Problem Statement
Currently, multi-user development for Semantic Model Extensions (SME) within Fusion Data Intelligence (FDI / Fusion AI Data Platform) suffers from a severe operational bottleneck. Because extensions rely on a linear server side Sandbox framework, only one custom sandbox can be compiled "Applied" and published at a time. Typically, we would not Merge to the Main Sandbox until there is confidence on the code. This could happen much later in the Deployment cycle to upper environments.
When multiple developers are working concurrently across different functional pillars (one extending ERP Financials AP and another extending HCM Workforce metrics), they are forced into a single-file queue for Unit Testing changes or even Report / Workbook development. This architecture creates several critical challenges:
- Parallel Work Stoppage: Simultaneous unit testing and report validation across separate sandboxes are blocked - while one developer completes Unit Testing or builds a report and the other developer has to wait.
- Publishing Latency: Applying features and publishing a sandbox takes an excessive amount of time because the platform must compile and validate the entire metadata footprint against for every single Publish merge.
Proposed Solution
I propose extending the native Git Integration to support a robust GIT Multi-User Development workflow for FDI Customisation and Extension Layers via a visual, branch-based interface directly within the FDI Console. The framework should include the following core capabilities:
- GIT Branch-Based Sandboxes: Opening a sandbox in FDI should implicitly provision or check out a Git "Sandbox" Branch and a "Push" to Publish/Merge to the Main "Live" Sandbox for multiple custom feature sandboxes
- Consistency Validation Check: Instead of validating the entire application schema upon publishing, the platform should isolate and validate only the delta and dependent objects introduced by that branch, drastically slashing applying and publishing wait times.
- Visual "Sync to Main" Conflict Resolution UI: Introduce a built-in merge-review interface similar to what we currently have under limited availability but under a GIT setting. When a developer submits their branch changes to the live mainline metadata trunk, the system should run a background diff check. If a collision is detected (e.g., two developers modifying the same presentation folder column), a visual conflict screen should allow the administrator to select the winning logic without breaking the model.
- Rollback on versioned or tagged metadata: I have a separate idea on this posted TBA but this typical of GIT versioning and rollback feature.
Business Value & Impact
- Accelerated Time-to-Value: Drastically shortens the implementation and deployment lifecycle for large enterprise rollouts by allowing Systems Integrators (SIs) and internal IT teams to transition to a true GIT Multi-User Development model, scaling from 1-2 developers to 5-10+ developers working in parallel.
- Reduced Operational Risk & Instant Rollbacks: Moving extensions to a Git lifecycle means any broken report or faulty custom metric can be instantly rolled back using standard version control history, ensuring zero operational downtime for Deployments to upper environments.
- Seamless Patch Upgradability: By isolating and tracking custom deltas via precise Git commits, the platform can more intelligently reconcile custom extensions against Oracle's automated monthly or quarterly factory upgrades, cutting down post-patch maintenance overhead to near zero.
