Summary
Follow-up plan after the Slim LLM, Tool, and Code DSL import work. Current default DSL loading supports start, end, answer, if-else, template-transform, code, llm, and tool. This issue tracks the remaining node coverage by implementation shape.
Category 1: local node factory support
These nodes are mostly Graphon-local and should be the first expansion target:
http-requestvariable-aggregatorassignerlist-operator
Goal: add DSL factory creation, payload normalization only where needed, and focused load/run tests.
Category 2: Slim LLM-backed nodes
These nodes can reuse the existing Slim LLM runtime path, but need node-specific wiring:
question-classifierparameter-extractor
Goal: resolve model provider credentials through the same DSL Slim path as LLM nodes, then add import and execution tests.
Category 3: container nodes
These need recursive graph import and nested execution semantics before they can be considered loadable:
iterationiteration-startlooploop-startloop-end
Goal: define child graph normalization, internal start/end handling, variable scope, and error handling before enabling them in the default import whitelist.
Out of scope for now
These nodes do not have a clear support plan yet:
knowledge-retrievaldatasourceagenthuman-inputdocument-extractorvariable-assigner
Their runtime boundaries are still unclear or their implementation is incomplete in Graphon. More specifically, the node logic has not yet been consolidated into Graphon and mostly still remains in Dify, so this issue does not schedule support for them.
Acceptance criteria
- The default DSL import plan reports unsupported node types accurately.
- New node support is added incrementally with import tests and at least one execution-path test per node family.
- Documentation or examples are updated when a newly supported node changes the public DSL import surface.
Summary
Follow-up plan after the Slim LLM, Tool, and Code DSL import work. Current default DSL loading supports
start,end,answer,if-else,template-transform,code,llm, andtool. This issue tracks the remaining node coverage by implementation shape.Category 1: local node factory support
These nodes are mostly Graphon-local and should be the first expansion target:
http-requestvariable-aggregatorassignerlist-operatorGoal: add DSL factory creation, payload normalization only where needed, and focused load/run tests.
Category 2: Slim LLM-backed nodes
These nodes can reuse the existing Slim LLM runtime path, but need node-specific wiring:
question-classifierparameter-extractorGoal: resolve model provider credentials through the same DSL Slim path as LLM nodes, then add import and execution tests.
Category 3: container nodes
These need recursive graph import and nested execution semantics before they can be considered loadable:
iterationiteration-startlooploop-startloop-endGoal: define child graph normalization, internal start/end handling, variable scope, and error handling before enabling them in the default import whitelist.
Out of scope for now
These nodes do not have a clear support plan yet:
knowledge-retrievaldatasourceagenthuman-inputdocument-extractorvariable-assignerTheir runtime boundaries are still unclear or their implementation is incomplete in Graphon. More specifically, the node logic has not yet been consolidated into Graphon and mostly still remains in Dify, so this issue does not schedule support for them.
Acceptance criteria