◈ Aether Flow

Autonomous Multi-Agent AI Orchestrator

Aether Flow is a premium, low-latency AI orchestration engine that leverages a multi-agent "Sequential Handoff Pipeline" to transform raw technical requirements into secure, audited architecture blueprints. Built with React 19, TypeScript, and Gemini 2.5, it features a deterministic execution flow powered by a robust Finite State Machine (FSM).

---

🏗️ Architecture Flow

graph TD
    A[User Prompt] --> B[Researcher Agent]
    B -- "Research Data" --> C[Architect Agent]
    C -- "Architecture Blueprint" --> D[Security Reviewer/Auditor]
    D -- "Security Audit" --> E[Final Pipeline Output]
    
    style B fill:#3b82f6,stroke:#fff,stroke-width:2px,color:#fff
    style C fill:#10b981,stroke:#fff,stroke-width:2px,color:#fff
    style D fill:#f59e0b,stroke:#fff,stroke-width:2px,color:#fff
    style E fill:#8b5cf6,stroke:#fff,stroke-width:2px,color:#fff

Loading

---

� Key Features

• Real-time Scrambling UI: Dynamic text decryption effects for agent outputs, providing a high-fidelity "terminal-reveal" experience.
• Chain of Verification: Every architectural decision is audited by a specialized Security Reviewer to prevent "Context Drift" and ensure safety.
• Deterministic Execution Flow: A strict sequential pipeline managed via Zustand, ensuring predictability and state integrity.
• Low-latency Observability: Integrated with OpenTelemetry patterns for deep pipeline introspection and performance monitoring.
• Enterprise-Ready Integration: Pre-architected for Kafka/Pulsar data streams and OpenTelemetry telemetry.

---

⚙️ The Engine: Sequential Handoff Pipeline

The core orchestration logic resides in src/services/ai/geminiService.ts and src/features/agents/useOrchestrator.ts. It follows a rigid-yet-flexible pipeline where each agent consumes the specialized output of its predecessor.

1. Researcher (🔬): Scans technical requirements and identifies core metrics.
2. Architect (🏗️): Constructs a full-stack blueprint based on research findings.
3. Security Reviewer (🛡️): Conducts an automated vulnerability audit on the blueprint.

---

� The State: Persistent Shared Context

Aether Flow utilizes a Persistent Shared Context pattern managed in src/store/orchestratorStore.ts. This prevents the "Context Drift" common in long-running LLM chains by ensuring each agent has access to a verbatim, tamper-proof record of the pipeline's history.

---

🛠️ Technical Stack

• Frontend: React 19 (Vite), Framer Motion (Animations)
• State: Zustand (Store-level isolation)
• AI: Gemini 2.5 Flash (via @google/generative-ai)
• Infrastructure: OpenTelemetry (Observability), Kafka/Pulsar (Planned Stream Integration)
• Styling: Vanilla CSS with modern Glassmorphism foundations.

---

🏁 Getting Started

1. Clone the Repo:

   git clone https://github.com/your-repo/aether-flow.git
   cd aether-flow

2. Install Dependencies:

   npm install

3. Set Environment Variables: Create a .env file:

   VITE_GEMINI_API_KEY=your_api_key_here

4. Run Development Server:

   npm run dev

---

� Development

The project structure is optimized for feature-based development:

• src/features/agents: Orchestration hooks and agent UI components.
• src/services/ai: Core Gemini integration and prompt engineering.
• src/store: Global state management and FSM logic.

---

🧪 Testing

Aether Flow maintains a rigorous testing suite using Vitest and React Testing Library.

# Run unit tests
npm run test

# Generate coverage report
npm run coverage

---

🗺️ Roadmap

• Multi-path Parallel Orchestration
• Visual Logic Designer (Node-based)
• Real-time Kafka Stream Connectors
• Extended Agent Personalities (FinOps, DevOps)

---

📄 License

MIT © 2026 Aether Flow Team