OpenAIPromptExecutionSettings settings = new() { ToolCallBehavior = ToolCallBehavior.AutoInvokeKernelFunctions };\r\nConsole.WriteLine(await kernel.InvokePromptAsync(\"How many days until Christmas? Explain your thinking.\", new(settings)));<\/code><\/pre>\nTo allow the AI Model to call your functions you need to provide in the settings that you want a specific ToolCallingBehavior, not providing it the AI won\u2019t have visibility to your functions.<\/p>\n
Note that this functionality will only work with AI models and connectors that support Tool\/Function Calling like Open AI latest chat completion models.<\/p>\n
In this example the AI Model will have visibility to your plugins and will call the functions it sees as necessary to fulfil the ask.<\/p>\n
2.4 Auto Function Invoking with your Complex Object Plugin Functions<\/strong>\nOne interesting aspect of complex objects in regards to the AI Model is that it knows the structure of the parameters of your functions and provide them following the description and the deserialization representation of those parameters.<\/p>\n\r\nConsole.WriteLine(await kernel.InvokePromptAsync(\"Create a handy lime colored widget for me.\", new(settings)));\r\nConsole.WriteLine(await kernel.InvokePromptAsync(\"Create a beautiful scarlet colored widget for me.\", new(settings)));\r\nConsole.WriteLine(await kernel.InvokePromptAsync(\"Create an attractive maroon and navy colored widget for me.\", new(settings)));<\/code><\/pre>\nThe above examples send prompts that will influence the AI decision on different parameter combinations to call the CreateWidget function.<\/p>\n
3. Using Yaml Prompt Files<\/strong><\/p>\nWhile you can provide templates in the code you can also import those from Yaml files. This sample show how can you create a Prompt Function in that way.<\/p>\n
Semantic Kernel Format Configuration<\/strong><\/p>\n\r\nname: GenerateStory\r\ntemplate: |\r\n\u00a0 Tell a story about {{$topic}} that is {{$length}} sentences long.\r\n\r\ntemplate_format: semantic-kernel\r\ndescription: A function that generates a story about a topic.\r\ninput_variables:\r\n\u00a0 - name: topic\r\n\u00a0\u00a0\u00a0 description: The topic of the story.\r\n\u00a0\u00a0\u00a0 is_required: true\r\n\u00a0 - name: length\r\n\u00a0\u00a0\u00a0 description: The number of sentences in the story.\r\n\u00a0\u00a0\u00a0 is_required: true\r\noutput_variable:\r\n\u00a0 description: The generated story.\r\nexecution_settings:\r\n\u00a0 default:\r\n\u00a0\u00a0\u00a0 temperature: 0.6<\/code><\/pre>\nHandlebars Format Configuration<\/strong><\/p>\n\r\nname: GenerateStory\r\ntemplate: |\r\n\u00a0 Tell a story about {{topic}} that is {{length}} sentences long.\r\ntemplate_format: handlebars\r\ndescription: A function that generates a story about a topic.\r\ninput_variables:\r\n\u00a0 - name: topic\r\n\u00a0\u00a0\u00a0 description: The topic of the story.\r\n\u00a0\u00a0\u00a0 is_required: true\r\n\u00a0 - name: length\r\n\u00a0\u00a0\u00a0 description: The number of sentences in the story.\r\n\u00a0\u00a0\u00a0 is_required: true\r\noutput_variable:\r\n\u00a0 description: The generated story.\r\nexecution_settings:\r\n\u00a0 service1:\u00a0\r\n\u00a0\u00a0\u00a0 model_id: gpt-4\r\n\u00a0\u00a0\u00a0 temperature: 0.6\r\n\u00a0 service2:\r\n\u00a0\u00a0\u00a0 model_id: gpt-3\r\n\u00a0\u00a0\u00a0 temperature: 0.4\r\n\u00a0 default:\r\n \u00a0\u00a0 temperature: 0.5<\/code><\/pre>\nThe Yaml file structures above provides important details. about the prompt used, description to the AI Model, specially:<\/p>\n
![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture1-red-box.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture2.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture3.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture4-1.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture5.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture6.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture8.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture9.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture10.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture11.png\")
<\/a><\/p>\n![\"Image](\"https:\/\/devblogs.microsoft.com\/semantic-kernel\/wp-content\/uploads\/sites\/78\/2024\/05\/Picture12.png\")
<\/a><\/p>\n