Skip to content

tmp_arch.md

Latest commit

 

History

History
26 lines (19 loc) · 2.26 KB

tmp_arch.md

File metadata and controls

26 lines (19 loc) · 2.26 KB

BloomPath Technical Architecture

BloomPath is an AI-enhanced bridge application connecting agile project management tools with real-time 3D simulation in Unreal Engine 5. The primary purpose of the architecture is to visually and interactively represent software engineering tasks as organic elements inside a virtual "Garden".

System Components

  1. Project Data Sources

    • Linear & Jira: The raw data ingress point (CRITICAL: You MUST use the official Linear app logo and Atlassian Jira software logo for these, not generic software icons). These services send HTTP POST Webhook Events to the middleware.

  2. BloomPath Middleware (Python Flask)

    • Webhook Receiver: Ingests raw JSON payloads from the project management tools.
    • Data Normalizer: Standardizes the disparate ticket formats into a Unified Ticket representation.
    • Event Queue: A buffer regulating the flow of new Unified Tickets.
    • Core Orchestrator: The central brain driving the business logic and managing the state. It handles events popped from the queue.

  3. AI Generation & Analysis Layer

    • The Core Orchestrator sends context prompts (like feature requests from tickets) to World Labs Marble, which synchronously generates .GLB 3D model assets.
    • A quality validation requests and spatial planning metadata are processed by Gemini 3 Flash (Vision Validation), which returns semantic layout and navigation JSON to the Orchestrator.
    • The architectural state is periodically fed to DiagramAgent (Multi-Agent Architect) to update the documentation.

  4. Engine Presentation Layer

    • The Garden (Unreal Engine 5.7): Receives remote API commands (Spawn, Scale, Adjust Weather) from the Core Orchestrator.
    • Local Client Simulation: Represents the end-user interacting with the UE5 Garden (e.g., watering a plant). Object interaction sync operations are sent back to the Webhook Receiver to trigger return mutations (resolving an incident in Linear/Jira).

The Data Flow Loop

The workflow follows a cyclic pattern: Data comes from Data Sources to Middleware, get augmented by AI Generation, gets rendered in the Engine Presentation Layer, and then client interaction feedback updates the original Data Sources.