Distributed Agentic AI Platform

A Docker-based distributed platform for deploying and orchestrating multi-agent AI workflows. Upload an application bundle, and the platform automatically containerizes, scales, and manages all services.

Platform Architecture

graph TD
    subgraph Platform Layer
        UI[Web Dashboard]
        API[FastAPI Service]
        Inspector[App Inspector]
        Runtime[Runtime Generator]
        Repo[(Artifact Store<br>MongoDB + GridFS)]
        CMD[Docker Engine]
    end

    UI -->|Upload / Build / Deploy| API
    API --> Inspector
    API --> Runtime
    API --> CMD
    Runtime --> Repo
    CMD -->|docker compose| Containers

    subgraph Deployment
        Containers[Pod Containers]
        Kafka[Kafka Event Bus]
    end

Loading

The platform follows a linear Inspect → Build → Deploy pipeline:

1. Inspect — Validates the uploaded app zip and discovers components from config.yaml.
2. Build — Generates Dockerfiles and Compose specs, builds images, and stores artifacts in MongoDB/GridFS.
3. Deploy — Restores images from the artifact store and runs docker compose up across pods.

Application Workflow

graph LR
    Input[Email Input] --> Orchestrator

    subgraph Agentic App
        Orchestrator[Orchestrator<br>LangGraph FSM]

        Orchestrator --> Classify[Classify Agent]
        Orchestrator --> Fetch[Fetch Agent]
        Orchestrator --> Respond[Respond Agent]

        Fetch -.->|Kafka| IMAP[IMAP Tool]
        Respond -.->|Kafka| Calendar[Calendar Tool]
        Respond -.->|Kafka| Draft[Draft Tool]
    end

    Respond --> Output[Reply + Trace]

Loading

The reference app (InboxPilot) is an email-response automation workflow:

• Classify Agent — LLM-based email categorization (meeting / billing / general).
• Fetch Agent — Retrieves thread context via IMAP.
• Respond Agent — Generates a reply draft with an available calendar slot.

All inter-service communication happens over Kafka. The platform is app-agnostic — any app following the config.yaml contract can be deployed.

Repository Structure

├── Platform/          # Deployment control plane, runtime engine, web UI
├── App/               # Reference agentic app (InboxPilot)
├── docker-compose.yml # Containerized platform + MongoDB
├── run.sh             # Local startup script
├── build.sh           # Build and deploy helper
└── .env.example       # Environment variable template

Getting Started

Prerequisites

• Docker & Docker Compose
• Python 3.11+

Run with Docker Compose

docker compose up --build

Platform UI will be available at http://localhost:8000.

Run Locally

# 1. Set up environment variables
cp .env.example .env
# Edit .env with your GROQ_API_KEY

# 2. Start the platform (launches MongoDB + FastAPI)
bash run.sh

Deploy an App

1. Open the Platform UI.
2. Upload the app zip (e.g., App/).
3. Inspect → configure ports → Build → Deploy.
4. Access the deployed app from the status page.

Contributors

Name	Role