openAIFunction("sind","Sine of argument in degrees");
f = addParameter(f,"x",type="number",description="Angle in degrees");

Connect to the OpenAI Chat Completion API. Pass the openAIFunction object f as an input argument.

model = openAIChat("You are a helpful assistant.",Tools=f);

Initialize the message history. Add a user message to the message history.

messages = messageHistory;
messages = addUserMessage(messages,"What is the sine of thirty?");

Generate a response based on the message history.

[~,completeOutput] = generate(model,messages)
completeOutput = struct with fields:
          role: 'assistant'
       content: []
    tool_calls: [1x1 struct]
       refusal: []

The model has not generated any text. Instead, it has created a function call, completeOutput.tool_calls.

Add the response to the message history.

messages = addResponseMessage(messages,completeOutput);

Extract the tool call ID and the name of the called function.

toolCallID = string(completeOutput.tool_calls.id)
toolCallID = "call_fnCZwyltX0jJmVweBTAgC4qI"

functionCalled = string(completeOutput.tool_calls.function.name)
functionCalled = "sind"

Make sure that the model is calling the correct function. Even with only a single function, large language models can hallucinate function calls to fictitious functions.

Extract the input argument values from the complete output using the jsondecode function. Compute the sine of the generated argument value and add the result to the message history using the addToolMessage function.

if functionCalled == "sind"
    args = jsondecode(completeOutput.tool_calls.function.arguments);
    result = sind(args.x)
    messages = addToolMessage(messages,toolCallID,functionCalled,"x="+result);
end
result = 0.5000

Finally, generate a natural language response.

generatedText = generate(model,messages)
generatedText = "The sine of 30 degrees is 0.5."

Input Arguments

messages — Message history

messageHistory object

Message history, specified as a messageHistory object.

toolCallID — Tool call ID

string scalar | character vector

Tool call ID, specified as a string scalar or character vector.

If an LLM creates a function call during generation, then the tool call ID is part of the complete output of the generate function.

name — Tool name

string scalar | character vector

Name of the tool, specified as a string scalar or character vector. The name must be nonempty and must only contain letters, numbers, underscores (_), and dashes (-).

content — Message content

string scalar | character vector

Message content, specified as a string scalar or character vector. The content must be nonempty.

Output Argument

updatedMessages — Updated message history

messageHistory object

Updated message history, specified as a messageHistory object. The updated message history includes a new structure array with these fields:

  • tool_call_id — Set by the toolCallID input argument
  • role —"tool"
  • name — Set by the name input argument
  • content — Set by the content input argument

See Also

messageHistory | openAIFunction | generate | addResponseMessage

Copyright 2024 The MathWorks, Inc.