“Which branch actually ran, and why?”
Correlated graph and node identity keeps runtime evidence attached to the modeled path.
Visual intent · programmable control
Orchestrate the difficult parts.
Ravenroot turns a graph into the behavioral structure of a backend workflow. Functional specialists can shape the flow visually while engineers retain explicit contracts, policies, observability and deployment control.
- Executable GraphML
- Catalog-backed UI
- JVM application boundary
AI, governed by the graph
LLM Prompt · external provider
Bring model output into flows you can inspect.
Use an external model provider for prompt steps, then route the result through explicit shape checks, policy gates, human review and deterministic actions. Ravenroot bundles no model and arms no provider by default.
The operational pain
The process is clear—until the exceptions arrive.
An order passes normal checks, then a provider times out. A policy changes. A specialist needs to intervene. The diagram, code and operational truth drift apart exactly when the workflow becomes difficult.
“Can the specialist change intent without bypassing engineering?”
Visual authoring shares one executable GraphML contract with schemas, policies and reviewable configuration.
“Where does control return after an external failure?”
Explicit outcomes and fallback edges keep recovery paths visible instead of burying them in glue code.
One shared model
Visual enough to compose. Explicit enough to operate.
Ravenroot is a programmable no-code/low-code foundation, not a promise that engineering disappears. GraphML carries the executable workflow; runtime behavior remains inspectable and testable.
01 / AUTHOR
Graph-first workflows
Build with the browser editor or produce GraphML in your own tooling. Nodes represent transformations, controlled actions and bounded extension points.
02 / EXECUTE
One application boundary
The UI, CLI, HTTP server and embedding applications compose the same framework-neutral application contracts and graph core.
03 / OBSERVE
Correlated runtime events
Follow executions through Server-Sent Events and structured logs keyed by execution, graph version and node identity.
Architecture
A stable center with replaceable edges.
Core graph semantics stay independent from presentation, transport and actor-framework adapters.
INPUT
UI · CLI · host app
Choose the surface that fits the team.
CORE
Application + graph
Shared use cases, policies and semantics.
ENGINE
Pekko · optional Akka
Execution behind a framework-neutral SPI.
Where it fits
Complex flows without a black box.
Model personal automations, operational processes and enterprise decision flows as graphs. Connect real systems through allowed HTTP actions or extensions you explicitly provide.
Personal pattern
Collect → classify → notify
Normalize an event, evaluate a CEL decision and route a notification while keeping each step visible.
Enterprise pattern
Validate → approve → act
Make decision points and controlled external actions explicit, with live execution evidence for operators.
Release status
Open source · Apache 2.0.
Apache License 2.0 is selected for Ravenroot’s first public repository release. The current source builds locally, while the public product repository, JAR, container image and hosted sandbox remain unpublished.