The chat example below is meant to demonstrate the use of an OpenAPI-based skill (e.g., GitHub), a planner (e.g., ActionPlanner) and chat completions to create a conversational experience with additional information from a skill when needed.
The GitHub OpenAPI specific used is a reduced definition supporting viewing of pull requests.
Before you get started, make sure you have the following requirements in place:
- .NET 6.0 SDK
- Azure OpenAI resource or an account with OpenAI.
- Clone the repository
- Open the
./appsettings.json
and configure your AI service and GitHub credentials. - Run the sample
- In Visual Studio 2019 or newer, right-click on the
OpenApiSkillsExample
project, select "Set as Startup Project", then pressF5
to run and debug the application - OR open a terminal window, change directory to the
OpenApiSkillsExample
project, then rundotnet run
.
- In Visual Studio 2019 or newer, right-click on the
The sample will provide a simple chat interface the large language model (e.g., gpt-3.5-turbo) and the planner will invoke the GitHub skill when needed.
When asking for results from GitHub, try using bounded asks such as "List the three most recent open pull requests and include the name, number, and state for each."