[ { "id": "F001", "description": "Workflow Designer exposes an AI grouped palette tile backed by the shared designer action catalog", "implemented": true, "prdRefs": [ "Summary", "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F002", "description": "The shared Workflow V2 action registry exposes a first-class AI inference action such as `ai.infer`", "implemented": true, "prdRefs": [ "Requirements > Functional Requirements", "Data / API / Integrations" ] }, { "id": "F003", "description": "AI inference steps remain modeled as `action.call` rather than introducing a new workflow node type", "implemented": true, "prdRefs": [ "Summary", "Goals", "Non-goals" ] }, { "id": "F004", "description": "Workflow authors can provide the AI prompt through the existing action input mapping authoring flow", "implemented": true, "prdRefs": [ "Users and Primary Flows", "Requirements > Functional Requirements" ] }, { "id": "F005", "description": "AI step config persists an inline output schema and schema-authoring mode per step", "implemented": true, "prdRefs": [ "Goals", "Requirements > Functional Requirements", "Data / API / Integrations" ] }, { "id": "F006", "description": "The AI step properties panel includes a dedicated output schema authoring section in addition to the normal grouped action and save-as controls", "implemented": true, "prdRefs": [ "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F007", "description": "Simple schema mode lets authors add and remove top-level object fields for AI output", "implemented": true, "prdRefs": [ "Users and Primary Flows", "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F008", "description": "Simple schema mode supports primitive field types, required flags, and field descriptions", "implemented": true, "prdRefs": [ "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F009", "description": "Simple schema mode supports nested object fields inside the root output object", "implemented": true, "prdRefs": [ "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F010", "description": "Simple schema mode supports arrays of primitives and arrays of objects inside the root output object", "implemented": true, "prdRefs": [ "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F011", "description": "The designer generates canonical inline JSON Schema from the simple schema builder and stores it on the AI step", "implemented": true, "prdRefs": [ "Goals", "Requirements > Functional Requirements", "Data / API / Integrations" ] }, { "id": "F012", "description": "Advanced mode lets authors edit raw inline JSON Schema for the AI step output", "implemented": true, "prdRefs": [ "Users and Primary Flows", "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F013", "description": "The designer can rehydrate simple mode from saved inline AI schemas that fit the supported simple-builder subset", "implemented": true, "prdRefs": [ "Users and Primary Flows", "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F014", "description": "The designer falls back to advanced mode without data loss when a saved inline AI schema is outside the supported simple-builder subset", "implemented": true, "prdRefs": [ "Users and Primary Flows", "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F015", "description": "The properties panel shows the resolved AI output schema preview rather than only the static registry schema for the AI action", "implemented": true, "prdRefs": [ "UX / UI Notes", "Requirements > Functional Requirements" ] }, { "id": "F016", "description": "Workflow designer data-context construction resolves an AI step's output schema from step config rather than from the action registry's static output schema", "implemented": true, "prdRefs": [ "Problem", "Goals", "Requirements > Functional Requirements" ] }, { "id": "F017", "description": "Downstream reference browsing, mapping, and expression autocomplete expose AI output fields through normal `vars.` behavior", "implemented": true, "prdRefs": [ "Goals", "Users and Primary Flows", "Requirements > Functional Requirements" ] }, { "id": "F018", "description": "Publish validation accepts AI step config only when the inline output schema is valid and within the supported subset", "implemented": true, "prdRefs": [ "Requirements > Functional Requirements", "Security / Permissions", "Acceptance Criteria (Definition of Done)" ] }, { "id": "F019", "description": "Publish-time type inference uses the resolved per-step AI output schema for downstream `vars` typing and compatibility checks", "implemented": true, "prdRefs": [ "Problem", "Requirements > Functional Requirements" ] }, { "id": "F020", "description": "Workflow save/load and edit flows preserve AI prompt, schema mode, and inline output schema configuration", "implemented": true, "prdRefs": [ "Users and Primary Flows", "Requirements > Functional Requirements", "Acceptance Criteria (Definition of Done)" ] }, { "id": "F021", "description": "Workflow bundle export, import, and definition persistence preserve inline AI schema config without introducing schema-ref dependencies", "implemented": true, "prdRefs": [ "Requirements > Functional Requirements", "Rollout / Migration", "Data / API / Integrations" ] }, { "id": "F022", "description": "A reusable server-side workflow inference service is built on top of the current OpenAI-compatible provider resolver", "implemented": true, "prdRefs": [ "Goals", "Requirements > Functional Requirements", "Data / API / Integrations" ] }, { "id": "F023", "description": "The workflow inference service submits structured-output requests using the step's inline JSON Schema as the declared output contract", "implemented": true, "prdRefs": [ "Goals", "Requirements > Functional Requirements" ] }, { "id": "F024", "description": "The workflow inference service validates model output against the inline AI schema before returning it to workflow runtime", "implemented": true, "prdRefs": [ "Goals", "Requirements > Functional Requirements", "Acceptance Criteria (Definition of Done)" ] }, { "id": "F025", "description": "The workflow inference service normalizes transient provider, provider-configuration, and schema-validation failures into actionable workflow errors", "implemented": true, "prdRefs": [ "Requirements > Functional Requirements", "Observability" ] }, { "id": "F026", "description": "The `ai.infer` action handler invokes the workflow inference service and returns structured results through the normal `action.call` runtime path", "implemented": true, "prdRefs": [ "Data / API / Integrations", "Requirements > Functional Requirements" ] }, { "id": "F027", "description": "AI action outputs are stored via existing `saveAs` semantics and become available to later steps exactly like other saved action outputs", "implemented": true, "prdRefs": [ "Goals", "Requirements > Functional Requirements", "Acceptance Criteria (Definition of Done)" ] }, { "id": "F028", "description": "V1 uses the current app-level provider and model resolution with no per-step provider or model selector", "implemented": true, "prdRefs": [ "Non-goals", "Security / Permissions" ] }, { "id": "F029", "description": "Existing non-AI workflows and existing grouped action flows continue to load, publish, and execute unchanged after AI support is added", "implemented": true, "prdRefs": [ "Goals", "Rollout / Migration", "Acceptance Criteria (Definition of Done)" ] } ]