Hello everyone!

We have been busy working on the AiFGE app, and I wanted to share some updates with you. Our most recent update today is changes to the annotations system. We are relying on OpenAI's Assistants endpoints to provide Retrieval Augmented Generation, or RAG, to search the Master Agreement and other documents that we've uploaded. It's considered beta right now, and very buggy, and we wanted to make sure that we were managing all of the errors properly. As a result, we were forced to re-write our program logic to handle edge cases. Here's how the application works now.

If the assistant returns a message with annotations, we will update the content to format the annotation, and provide each citations as independent elements. Sometimes the agent will return a reference in the content, such as 【8:9†VA-AFGE-2023-Master-Agreement.pdf】, but will not return any annotations, so you may still see this reference format. When the the tools return citations with no quotes, we assume the source is apparent in context via the thread or message, and do not return the annotation at all.

We are continuing to obsess over code quality, and really understanding the frameworks that the app is built on, namely OpenAI, and Chainlit, which we use for our frontend and backend data layer. We're really focusing on making AiFGE simple to use, and really good at what it does, which is help stewards find the relevant information necessary to help their BUEs. Right now we're focusing on grievance management, with an eye toward expanding the capabilities of our base assistant, while adding additional assistants customized for users and groups of users based on agency or locations. The sky is the limit here, but we're being very selective and methodical about how we approach development.

Your feedback is crucial here. Make sure to use the 👍👎 feedback buttons for any response that is inaccurate or buggy, or drop me a line at help@wade.digital if there's anything else you'd like to report.

Thank you so much for using AiFGE!

In Solidarity,

MW

Hello Everyone!

Thank you for checking out our site. I hope you've had a chance to try our AiFGE assistant app. Here at Wade.Digital we're building AI-enabled applications and automation pipelines, and AiFGE is a project that I started earlier this year to help my wife Stacy, who is the Vice-President of AFGE Local 2328 here at the Hampton Virginia Medical Center.

As Stacy’s husband, I understand the challenges she goes through in order to provide support for her members. The Master Agreement is a large document, and I understand all the resources the AFGE puts into training its officers and stewards on how to wield that powerful document properly.

I'm a software developer by trade, and for the last year and a half have been fascinated by ChatGPT and advances in artificial intelligence. So to that end, I decided to build a demo ChatGPT-type app that has the master agreement loaded in it. You can see it for yourself at http://app.aifge.xyz.

We've prompted it to serve as a steward assistant, showcasing how this product can be used to take problem statements in natural language and provide next steps toward resolution. If there is a conflict with the Master Agreement, it will reference the relevant portion with it's response. It can even output form letters, draft emails, and draft grievances.

In addition to the Master Agreement, we've also indexed supplemental agreements relevant to 2328. We've added some basic Title 38 information as well, and will be adding case notes, legal briefs and even VA published field manuals as needed. Right now we have links to these documents within the app's readme tab, and we'll be adding a codex viewer to display the documents in the future.

We hope that the value of such an application is apparent, and we are actively seeking funding from the AFGE to build out a full-featured app that can be scaled and made available to the entirety of their organization. We have the following short-term objectives:

• Continue scaling and collecting user feedback.
• Determine where the bot is providing incorrect information, and adjust our models as needed.
• Refine Title 38 Data and output
• Implement a multi-agent system that will allow us to customize document codexes on a council, local, or individual level

Long term, our main priority is to help the AFGE, as well as other labor organizations, support the growing demands of their members by giving their leadership these advanced tools and others like it. My wife and I are both committed to the resurgence of worker power in the United States, as well as the development of ethical AI applications.

If you'd lke to get in touch with us, please send an email to contact at wade dot digital. I'd love to hear from you.

In Solidarity,

MW