ecosystem.Ai Platform 2.0
Build Your Own Contextual Interactions — Inside Your Cloud Perimeter.
ecosystem.Ai Platform 2.0 combines behavioral social science with real-time machine learning and agentic operations so enterprises can score, personalize, and learn from every customer interaction in the moment. Deployed entirely inside your Kubernetes 1.37 cluster on Microsoft Azure AKS, Amazon EKS, Google Cloud GKE, or Self-Hosted Kubernetes / Docker Compose (supporting both ARM64 and x86_64 AMD64 worker nodes, with gpt-6-luna configured as the default generative model across all containers), subscriber features, models, and interaction data remain inside your cloud perimeter.
Purpose-built for telecommunications (CVM, next-best-offer, and MVNOs), retail and commercial banking, insurance, retail, and gaming, the platform covers Intelligent Discovery, Smart Engagement, Intelligent Activation, and Proactive Retention.
Complete 10-Container Architecture Explained
| # | Container / Service | Ports & Endpoints | Role in the Platform |
|---|---|---|---|
| 1 | ecosystem-workbench2ecosystemai/ecosystem-workbench2 | :5270 SPA & Networks UI:8001 Private FastAPI:8004 Public API Gateway:9090 Offline Scorer | Primary operator workbench. Hosts Projects, Offer Matrices, Dynamic Engagement, Ontology-mapped data queries, Interactive Analytics Dashboards (/analytics/), Computational Social Science Network Graphs (/networks/), Spend & Money Personalities, Simulations Lab, Runtime Console, Runtime Testing, and the Ecosystem Agent. |
| 2 | ecosystem-serverecosystemai/ecosystem-server | :3001 Core Server API:54445 Embedded MongoDB 9.0:27028 mongot Vector Search:54321 H2O AutoML Engine | Core Java/Spring modeling, feature engineering, PlantUML/EcosystemCode generation, and central persistence engine. Runs self-contained MongoDB 9.0 replica set (ecosystem-server:54445), mongot community vector search, and H2O 3.46 AutoML for static model training. |
| 3 | ecosystem-runtime-soloecosystemai/ecosystem-runtime-solo | :8091 Real-Time ScorerPOST /invocationsPOST /responsePOST /mcp | Horizontally autoscaled (HPA 1..N replicas) Java 25 low-latency scoring engine. Executes pre-score feature lookups, real-time ML/bandit convergence (Epsilon Greedy, Bayesian Probabilistic, Ecosystem Rewards, Q-Learning, Two-Tower, Spend/Money/Interaction Personality), post-score eligibility plugins, closed-loop feedback via POST /response, and Model Context Protocol (POST /mcp). |
| 4 | ecosystem-campaign-managementecosystemai/ecosystem-campaign-management | :54310 Campaign UI:54311 Express REST/MCP API | Agentic Campaign Management platform. Plans, governs, schedules, activates, and measures multi-stage customer outreach campaigns wired directly to Workbench 2 projects and real-time runtime scoring. |
| 5 | ecosystem-data-agentsecosystemai/ecosystem-data-agents | :54325 Data Agents SPA:54324 FastAPI / LangGraph | Autonomous LangGraph Data Analysis Agents. Runs multi-stage extraction, preparation, categorical, actuarial risk, feature engineering, personality, recommender, time-series, marketing-mix, and financial analytics across MongoDB and Trino. |
| 6 | ecosystem-notebooksecosystemai/ecosystem-notebooks | :8010 EcoGentic Journey API:8000 JupyterHub | Agentic conversational journey server (EcoGentic) and interactive Python data-science notebook environment connected to the shared /data volume and LLM gateway. |
| 7 | ecosystem-custom-gptsecosystemai/ecosystem-custom-gpts | :54320 Specialist Agents UI:54349 Managed Agents API | Domain-grounded specialist GPT and RAG agents with dedicated knowledge bases, custom tools, and live runtime scoring integration. |
| 8 | ecosystem-domain-catalogecosystemai/ecosystem-domain-catalog | :8080 Catalog Service | Authoritative semantic catalog serving OWL/SKOS domain ontologies, SHACL validation shapes, and industry overlays (Telecommunications CVM, Banking, Insurance, Retail, Gaming). |
| 9 | trinoecosystemai/trino | :8080 Distributed SQL Engine | Distributed ANSI SQL query engine pre-configured with the MongoDB connector pointing to ecosystem-server:54445 for high-speed analytical and ontology-mapped SQL queries. |
| 10 | busyboxbusybox:1.36 | Kubernetes initContainer | Lightweight startup dependency gate that waits for ecosystem-server:3001 and embedded MongoDB ecosystem-server:54445 to accept TCP connections before dependent pods start. |
Subdomain Routing & Internal Container Communication
Every application in ecosystem.Ai Platform 2.0 runs at root path: / on its own dedicated subdomain (https://<app>.<baseDomain>/ on Azure AKS, AWS EKS, and Google Cloud GKE, or http://<app>.localhost/ locally). No sub-path prefixes are used, ensuring SPAs, static assets (/assets/...), OAuth callbacks, and API routes resolve cleanly at /:
| Service | Subdomain Host (<baseDomain> = <IP>.nip.io, <dnsZone>, or localhost) | Pure Origin Env Var | Internal Container-to-Container DNS |
|---|---|---|---|
| Workbench 2 (Operator UI & Public API) | workbench.<baseDomain> (/, /public/v1) | WORKBENCH_PUBLIC_URL, WORKBENCH_UI_URL, PUBLIC_BASE_URL | http://workbench2:5270, http://workbench2:8004/public/v1, http://workbench2:8001/api/v1 |
| Prediction Server & Modeling API | server.<baseDomain> (/, /api/v1/health) | SERVER_PUBLIC_URL, IP | http://ecosystem-server:3001, MongoDB ecosystem-server:54445, H2O http://ecosystem-server:54321 |
| Runtime 0.9.7 (Real-Time Scoring) | runtime.<baseDomain> / runtime1.<baseDomain> (/invocations, /response, /mcp, /ping) | RUNTIME_PUBLIC_URL | http://ecosystem-runtime-solo:8091 |
| Campaign Management | campaign.<baseDomain> (/) | CAMPAIGN_PUBLIC_URL | http://ecosystem-campaign:54310 |
| Data Analysis Agents | data-agents.<baseDomain> (/) | DATA_AGENTS_PUBLIC_URL | http://ecosystem-data-agents:54325 |
| EcoGentic & Notebooks | notebooks.<baseDomain> (/) | NOTEBOOKS_PUBLIC_URL, VITE_ECOGENTIC_AGENT_URL | http://ecosystem-notebooks:8010 |
| Custom GPTs / Managed Agents | custom-gpts.<baseDomain> (/) | CUSTOM_GPTS_PUBLIC_URL | http://ecosystem-custom-gpts:54320 |
| Domain Catalog | domain-catalog.<baseDomain> (/, /v1/health) | DOMAIN_CATALOG_PUBLIC_URL | http://ecosystem-domain-catalog:8080 |
| Trino SQL Engine | trino.<baseDomain> (/v1/info) | — | trino:8080 |
Operator Workbench & Container Modules — Visual Tour
1. Workbench 2 — Unified Operator Dashboard
Central launchpad across all 523+ configurations: Projects, Network Selector, Offer Matrix, Contacts, Channels, Ontology, Data Management, Static/Dynamic/Generative Models, EcoGentic, Campaigns, Analytics, and Runtime Console.
2. Projects & Deployment Configuration
End-to-end project configuration linking feature store lookups, H2O frames, dynamic engagement algorithms, offer matrices, pre/post-score Java plugins, and deployment steps.
3. Dynamic Engagement & Real-Time Bandit Learning
Configure Thompson Sampling / Bayesian Probabilistic, Epsilon Greedy, Ecosystem Rewards, Loss Aversion, Prospect Theory, and Q-Learning recommenders with contextual variables and cold-start priors.
4. Agentic Campaign Management
Author, generate, import/export, and execute multi-stage customer outreach and decisioning campaigns wired to real-time scoring and generative AI prompts.
5. Domain Ontology Catalog & Semantic Mappings
Sync OWL/SKOS/SHACL domain ontologies from ecosystem-domain-catalog, infer ontologies from live MongoDB collections, and map business terms to physical data fields.
6. EcoGentic Conversational Agent Journeys
Build and version multi-channel conversational AI agents (WhatsApp, Web, API, CLI) backed by ecosystem-notebooks (:8010) and live runtime context.
7. Simulations Lab & Convergence Testing
Run closed-loop recommender and log-replay simulations to validate bandit convergence, reward lift, and behavioral drift prior to production rollout.
8. Spend & Money Personalities Pipeline
Guided 6-stage behavioral scoring pipeline (SP-01 through SP-06) that extracts customer Spend and Money Personality traits from transaction histories.
9. Interactive Analytics Dashboards (/analytics/)
Compile and explore multi-dashlet operational, financial, personality, and runtime performance boards powered by MongoDB 9.0 and Trino SQL.
10. Computational Social Science Network Graph Analysis (/networks/)
Interactive 2D/3D graph and social network topology explorer for customer segments, credit bureau bands, payment flows, and runtime offer affinities.
