https://github.com/craigomatic/sk-researcher

Using Semantic Kernel to power a research agent that will work on your behalf to research a topic
https://github.com/craigomatic/sk-researcher

Last synced: about 1 month ago
JSON representation

Using Semantic Kernel to power a research agent that will work on your behalf to research a topic

• Host: GitHub
• URL: https://github.com/craigomatic/sk-researcher
• Owner: craigomatic
• License: mit
• Created: 2023-07-26T00:27:57.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-09-01T02:07:57.000Z (over 1 year ago)
• Last Synced: 2024-08-01T13:18:22.859Z (4 months ago)
• Language: C#
• Size: 32.2 KB
• Stars: 7
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-semantickernel - sk-researcher

README

        
# sk-researcher

Using Semantic Kernel to power a research agent that will work on your behalf to research a topic

## Pre-requisites

You will Azure OpenAI or OpenAI keys. In addition it's recommended that you have a deployed version of the [websearch-plugin](https://github.com/craigomatic/websearch-aiplugin/) and [webscraper-plugin](https://github.com/craigomatic/webscraper-aiplugin/) available 

## Usage

Assuming you will be running the app locally, enter your keys and the path to the websearch-plugin and webscraper-plugin into appsettings.json (or even better, create a copy of appsettings.json and rename it to appsettings.development.json).

Then edit the research request Program.cs to the topic you would like to research and run the code!



### Running on Azure

The configuration convention is a little tricky when it comes to collections, to reference your plugins you will need to follow this pattern:

```javascript

name: 'AppConfig:Plugins:0:Name'

value: 'WebSearch'

name: 'AppConfig:Plugins:0:Url'

value: ''

name: 'AppConfig:Plugins:1:Name'

value: 'WebScraper'

name: 'AppConfig:Plugins:1:Url'

value: ''

//the next plugin would use the number 2, and so on...

```

### For extra fun

Here are two additional plugins to configure for the researcher to use:

[azblob-aiplugin](https://github.com/craigomatic/azblob-aiplugin)

[worddoc-aiplugin](https://github.com/craigomatic/worddoc-aiplugin)

Then make a slight modification in ResearchService.cs so that instead of:

```csharp

var researchRequest = $"I would like to learn about {topic}" +

            "The more detail you are able to provide the better. " +

            "If you do a web search, don't just take the summary from the search results, scrape each linked page and summarise for best context." +

            "While the opinion of a journalist is interesting, bias towards original sources where possible.";

```

You have:

```csharp

var researchRequest = $"I would like to learn about {topic}" +

            "The more detail you are able to provide the better. " +

            "If you do a web search, don't just take the summary from the search results, scrape each linked page and summarise for best context." +

            "While the opinion of a journalist is interesting, bias towards original sources where possible. " +

            "Please write your analysis to a Word Document stored on an Azure Blob";

```

Now when you run the agent, you will be given the URI to a SAS protected blob on Azure, with the analysis contained within :)