Plasma — Open-Source Real-Time Intelligence Platform

Core Concepts

How Plasma works

Plasma implements standardized building blocks for real-time intelligence platforms. Ontology organizes information. Topology organizes computation. Applications interface the world. Agents react to it. Autonomous from infrastructure to dashboard.

Nature

Sovereign

You own the full stack. No managed dependency, no external control plane, no vendor lock. Plasma provisions bare metal (OpenTofu), configures everything (Ansible), builds, and runs — from OS to dashboards. Air-gapped if you need.

# Provision — any substrate
on-premise, dedicated, VMs, cloud

# Configure — full stack
plasmactl node:provision  # OpenTofu
plasmactl platform:up     # Ansible

# Result — you own everything
bare metal → running platform

Nature

Reactive

Event-driven at every level. Events carry complete state (ECST) — consumers never query back. Each layer has its own bus tuned to its semantics. Behavior emerges from channel topology, not a central coordinator. Cross-layer data flows through explicit relays with backpressure.

# Bus per layer — tuned semantics
integration: NATS   # business events
cognition:   Kafka  # analytics streams
conversation: Matrix # utterances

# Cross-layer via relays only
relay: Bento # backpressure

Structure

Ontology

Organizes information. Entities define WHAT exists, Metrics define HOW to measure. Together they form the stable data contract between all systems. Protobuf schemas are the source of truth, enabling automatic schema evolution.

# Entity — describes the world
entity: Person, Project, Utterance

# Metric — measures the world
metric: Engagement, Sentiment, Risk

# Schema = contract (protobuf)
package platform.person;

Structure

Topology

Organizes computation. Topology sections map logical architecture to physical resources. Nodes allocate to topology, Agents and Applications attach to it. Each section gets the right compute profile.

# Topology distributes to components
cognition.data    → GPU nodes
foundation.kv    → storage nodes
interaction.obs  → balanced nodes

# Nodes allocate, components attach
node → topology ← {app, agent}

Behavior

Application → Service → Software

Applications interface the world. An Application orchestrates WHY (business purpose), Services configure WHAT (deployment context), Software defines HOW (programs).

# Application (WHY) — orchestrates
dashboards: monitoring system

# Service (WHAT) — configures
dashboards-grafana: prometheus + oauth2

# Software (HOW) — computes
grafana: visualize time-series data

Behavior

Agent → Skill → Function

Agents react to the world. An Agent decides WHEN (event-driven triggers), Skills configure WHAT (parameters), Functions define HOW (generic computation). One function serves many use cases through different skills.

# Function (HOW) — generic computation
score(input, model, threshold)

# Skills (WHAT) — one function, many configs
risk_scoring: bayesian + threshold 0.8
quality_gate: coverage + threshold 80%

# Agent (WHEN) — event-driven
sprint_reviewer: on sprint_completed

Ecosystem

Packages, models, and services

Packages bundle components for a domain. Models compose packages into platforms. Integrators deploy them. Managed services run them.

Packages

1 Co plasma-core

plasma-core soon

Foundation, Integration bus, Conversation — the kernel everything depends on. Kubernetes, IAM, NATS, mail bridge, observability.

2 Wk plasma-work

plasma-work soon

Project management, team messaging, calendar, video calls, initiative tracking. Plane, Element, Jitsi.

3 Da plasma-data

plasma-data soon

Analytics pipeline, data lake, OLAP engine, dashboards. Spark, Kafka, Druid, Superset.

4 Tk plasma-talk

plasma-talk soon

AI conversational interface. MCP server for natural language interaction with your platform.

5 Cd plasma-code

plasma-code soon

Git repositories, CI/CD pipelines, code review, artifact management. GitLab integration.

6 Ln plasma-learn

plasma-learn soon

Learning paths, knowledge base, skill assessments, certifications. Moodle LMS.

7 Lb plasma-lab

plasma-lab soon

Low-code experimentation — data connectors, workflow automation, spreadsheet database. Airbyte, n8n, NocoDB.

Models

7 Ta The Agency

soon

A complete digital operating system for service companies. Manages projects, tracks time, handles invoicing, analyzes team communication, and runs CI/CD — all on your own infrastructure.

plasma-core plasma-work plasma-data plasma-talk plasma-code plasma-learn plasma-lab

Integrators

Skilld

Plasma creator. Turns recurring operational decisions into autonomous, auditable responses — from situation identification to live execution in 6 weeks.

Deploy Plasma for your clients?

Become an integrator →

Managed services