Tactical radar scene with AI command nodes over a dark battlefield
BY SASZEL STUDIO · AI BATTLEFIELD COMMAND

DCS World without the Mission Editor
with AI command of forces.

IASER (IAgent State Architecture for Strategy Engineering and Rebalancing) is a DCS World distribution and cluster solution that turns a stock Eagle Dynamic Dedicated Server into an autonomous battlefield. It adds a War Manager that owns economy and strategy, plus an AI offloading system that keeps the heavy compute off the DCS tick-loop.

Explore IASER Read the docs →

Runs alongside

Eagle Dynamic DCSWar ManagerRay / CeleryFastAPIRedisPyTorchPrometheus · Grafana
THE PROBLEM

DCS is a physics engine — not a strategic mind.

DCS World Dedicated Server is excellent at realistic flight, ballistics, damage, and per-unit state. It is not built for long-horizon AI command: grand-strategy planning, an economy that produces and consumes fuel, ammo, and funds, or a command-and-control mind that decides what to attack and when.

Running a full strategic mind inside the DCS Lua loop also stalls the sim — expensive planning blocks the tick, couples gameplay and AI, and cannot spread work across machines.

IASER fixes all three by splitting the problem into two cleanly decoupled halves joined by an asynchronous bridge.

IASER architecture: game layer vs AI layer
THE THESIS

Separate game. Separate mind.

Two cleanly decoupled halves, joined by an asynchronous bridge. The game layer keeps doing what DCS does best; the AI layer does what DCS cannot.

01

Game layer

Physics, visuals, unit state. Runs on the unmodified Eagle Dynamic DCS Dedicated Server.

02

AI layer

Economy, strategy, command, heavy compute. Runs on a Python cluster (War Manager + offloaders).

03

Async bridge

Lua + Redis Streams. Non-blocking telemetry and control — DCS never blocks the tick-loop.

04

Control layer

REST / gRPC API, SDKs, dashboards. Mission control, monitoring, external integrations.

THE BRAIN

The War Manager owns the war.

The War Manager is the Autonomous Battlefield Manager (ABM) — a Python core that commands entire theaters. It owns the economy and the strategy, and issues high-level command through a layered AI.

  • Economy Engine — resource generation, production, logistics. Funds, fuel, ammo, factories.
  • Strategy Planner — MCTS-based operational planning with dynamic re-planning.
  • Force Commander — when to attack, retreat, or reinforce. Layered decision trees + learned heuristics.
  • Recon & Intel — Bayesian inference from telemetry to estimate enemy strength.
  • Reactive AI — low-latency emergency response: SAM suppression, CAS, threats.

The planning loop runs every 10–30 seconds: recon → economy → strategy → force command → logistics.

War Manager core modules
THE PIPELINE

Offload the heavy compute — keep DCS responsive.

The AI offloading system is a scalable cluster of worker processes that absorb heavy computation — pathfinding, low-level tactics, physics-aware actions — off the DCS tick-loop.

Communication is non-blocking: the DCS Lua bridge pushes tasks onto a Redis Streams queue, a FastAPI + Ray / Celery worker pool crunches them, and results are ingested safely through a Lua bridge with timeouts and bounded payloads.

  • Redis Streams queue — XADD/XREAD, non-blocking, persistent.
  • Worker pool — auto-scaling, load-balanced, distributed across machines.
  • Lua bridge — safe result ingestion, bounded payloads, timeouts.
AI offloading pipeline
WHY IASER

From scripted triggers to a strategic mind

How IASER compares to a traditional DCS mod.

ConcernTraditional DCS modIASER
AI CommandScripted triggerAction, static scriptsWar Manager + layered AI, dynamic re-planning
EconomyManual / noneFirst-class engine: funds, factories, logistics
StrategyNoneMCTS-based operational planning with re-planning
Compute LoadRuns inside DCS Lua (blocks sim)Offloaded to a cluster worker pool (non-blocking)
ScaleSingle machineCluster solution, auto-scaling workers
SYSTEM LAYERS

Five layers, one battlefield

From the game layer up to the control layer.

  1. 01

    Game layer

    DCS Dedicated Server (unmodified ED build) — physics, visuals, unit state.

  2. 02

    AI layer

    War Manager (ABM core) — Python + PyTorch + Ray. Strategy, economy, high-level command.

  3. 03

    Offload layer

    AI Worker Pool — FastAPI + Redis + Celery. Pathfinding, low-level tactics, physics-aware actions.

  4. 04

    Integration layer

    DCS Bridge — Lua + Redis Streams. Asynchronous telemetry & control.

  5. 05

    Control layer

    REST/gRPC API — FastAPI + Protobuf. Mission control, monitoring, external integrations.

// War Manager state schema (MongoDB / SQLite)
{
  "theater": "Caucasus",
  "factions": {
    "red": {
      "funds": 1250000,
      "factories": { "air": 3, "armor": 2, "logistics": 4 },
      "forces": [
        { "type": "Su-27", "available": 12, "deployed": 8 },
        { "type": "T-90",   "available": 40, "deployed": 24 }
      ]
    },
    "blue": { /* symmetric */ }
  },
  "missions": [
    { "id": "M-001", "type": "strike",
      "target": "Airfield 3", "status": "active" }
  ],
  "recon": {
    "enemy_air_strength": "high",
    "estimated_frontline": "lat 45.2, lon 40.1"
  }
}
DOCUMENTATION

Every module, mapped

The full IASER specification — ten sections, thirty documents. Explore the architecture, deployment, and roadmap.

Introduction4 docs

What IASER is

What IASER is, goals & non-goals, architecture overview, and design principles.

Read →
Distribution6 docs

The IASER distribution

The DCS distribution, cluster solution, Eagle Dynamic setup, mission & script pack, economy config, and telemetry wiring.

Read →
War Manager7 docs

Economy, strategy, command

War Manager overview, economy engine, strategy planner, force command, recon & intel, reactive AI, and the lifecycle loop.

Read →
AI Offloading4 docs

Offload pipeline

Offloading overview, task taxonomy, worker pool, and scheduling & load balancing.

Read →
DCS Bridge1 doc

Lua bridge

The Lua bridge, Redis Streams, event hooks, and the safety model that keeps DCS non-blocking.

Read →
API4 docs

REST, gRPC, SDKs

API overview, REST endpoints, gRPC contracts, and SDKs for external integrations.

Read →
Deployment4 docs

Infrastructure & scale

Infrastructure stack, Docker Compose, Kubernetes, and the scaling strategy.

Read →
Data & Models4 docs

Data model, ML, telemetry

Data model, machine learning models, the telemetry pipeline, and the state store.

Read →
Operations3 docs

Observability & runbooks

Monitoring, observability, troubleshooting, and operational runbooks.

Read →
Roadmap3 docs

Product roadmap & MVP

The product roadmap, MVP phases, and the prioritized requirements backlog.

Read →
Full documentation index →
THE JOURNEY

From foundation to production

Seven stages — from a documentation scaffold to a launched, monitored cluster.

0

Foundation

Project scaffolding, docs, version control, conventions.

1

Design & architecture

Freeze architecture, data model, contracts, deployment targets.

2

Build — AI layer

War Manager core: economy, strategy, recon, reactive AI.

3

Build — Offloading & bridge

Lua bridge, worker pool, end-to-end offload pipeline.

4

Build — Control & deployment

REST/gRPC API, SDKs, Docker Compose, Kubernetes.

5

Integration, test & hardening

Demo, performance, reliability, security, docs.

6

Production launch & operations

Release, monitor, operate, iterate.

QUESTIONS

Common questions

What is IASER?

IASER is a DCS World distribution and cluster solution by Saszel Studio. It turns a stock Eagle Dynamic Dedicated Server into an autonomous battlefield with an AI War Manager (economy + strategy) and an AI offloading system that keeps heavy compute off the DCS tick-loop.

Does IASER replace DCS?

No. IASER runs alongside an unmodified Eagle Dynamic DCS Dedicated Server. The game layer stays lean; the AI layer owns economy, strategy, and command through a lightweight, asynchronous Lua bridge.

How does IASER relate to the Saszel Framework?

The Saszel Framework is our custom multiplayer engine for large-scale, human-player battles. IASER is a separate product from Saszel Studio — an AI distribution that adds autonomous command, economy, and strategy to DCS Dedicated Servers. Both are built by Saszel Studio.

Is IASER runnable yet?

At this stage IASER is a documentation-first specification — the full architecture, War Manager, offloading, deployment, and roadmap are documented. Application source code is on the roadmap for later production stages.

Can I host my own IASER server?

Yes — the deployment docs cover Docker Compose for local dev and Kubernetes for production, with an auto-scaling worker pool strategy. The full specification is available in the documentation.

READY TO COMMAND?

Explore IASER

Dive into the full specification — architecture, War Manager, offloading, deployment, and roadmap.

Read the docs Join the mission