Replies: 2 comments
-
|
My colleague had answered a question similar to this in a separate thread, so borrowing his answer: "In this sample they are being used to add memory to an agent, but the InvokingAsync method can be used to also inject tools into the agent based on the current user request. You can therefore use an AIContextProvider to choose the most appropriate functions to advertise to the agent for a specific run, based on the user input for that run. It's also possible to maintain some recent chat history to improve the breadth of advertised functions, since the last message is not always deterministic enough." cc: @westey-m who originally answered (in a different thread) |
Beta Was this translation helpful? Give feedback.
-
|
Related issue: #3083 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hey,
There is an evolving "skills" pattern - allowing an agent to dynamically load a skill, which is basically a prompt + tools for a specific expertise.
Loading a skill could be implemented using a custom
load_skill(skillName)tool. But I wonder if it's possible make such a tool invocation change the agent's tools, during the same turn?I'm using .NET, Azure OpenAI, and both Completions and Response APIs.
Thanks!
Beta Was this translation helpful? Give feedback.
All reactions