Is your feature request related to a problem? Please describe.
When the new AI Agent connector types and element templates ship, existing BPMN diagrams reference the old template IDs and carry the legacy provider configuration shape. Customers can keep running those diagrams via the deprecated input-rewrite shim, but they have no automated way to migrate the BPMN files themselves to the new template format.
Describe the solution you'd like
Ship a small migration tool — either a c8ctl subcommand, a Camunda Desktop Modeler plugin, or both — that rewrites BPMN files in place:
- Replace the
modelerTemplate reference from the old element template ID to the new one (io.camunda.connectors.agenticai.aiagent.task.v2 / .subprocess.v2).
- Rewrite the input mappings to the new
ProviderConfiguration shape: collapse openai / azureOpenAi / openaiCompatible into openai{backend: openai|foundry|custom}, project googleVertexAi → googleGenAi{backend: vertex}, split bedrock with Anthropic model IDs into anthropic{backend: bedrock}.
- Leave existing running process instances untouched — they continue to run through the deprecated shim until their diagram is migrated and redeployed.
Describe alternatives you've considered
Leaving migration entirely manual; users would have to re-create their AI Agent task / sub-process from scratch on the new template.
Additional context
The Jackson migration deserializer landing alongside the new ProviderConfiguration shape covers the same field transformations; its rules can be reused or adapted for the BPMN-level rewrite.
Is your feature request related to a problem? Please describe.
When the new AI Agent connector types and element templates ship, existing BPMN diagrams reference the old template IDs and carry the legacy provider configuration shape. Customers can keep running those diagrams via the deprecated input-rewrite shim, but they have no automated way to migrate the BPMN files themselves to the new template format.
Describe the solution you'd like
Ship a small migration tool — either a
c8ctlsubcommand, a Camunda Desktop Modeler plugin, or both — that rewrites BPMN files in place:modelerTemplatereference from the old element template ID to the new one (io.camunda.connectors.agenticai.aiagent.task.v2/.subprocess.v2).ProviderConfigurationshape: collapseopenai/azureOpenAi/openaiCompatibleintoopenai{backend: openai|foundry|custom}, projectgoogleVertexAi→googleGenAi{backend: vertex}, splitbedrockwith Anthropic model IDs intoanthropic{backend: bedrock}.Describe alternatives you've considered
Leaving migration entirely manual; users would have to re-create their AI Agent task / sub-process from scratch on the new template.
Additional context
The Jackson migration deserializer landing alongside the new
ProviderConfigurationshape covers the same field transformations; its rules can be reused or adapted for the BPMN-level rewrite.