Skip to content

Commit 8348566

Browse files
authoredDec 9, 2024
Update README.md
Adding more descriptive context for developers learning about skillsets for the first time. Outlines the differences between skillsets and agents and includes and architectural map.
1 parent f6f90d2 commit 8348566

File tree

1 file changed

+29
-3
lines changed

1 file changed

+29
-3
lines changed
 

‎README.md

+29-3
Original file line numberDiff line numberDiff line change
@@ -6,9 +6,35 @@
66
77
## Description
88

9-
This is an example skillset Copilot Extension, designed to create random test and example data for a number of development purposes. The purpose of this to showcase how an extension can be created as a skillset.
9+
This code sample demonstrates building a Copilot Extension using the skillsets approach rather than a traditional agent. This extension is designed to generate random test and example data for a number of development purposes, by calling publicly available APIs.
1010

11-
## Installation:
11+
### Architectural Model
12+
- **Skillsets**: Define up to 5 API endpoints that Copilot can call directly. Copilot handles all AI interactions, prompt engineering, and response formatting.
13+
- **Agents**: Provide full control over the interaction flow, including custom prompt crafting and specific LLM model selection.
14+
15+
![Architectural comparison between Skillsets and Agents](https://github.com/user-attachments/assets/462a5963-bce9-4a4e-8030-3434aa904f4e)
16+
17+
### When to Choose Skillsets
18+
Skillsets are ideal when you want to:
19+
- Quickly integrate existing APIs or services without managing AI logic
20+
- Focus purely on your service's core functionality
21+
- Maintain consistent Copilot-style interactions without extensive development
22+
- Get started with minimal infrastructure and setup
23+
24+
Use agents instead if you need:
25+
- Complex custom interaction flows
26+
- Specific LLM model control (using LLMs that aren't provided by the Copilot API)
27+
- Custom prompt crafting
28+
- Advanced state management
29+
30+
## Example Implementation
31+
32+
This extension showcases the skillset approach by providing three simple endpoints that generate random development data:
33+
- Random commit messages
34+
- Lorem ipsum text generation
35+
- Random user data
36+
37+
## Getting Started
1238
1. Clone the repository:
1339

1440
```
@@ -37,7 +63,7 @@ ngrok http http://localhost:8080
3763
go run .
3864
```
3965

40-
## Accessing the Agent in Chat:
66+
## Accessing the Extension in Chat:
4167

4268
1. In the `Copilot` tab of your Application settings (`https://github.com/settings/apps/<app_name>/agent`)
4369
- Set the app type to "Skillset"

0 commit comments

Comments
 (0)
Failed to load comments.