LLMs: deprecate Json

src/common/stores/llms/llms.types.ts

@@ -168,6 +168,7 @@ export type DModelInterfaceV1 =
 // FIXME: keep this in sync with the server side on modules/llms/server/llm.server.types.ts
 export const LLM_IF_OAI_Chat: DModelInterfaceV1 = 'oai-chat';
 export const LLM_IF_OAI_Fn: DModelInterfaceV1 = 'oai-chat-fn';
+/** @deprecated we don't se this one anymore 2026-04-19; suspended until we have a reason or per-model continuos validation of this */
 export const LLM_IF_OAI_Json: DModelInterfaceV1 = 'oai-chat-json'; // for Structured Outputs (or JSON mode at worst)
 export const LLM_IF_ANT_ToolsSearch: DModelInterfaceV1 = 'ant-tools-search';
 // export const LLM_IF_OAI_JsonSchema: ... future?

src/modules/llms/server/llm.server.types.ts

@@ -139,7 +139,7 @@ export const ModelDescription_schema = z.object({
   updated: z.int().optional(),
   description: z.string(),
   contextWindow: z.int().nullable(),
-  interfaces: z.array(z.union([z.enum(LLMS_ALL_INTERFACES), z.string()])), // backward compatibility: to not Break client-side interface parsing on newer server
+  interfaces: z.array(z.enum(LLMS_ALL_INTERFACES).or(z.string())), // backward compatibility: to not Break client-side interface parsing on newer server
   parameterSpecs: z.array(ModelParameterSpec_schema).optional(),
   maxCompletionTokens: z.int().optional(), // initial parameter value for 'llmResponseTokens'
   // rateLimits: rateLimitsSchema.optional(),