Comparison with Other Solutions
Engine-Level Comparison
Legend: ✓ supported · △ partial / requires your own implementation · ✗ not supported
| Dimension | GFlow Engine | Activiti / Flowable | DingTalk/Lark Approval |
|---|---|---|---|
| Chinese-style approval semantics (countersign / add-sign / return / transfer / delegate / withdraw) | ✓ native, first-class | △ requires extension development | ✓ supported |
| Rule engine integration (a process is a rule chain; automation invokes any chain) | ✓ native | ✗ none | ✗ none |
| AI approval (aiAgent node) | ✓ built in | △ integrate it yourself | △ limited |
| Multi-tenant | ✓ tenant_id end to end | △ do it yourself | ✓ platform-level |
| Embedding into an existing system (library + 7 tables) | ✓ import one library | ✗ heavy middleware | ✗ not embeddable |
| Deployment form | ✓ single binary, no mandatory middleware | △ JVM application | — SaaS |
| Process definition format | RuleGo rule chain DSL (JSON) | BPMN 2.0 XML | Closed source |
| Tech stack | Go | Java | SaaS, closed source |
Why It Does Not Parse BPMN 2.0 XML
BPMN is a standard designed for drawing diagrams, while the heart of Chinese-style approval is task semantics (the parent-child chain of add-sign, countersign rules, the snapshot carried back on a return). GFlow Engine describes nodes and edges directly in JSON and puts the approval semantics into the node configuration:
- One less XML↔object-model conversion layer: the DSL is the storage format — what you see is what you get;
- Fully isomorphic with RuleGo rule chains: gateways / parallel / inclusive branches reuse the rule engine's native nodes directly;
- No information loss between the frontend designer (tree-based) and the DSL.
Why So Few Tables
The full Activiti suite easily runs to twenty or thirty tables; GFlow Engine has only 7. The reasons are covered in Data model: why only 7 tables — the ordering between tasks does not need a "transition table" to record it, because it is encoded in the rule chain's edges; candidate resolution needs no redundant expansion table — store the original references and expand them through the identity service at query time.
Relationship with the Commercial gflow Edition
The engine is free and open source (Apache-2.0); the platform is commercial at a clearly stated price. Simply put: embedding the engine into your own system is free; if you want a complete platform that business administrators can use directly, one payment gets you the full source code (the three designers — process / form / rule chain — plus the AI agent and the frontend application). See Pricing for details.