AIX: OpenAI: disable reasoning summaries when disabling Streaming as well. Fixes #932

src/common/stores/llms/llms.parameters.ts

@@ -69,6 +69,12 @@ export const DModelParameterRegistry = {
   /**
    * First introduced as a user-configurable parameter for the 'Verification' required by o3.
    * [2025-04-16] Adding parameter to disable streaming for o3, and possibly more models.
+   *
+   * [2026-01-21] OpenAI Responses API: Reasoning Summaries require organization verification.
+   * Per OpenAI docs, both streaming AND reasoning summaries require org verification for GPT-5/5.1/5.2.
+   *  - https://help.openai.com/en/articles/10362446-api-model-availability-by-usage-tier-and-verification-status
+   *  - Rather than adding a separate param, we piggyback on llmForceNoStream.
+   *  - AIX Wire type `vndOaiReasoningSummary` is derived from `llmForceNoStream` in aix.client.ts.
    */
   llmForceNoStream: {
     label: 'Disable Streaming',

src/modules/aix/client/aix.client.ts

@@ -46,7 +46,7 @@ export function aixCreateModelFromLLMOptions(
 
   // destructure input with the overrides
   const {
-    llmRef, llmTemperature, llmResponseTokens, llmTopP,
+    llmRef, llmTemperature, llmResponseTokens, llmTopP, llmForceNoStream,
     llmVndAnt1MContext, llmVndAntSkills, llmVndAntThinkingBudget, llmVndAntWebFetch, llmVndAntWebSearch, llmVndAntEffort,
     llmVndGeminiAspectRatio, llmVndGeminiImageSize, llmVndGeminiCodeExecution, llmVndGeminiComputerUse, llmVndGeminiGoogleSearch, llmVndGeminiMediaResolution, llmVndGeminiShowThoughts, llmVndGeminiThinkingBudget, llmVndGeminiThinkingLevel, llmVndGeminiThinkingLevel4,
     // llmVndMoonshotWebSearch,
@@ -100,6 +100,7 @@ export function aixCreateModelFromLLMOptions(
     ...(hotfixOmitTemperature ? { temperature: null } : llmTemperature !== undefined ? { temperature: llmTemperature } : {}),
     ...(llmResponseTokens /* null: similar to undefined, will omit the value */ ? { maxTokens: llmResponseTokens } : {}),
     ...(llmTopP !== undefined ? { topP: llmTopP } : {}),
+    ...(llmForceNoStream ? { forceNoStream: true } : {}),
     ...(llmVndAntThinkingBudget !== undefined ? { vndAntThinkingBudget: llmVndAntThinkingBudget } : {}),
     ...(llmVndAnt1MContext ? { vndAnt1MContext: llmVndAnt1MContext } : {}),
     ...(llmVndAntSkills ? { vndAntSkills: llmVndAntSkills } : {}),
@@ -121,7 +122,10 @@ export function aixCreateModelFromLLMOptions(
     // ...(llmVndGeminiUrlContext === 'auto' ? { vndGeminiUrlContext: llmVndGeminiUrlContext } : {}),
     // ...(llmVndMoonshotWebSearch === 'auto' ? { vndMoonshotWebSearch: 'auto' } : {}),
     ...(llmVndOaiResponsesAPI ? { vndOaiResponsesAPI: true } : {}),
-    ...((llmVndOaiReasoningEffort52Pro || llmVndOaiReasoningEffort52 || llmVndOaiReasoningEffort4 || llmVndOaiReasoningEffort) ? { vndOaiReasoningEffort: llmVndOaiReasoningEffort52Pro || llmVndOaiReasoningEffort52 || llmVndOaiReasoningEffort4 || llmVndOaiReasoningEffort } : {}),
+    ...((llmVndOaiReasoningEffort52Pro || llmVndOaiReasoningEffort52 || llmVndOaiReasoningEffort4 || llmVndOaiReasoningEffort) ? {
+      vndOaiReasoningEffort: llmVndOaiReasoningEffort52Pro || llmVndOaiReasoningEffort52 || llmVndOaiReasoningEffort4 || llmVndOaiReasoningEffort,
+      vndOaiReasoningSummary: llmForceNoStream ? 'none' /* we disable the summaries, to not require org verification */ : 'detailed',
+    } : {}),
     ...(llmVndOaiRestoreMarkdown ? { vndOaiRestoreMarkdown: llmVndOaiRestoreMarkdown } : {}),
     ...(llmVndOaiVerbosity ? { vndOaiVerbosity: llmVndOaiVerbosity } : {}),
     ...(llmVndOaiWebSearchContext ? { vndOaiWebSearchContext: llmVndOaiWebSearchContext } : {}),

src/modules/aix/server/api/aix.wiretypes.ts

@@ -472,12 +472,13 @@ export namespace AixWire_API {
     // Moonshot
     vndMoonshotWebSearch: z.enum(['auto']).optional(),
     // OpenAI
+    vndOaiImageGeneration: z.enum(['mq', 'hq', 'hq_edit', 'hq_png']).optional(),
     vndOaiResponsesAPI: z.boolean().optional(),
     vndOaiReasoningEffort: z.enum(['none', 'minimal', 'low', 'medium', 'high', 'xhigh']).optional(),
+    vndOaiReasoningSummary: z.enum(['none', 'detailed']).optional(),
     vndOaiRestoreMarkdown: z.boolean().optional(),
     vndOaiVerbosity: z.enum(['low', 'medium', 'high']).optional(),
     vndOaiWebSearchContext: z.enum(['low', 'medium', 'high']).optional(),
-    vndOaiImageGeneration: z.enum(['mq', 'hq', 'hq_edit', 'hq_png']).optional(),
     // OpenRouter
     vndOrtWebSearch: z.enum(['auto']).optional(),
     // Perplexity

src/modules/aix/server/dispatch/chatGenerate/adapters/openai.responsesCreate.ts

@@ -38,7 +38,6 @@ export function aixToOpenAIResponses(
   const isOpenAIOFamily = ['gpt-6', 'gpt-5', 'o4', 'o3', 'o1'].some(_id => model.id === _id || model.id.startsWith(_id + '-'));
   const isOpenAIChatGPT = ['gpt-5-chat'].some(_id => model.id === _id || model.id.startsWith(_id + '-'));
   const isOpenAIComputerUse = model.id.includes('computer-use');
-  const isOpenAIO1Pro = model.id === 'o1-pro' || model.id.startsWith('o1-pro-');
 
   const hotFixNoTemperature = isOpenAIOFamily && !isOpenAIChatGPT;
   const hotFixNoTruncateAuto = isOpenAIComputerUse;
@@ -75,7 +74,8 @@ export function aixToOpenAIResponses(
     // Operations Config
     reasoning: !model.vndOaiReasoningEffort ? undefined : {
       effort: model.vndOaiReasoningEffort,
-      summary: !isOpenAIO1Pro ? 'detailed' : 'auto', // elevated from 'auto' (o1-pro still at 'auto')
+      // 'none' = omit (for unverified orgs), 'detailed' = explicit, undefined = default per model
+      ...(model.vndOaiReasoningSummary !== 'none' ? { summary: model.vndOaiReasoningSummary } : {}),
     },
 
     // Output Config

src/modules/llms/models-modal/LLMParametersEditor.tsx

@@ -732,8 +732,8 @@ export function LLMParametersEditor(props: {
     {showParam('llmForceNoStream') && (
       <FormSwitchControl
         title='Disable Streaming'
-        description='Receive complete responses'
-        tooltip='Turn on to get entire responses at once. Useful for models with streaming issues, but will make responses appear slower.'
+        description='For unverified OpenAI orgs'
+        tooltip='Disables streaming and reasoning summaries, which both require OpenAI organization verification. Enable if you get verification errors with GPT-5 models.'
         checked={!!llmForceNoStream}
         onChange={checked => {
           if (!checked)

src/modules/llms/server/openai/models/openai.models.ts

@@ -545,7 +545,7 @@ export const _knownOpenAIChatModels: ManualMappings = [
     maxCompletionTokens: 100000,
     trainingDataCutoff: 'Sep 30, 2023',
     interfaces: [LLM_IF_OAI_Responses, ...IFS_CHAT_MIN, LLM_IF_OAI_Reasoning, LLM_IF_HOTFIX_NoTemperature],
-    parameterSpecs: [{ paramId: 'llmVndOaiReasoningEffort' }],
+    parameterSpecs: [{ paramId: 'llmVndOaiReasoningEffort' }, { paramId: 'llmForceNoStream' }],
     chatPrice: { input: 150, output: 600 },
     // benchmark: has not been measured yet by third parties
   },