VIEW26 KB Insights

Make your Knowledge Base top-notch with the new Forge App VIEW26 KB insights. Quickly identify gaps and leverage AI suggestions to improve your KB articles.

Comment

Inspiration

Creating content, especially in platforms like Confluence, is often a rewarding experience. But as many of us know, ensuring content to remain up-to-date is always a challenge. This becomes even more crucial when dealing with Knowledge Base (KB) articles. When Codegeist had the AI theme this time, our team saw an opportunity. We thought, why not leverage AI to help enhance and update content? That's how we came up with the idea of VIEW26 KB Insights

What it does

Block Diagram

The Forge App needs to be installed on both Confluence & JSM of the same cloud instance. Doing so ensures that the KB articles and user sessions from the portal page are tracked and reported.

The App can be accessed via the Confluence space-sidebar. You are presented with a Dashboard which consists of the following sections -

Key KB Metrics:

  • KB article view trends.
  • Top-performing content by page views.
  • Recent customer portal requests. KB Metrics

AI-Driven Insights: The App doesn't just provide KB analytics. Its AI recommendation engine analyses KB views and correlates incoming JSM requests to provide the below recommendations

  • Label suggestions to enhance searchability.
  • Content gaps in your KB by analysing incoming requests.
  • Missing FAQs by examining incoming JSM requests.
  • Outdated or irrelevant KB articles. AI-Driven Insights Updating KB articles: Whether you want to accept, modify, or decline the AI's suggestions, you're in control. Updating KB articles

The App not only helps you with data driven insights for understanding your KB usage but also brings in generative AI recommendation capability to make your KB top notch.

How we built it

We've harnessed the full potential of Forge in the development of the App. Given that the App is designed to track events from both JSM and Confluence, we've integrated modules for both products. This enables separate installations of the App on JSM and Confluence. Due to Forge's limitations with cross-product calls and shared storage, we've utilized MongoDB to store and consolidate events from both platforms.

For the user interface, we've leveraged Forge's Custom UI capabilities.

The Recommendation engine is powered by the OpenAI API. This AI mechanism is set to operate every hour, courtesy of the Forge scheduler. Given the intensive nature of AI rule execution, we employ the Forge async events API to handle requests across multiple function invocations.

We use Forge storage to cache the OpenAPI responses and rely on Atlassian APIs to update pages when users opt for the AI-driven recommendations.

Challenges we ran into

Cross-Product Limitations: Forge's inability to support cross-product calls and shared storage presented a hurdle. We overcame this by integrating MongoDB to capture and amalgamate events from both JSM and Confluence.

Custom-UI Routing issues faced with Forge:
FRGE-1285: Not able to go to parent Route using confluence:spacePage module

Handling AI Intensity: The demanding nature of AI rule processing required us to tap into the Forge async events API, ensuring requests were distributed across several function invocations.

Navigating the OpenAI API: Venturing into the OpenAI API was uncharted territory for us. We faced challenges in crafting the ideal prompt and had to extensively test with varied datasets in the Open API playground.

Module Integration: Combining the numerous modules, especially given the diverse product events, added another layer of complexity to our development process.

Accomplishments that we're proud of

When we first heard about AI theme, we had reservations given the lack of experience in AI App development. However we were quick to ideate and and started small with the concept of using generative AI capabilities for managing Knowledge base articles. We are pretty happy with the clean and simple UI to make AI recommendation in Confluence KB space. This was the very first time we worked on a cross platform Forge App using Async events API.

Every step, from integration to testing, was a delightful journey. Seeing the AI's insights spring to life using actual data has been truly rewarding!

What we learned

Diving into the world of AI was a journey of discovery for us. With platforms like OpenAI and Forge at our disposal, we found ourselves rapidly transforming our concept into a working solution.

What's next for VIEW26 KB Insights

Due to time constraints, there are a few features we couldn't fully implement, but they're high on our priority list:

Dashboard Enhancements: We're working on introducing filter capabilities to refine and customize the dashboard views.

Improved Editor: A advanced editor where in we would be able to modify AI suggested improvements.

Advanced AI Integration: Rather than relying on the default 3.5 model API, we're exploring the potential of training and deploying a fine-tuned model for more precise recommendations.

Deep Dive into Generative AI: We're keen on delving deeper into the fine-tuning concepts of OpenAI and other generative AI platforms to harness their full potential.

Scaling Up: As we look to cater to enterprise customers, rigorous testing will be conducted to ensure the App scales efficiently and meets the demands of larger organizations.

Looking ahead, our roadmap includes releasing the App to the Atlassian marketplace. We're also committed to incorporating responsible AI principles before releasing the App to a wider audience

Built With

Share this project:

Updates