Initial Release
The IRIS Agent project is an example of how generative AI and automated data retrieval can be integrated into applications developed with InterSystems technologies. The developed application is designed to interact with the Management Portal Iris, assisting users in retrieving system data and modifying non-critical settings.
InterSystems IRIS: A platform for application development and data integration, utilizing ObjectScript for server-side solutions.
Custom GPT Bot: A customized GPT-based bot designed for specific tasks.
Claude AI Desktop: Used to interact with the server via the Model Context Protocol (MCP).
Node.js: Platform for running and managing MCP servers.
OpenAPI: A standard for describing and interacting with APIs.
Docker: Used for containerizing the application, simplifying deployment and managing dependencies.
git clone https://github.com/banksiaglobal/bg-iris-agent
https://www.docker.com/products/docker-desktop/
docker-compose up -d
Now you can open the management portal page by following the
Production link
_SYSTEM
SYS
The Production consist of the following elements:
The functioning of the system can be seen most clearly in the visual traces. There is a visual trace of receiving the request and generating a response when searching for medicine:
You can get acquainted with the open API scheme by following
the link:
http://localhost:52773/interoperability/LanguageModelIn/openapi
Generate an API key to securely connect your portal to AI
tools:
http://localhost:52773/interoperability/LanguageModelIn/createapikey
After generating the key, it will automatically be available
for connection to the Claude AI.
https://nodejs.org/en
cd mcp-nodejs-server
npm install node-fetch @modelcontextprotocol/sdk zod --savedev @types/node
https://claude.ai/download
node build/index.js
http://localhost:52773/interoperability/LanguageModelIn/createapikey
You can also open the file via terminal. For example, if you have VS Code installed:
MacOS/Linux
code ~/Library/Application\ Support/Claude/
claude_desktop_config.json
Windows
code $env:AppData\Claude\claude_desktop_config.json
Add server configuration. Use this example to fill
the claude_desktop_config.json file you opened before:
{
"mcpServers": {
"bg-iris-agent": {
"command": "node",
"args": [
"<your_local_path_to_the_file_on_your_computer>/bg-iris-agent/mcp-nodejs-server/build/index.js"
],
"env": {
"API_KEY": "YourApiKey"
}
}
}
}
Specify the path to the file index.js on your computer and your api key you’ve generated before (after generation copy your key from the management portal of your local project: System Explorer -> Globals -> find your api key via SecretApiKey name). Save the
claude_desktop_config.json file you just filled,
and restart Claude for Desktop.
You could see if Claude for Desktop picked up the tools by
clicking on the hammer icon.
If the hammer icon has shown up, you can now test your
server via Claude for Desktop chat.
Open ChatGPT Custom GPT and create a new bot.
In the “Configure” section, scroll down to “Actions” and pick “Create new action”.
Next click “Import from URL” and fill using this link:
https://<YOUR_REMOTE_HOST>/interoperability/LanguageModelIn/openapi
Check that servers have an address via https protocol. If not, correct http to https.
Then go to “Authentication” section.
To set “Authentication” options you need to follow this link, to create your own Api key:
https://<YOUR_REMOTE_HOST>/interoperability/LanguageModelIn/createapikey
Copy your key from the management portal of your remote project - System Explorer -> Globals -> find your api key via SecretApiKey name
Enter your Api key in Authentication form in the Api Key field. Use Custom Header Name X-Api-Key like on screenshot.
We are the team of Banksia Global company and we always are opened for something new and interesting!
#BanksiaGlobal #AITools #GenerativeAI #AI #ObjectScript #InterSystems