arcane-os 0.34.0 → 0.34.2

package/CHANGELOG.md

@@ -1,5 +1,26 @@
 # Changelog
 
+## 0.34.2
+
+- Remove temporary `persist:false` Chat cards on completion, failure,
+  cancellation, destruction and session replacement. Temporary input is not
+  restored into the composer after a rejected request.
+- Track each submission's exact created cards, preserving earlier persistent
+  cards and newer submissions even when operation identifiers are reused.
+  Persistent transcript and draft behavior remain unchanged; existing saved
+  history is not migrated or rewritten.
+
+## 0.34.1
+
+- Honor an optional request-local `model` in `AI.fetchRequest()` and
+  `AI.streamRequest()`. Direct and registered built-in TWiN/native Ollama
+  requests carry the supplied model without changing the shared selection.
+  Calls that omit the option retain the selected default.
+- Preserve complete messages, streaming, tools, cancellation and request
+  observers. Provider-reported stream model metadata remains authoritative;
+  generated stream completion metadata falls back to the request's model.
+  The option does not load or switch a browser-WASM model.
+
 ## 0.34.0
 
 - Add `createDigitalOceanFalTTSProvider` through `arcane-os/ai/browser-speech`

package/README.md

@@ -19,7 +19,7 @@ version-locked SDK runtime, while an integrated Arcane checkout uses its live
 `arcane/` runtime. Both profiles share the theme, packaging, event, cancellation,
 and browser run contracts while retaining their selected application layout.
 
-This checkout defines the `0.34.0` SDK contract. Applications pin one exact npm
+This checkout defines the `0.34.2` SDK contract. Applications pin one exact npm
 version and lockfile; registry state is deliberately not baked into application
 artifacts.
 
@@ -85,7 +85,7 @@ Create one browser application, install its pinned SDK, and start its source
 server:
 
 ```bash
-npx arcane-os@0.34.0 new hello-speech --path ./hello-speech --target browser
+npx arcane-os@0.34.1 new hello-speech --path ./hello-speech --target browser
 cd hello-speech
 npm install
 ```

package/docs/reference/ai/twin-cloud.md

@@ -211,6 +211,38 @@ controller's `abort()` when the operation is cancelled or its page detaches.
 `streamRequest()` is the corresponding object-form streaming API. See the
 [AI module reference](../runtime-modules.md#aijs) for its complete options.
 
+## Use a model for one browser request
+
+Both object-form methods accept `model` for that request. For example, an
+application can use 20B for document search while retaining its configured
+120B conversation model:
+
+```javascript
+const result = await ai.fetchRequest({
+  model: 'openai-gpt-oss-20b',
+  messages: searchMessages
+});
+
+await ai.streamRequest({
+  model: 'openai-gpt-oss-20b',
+  messages: summaryMessages,
+  onChunk: displaySummaryChunk
+});
+```
+
+`searchMessages`, `summaryMessages`, and `displaySummaryChunk` are supplied by
+the application. Their content and callbacks keep their ordinary contracts.
+The exact model goes into each TWiN request payload without changing `ai.model`
+or selecting another provider. Concurrent calls can choose different models;
+later calls that omit `model` (or pass `undefined`) use the configured model.
+Retries keep the same request model. Provider-returned model metadata remains
+authoritative; an assembled stream completion uses the request model only when
+the response omits it.
+
+The browser API forwards an explicitly supplied model to registered providers,
+whose own request contract determines its meaning. This does not switch or
+activate the loaded model of a browser-WASM provider.
+
 ## Migrate saved preference tuples before using them
 
 The six tuple slots consumed by `ai.setAI(...tuple)` are:

package/docs/reference/runtime-components.md

@@ -413,6 +413,15 @@ the public result term; `disposition` remains an accepted input spelling.
 When a submitted turn uses `persist:false`, Chat may show its cards while that
 one operation is active, then removes them when the operation settles; neither
 the input nor response remains in the retained transcript or model context.
+This cleanup also applies to failed, cancelled, detached, or replaced-session
+requests. Temporary input is not restored into the composer, and a temporary
+failure does not become an assistant history card. The request still rejects
+normally and keeps its diagnostic/status reporting. Cleanup owns the exact
+operation's nodes so a completion listener can start another turn safely.
+Submission code passes those card references directly when it creates them;
+reusing an operation identifier never selects earlier conversation cards for
+cleanup or timestamp updates.
+Persistent-turn recovery and existing saved history remain unchanged.
 
 For streaming sessions, every provisional structural card and the terminal
 result must preserve the same choice, ordered call position, exact ID, type,

package/docs/reference/runtime-modules.md

@@ -164,6 +164,21 @@ entry points. `streamMessage(...)` and `streamRequest(options)` deliver
 incremental responses. The positional and object forms share the existing
 provider implementations; neither form is a retired compatibility API.
 
+`fetchRequest({model,...})` and `streamRequest({model,...})` accept an optional
+request-local model. On the built-in TWiN Cloud route, its exact value becomes
+the outbound `model` field; native Ollama receives the same request-local value.
+Omitting it or passing `undefined` preserves the configured model. Each built-in
+request captures its model before asynchronous callbacks, so concurrent requests
+may use different models without assigning `ai.model`, changing provider
+selection, or loading/unloading another model. Request observers and retries
+receive that request's model. Stream/native completion metadata uses it only
+when the provider omits its own `model` field.
+
+Registered providers receive a supplied `model` unchanged in their request
+payload; an omitted value remains absent. The provider owns interpretation.
+This option does not switch a loaded browser-WASM model or its lifecycle:
+explicit provider/model selection and activation remain necessary there.
+
 Built-in cloud chat decodes an HTTP error body once as JSON or text and rejects
 with that complete value unchanged. It does not reconstruct an Error, replace
 the message, or add `providerMessage`, `status`, or an SDK failure code to the

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "arcane-os",
-  "version": "0.34.0",
+  "version": "0.34.2",
   "description": "Arcane OS JavaScript SDK, project-local CLI, browser runtime, and repository-portable application packager.",
   "type": "module",
   "main": "./src/index.mjs",
@@ -86,7 +86,7 @@
     "test": "npm run test:unit && npm run test:functional && npm run test:integration && npm run test:regression",
     "test:release": "node ./bin/arcane-test.mjs test/npm-release.test.mjs",
     "test:unit": "node ./bin/arcane-test.mjs test/app-descriptor.test.mjs test/app-schema.test.mjs test/app-selection.test.mjs test/contracts.test.mjs test/doctor.test.mjs test/mail-credentials.test.mjs test/mail-outbox.test.mjs test/mail-public-api.test.mjs test/mail-send.test.mjs test/mail-transport.test.mjs test/targets.test.mjs test/workspace-operation-lock.test.mjs",
-    "test:functional": "node ./bin/arcane-test.mjs test/ai-request-retry.test.mjs test/ai-stream-cancellation.test.mjs test/browser-speech-providers.test.mjs test/browser-wasm-gpu-notice.test.mjs test/browser-wasm-download-resume.test.mjs test/chat-stream-speech.test.mjs test/chat-transcript-scroll.test.mjs test/cli.test.mjs test/dbopfs-document-library.test.mjs test/dbopfs.test.mjs test/dev-server.test.mjs test/dev-pwa.test.mjs test/dom-event-instrumentation.test.mjs test/event-manager.test.mjs test/events.test.mjs test/import-map.test.mjs test/mail-cli.test.mjs test/mail-runtime.test.mjs test/mail-server.test.mjs test/modal.test.mjs test/packaging.test.mjs test/pwa-packaging.test.mjs test/pwa-client.test.mjs test/pwa-install.test.mjs test/pwa-worker.test.mjs test/persistent-ai-chat-session.test.mjs test/reference-completeness.test.mjs test/runtime-api-behavior.test.mjs test/runtime.test.mjs test/scaffold.test.mjs test/speech-playback.test.mjs test/site.test.mjs test/twin-cloud.test.mjs test/update-check.test.mjs test/user-explicit-updates.test.mjs",
+    "test:functional": "node ./bin/arcane-test.mjs test/ai-request-model.test.mjs test/ai-request-retry.test.mjs test/ai-stream-cancellation.test.mjs test/browser-speech-providers.test.mjs test/browser-wasm-gpu-notice.test.mjs test/browser-wasm-download-resume.test.mjs test/chat-stream-speech.test.mjs test/chat-transcript-scroll.test.mjs test/chat-transient-retention.test.mjs test/cli.test.mjs test/dbopfs-document-library.test.mjs test/dbopfs.test.mjs test/dev-server.test.mjs test/dev-pwa.test.mjs test/dom-event-instrumentation.test.mjs test/event-manager.test.mjs test/events.test.mjs test/import-map.test.mjs test/mail-cli.test.mjs test/mail-runtime.test.mjs test/mail-server.test.mjs test/modal.test.mjs test/packaging.test.mjs test/pwa-packaging.test.mjs test/pwa-client.test.mjs test/pwa-install.test.mjs test/pwa-worker.test.mjs test/persistent-ai-chat-session.test.mjs test/reference-completeness.test.mjs test/runtime-api-behavior.test.mjs test/runtime.test.mjs test/scaffold.test.mjs test/speech-playback.test.mjs test/site.test.mjs test/twin-cloud.test.mjs test/update-check.test.mjs test/user-explicit-updates.test.mjs",
     "test:integration": "node ./bin/arcane-test.mjs test/installed-package-runtime.test.mjs test/root-app-layout.test.mjs test/integrated-shared.test.mjs test/integrated-workspace.test.mjs test/mail-browser.test.mjs test/native-plan.test.mjs test/native-provider-loader.test.mjs test/npm-release.test.mjs test/release-bundle.test.mjs test/release-capability-smoke.test.mjs test/shared-payload-batch.test.mjs test/tarball.test.mjs test/browser-wasm-cpu.test.mjs test/wllama-webgpu-runtime.test.mjs",
     "test:regression": "node ./bin/arcane-test.mjs test/channel-workflows.test.mjs test/html-import-registration.test.mjs test/logging-regression.test.mjs test/markdown-speech.test.mjs test/prepared-speech.test.mjs test/native-provider-generation.test.mjs test/speech-queue-regression.test.mjs test/testing.test.mjs test/test-sets.test.mjs",
     "check": "node tools/check-source.mjs && npm test",

package/runtime/arcane/components/chat.html

@@ -3188,15 +3188,12 @@
         ].includes(error?.code);
     }
 
-    function restoreRejectedStructuralDraft(messageId,operationId,text){
+    function restoreRejectedStructuralDraft(messageId,requestMessages,text){
         const response=[...chatOutput.children].find(
             candidate=>candidate.id===`message-${messageId}`
         )||null;
-        const request=[...chatOutput.children].find(
-            candidate=>candidate.dataset.operationId===operationId
-        )||null;
         response?.remove();
-        request?.remove();
+        for(const request of requestMessages) request.remove();
         const existing=textArea.value;
         if(existing!==text){
             textArea.value=existing
@@ -3206,14 +3203,15 @@
         textArea.style.height='auto';
         textArea.style.height=`${textArea.scrollHeight}px`;
         scrollTranscriptToBottom();
-        return Boolean(response||request);
+        return Boolean(response||requestMessages.length);
     }
 
     async function sendMessageThroughBoundSession(
         text,
         context,
         sessionRequest={content:text,role:'user'},
-        perTurnRequest=null
+        perTurnRequest=null,
+        requestMessages=[]
     ){
         const session=boundChatSession;
         const bindingGeneration=sessionBindingGeneration;
@@ -3248,12 +3246,21 @@
         const streamedStructuralToolCalls=[];
         const messageId=`session-${++sessionMessageSequence}`;
         const sessionMessageToken=Symbol(messageId);
+        const transientCards=retainTurn?[]:[...requestMessages];
+        function removeTransientSessionCards(){
+            if(!transientCards.length) return;
+            // The submitter supplies only nodes created for this submission.
+            for(const card of transientCards) card.remove();
+            transientCards.length=0;
+            scrollTranscriptToBottom();
+        }
         activeSessionMessageToken=sessionMessageToken;
         sessionMessagePending=true;
         applyAIAvailability(host.aiAvailability);
         setSessionStatus('sending','AI is responding…');
         try{
-            receivedMessage(host.aiName, '', messageId);
+            const responseCard=receivedMessage(host.aiName, '', messageId);
+            if(!retainTurn) transientCards.push(responseCard);
             await streamMessage('',messageId,true);
             setMessageProgress(
                 messageId,
@@ -3337,9 +3344,6 @@
                 result.message.content,
                 committedTurn?.response?.timestamp??result.message.timestamp
             );
-            const requestMessages=[...chatOutput.children].filter(
-                candidate=>candidate.dataset.operationId===context.operationId
-            );
             for(const [index,requestMessage] of requestMessages.entries()){
                 setTranscriptMessageTimestamp(
                     requestMessage,
@@ -3365,10 +3369,8 @@
                 }
                 setPendingStructuralToolCalls(terminalToolCalls);
             }else{
-                for(const requestMessage of requestMessages) requestMessage.remove();
-                message.remove();
+                removeTransientSessionCards();
                 setPendingStructuralToolCalls(previousPendingToolCalls);
-                scrollTranscriptToBottom();
             }
             setSessionStatus(
                 pendingStructuralToolMessage?'tool':'ready',
@@ -3417,16 +3419,13 @@
                     const response=[...chatOutput.children].find(
                         candidate=>candidate.id===`message-${messageId}`
                     )||null;
-                    const requestMessages=[...chatOutput.children].filter(
-                        candidate=>candidate.dataset.operationId===context.operationId
-                    );
                     response?.remove();
                     for(const requestMessage of requestMessages){
                         requestMessage.remove();
                     }
                     scrollTranscriptToBottom();
-                }else{
-                    restoreRejectedStructuralDraft(messageId,context.operationId,text);
+                }else if(retainTurn){
+                    restoreRejectedStructuralDraft(messageId,requestMessages,text);
                 }
                 arcaneLogging.error('Arcane structural tool protocol failure.',error);
                 setSessionStatus(
@@ -3436,14 +3435,11 @@
             }else if(!destroyed&&bindingGeneration===sessionBindingGeneration){
                 arcaneLogging.error('Arcane chat request failed.',error);
                 if(sessionRequestMessages.every(message=>message.role==='tool')){
-                    const requestMessages=[...chatOutput.children].filter(
-                        candidate=>candidate.dataset.operationId===context.operationId
-                    );
                     for(const requestMessage of requestMessages){
                         requestMessage.remove();
                     }
                 }
-                renderSessionMessageFailure(messageId,error);
+                if(retainTurn) renderSessionMessageFailure(messageId,error);
                 setSessionStatus(
                     'error',
                     visibleErrorMessage(error,'The AI request failed.')
@@ -3471,6 +3467,7 @@
             }
             throw error;
         }finally{
+            removeTransientSessionCards();
             if(activeSessionMessageToken===sessionMessageToken){
                 activeSessionMessageToken=null;
                 sessionMessagePending=false;
@@ -3625,6 +3622,7 @@
             return false;
         }
         try{
+            const requestMessages=[];
             for(const result of orderedResults){
                 const visibleMessage=appendTranscriptMessage(
                     'tool',
@@ -3633,12 +3631,14 @@
                 );
                 visibleMessage.dataset.operationId=operationId;
                 visibleMessage.dataset.toolCallId=result.tool_call_id;
+                requestMessages.push(visibleMessage);
             }
             const result=sendMessageThroughBoundSession(
                 orderedResults.map(message=>message.content).join('\n'),
                 eventContext,
                 orderedResults,
-                perTurnRequest
+                perTurnRequest,
+                requestMessages
             );
             return observeHostSubmission(result,eventContext,ownership);
         }catch(error){
@@ -3824,6 +3824,7 @@
             return false;
         }
         try{
+            const requestMessages=[];
             if(!eventContext.preserveDraft){
                 textArea.value='';
             }
@@ -3834,6 +3835,7 @@
                 hostSubmissionGeneration++;
                 const visibleMessage=appendTranscriptMessage('user',text,host.name);
                 visibleMessage.dataset.operationId=operationId;
+                requestMessages.push(visibleMessage);
             }
             scrollTranscriptToBottom(!eventContext.synthetic);
 
@@ -3854,7 +3856,9 @@
             let result;
             try{
                 result=boundChatSession
-                    ?sendMessageThroughBoundSession(text,eventContext)
+                    ?sendMessageThroughBoundSession(
+                        text,eventContext,undefined,null,requestMessages
+                    )
                     :host.sendMessage(text,eventContext);
             }catch(error){
                 result=Promise.reject(error);

package/runtime/arcane/modules/AI.js

@@ -4443,7 +4443,8 @@ class AI {
             function ignoreBuiltInLLMProviderRequest(){},
             signal,
             payload.reasoningEffort,
-            onRetry
+            onRetry,
+            payload.model
         );
     }
 
@@ -4474,7 +4475,8 @@ class AI {
             function ignoreBuiltInLLMStreamResult(){},
             payload.reasoningEffort,
             observeToolText,
-            onRetry
+            onRetry,
+            payload.model
         );
     }
 
@@ -4555,7 +4557,7 @@ class AI {
         });
     }
 
-    #openAICompatibleOllamaResponse(response={},id=Date.now()){
+    #openAICompatibleOllamaResponse(response={},id=Date.now(),model=this.model){
         const responseRecord=isPlainAIRecord(response)?response:{};
         const message=isPlainAIRecord(responseRecord.message)
             ?responseRecord.message
@@ -4597,7 +4599,7 @@ class AI {
                 :Math.floor(Date.now()/1000),
             model:Object.hasOwn(responseRecord,'model')
                 ?responseRecord.model
-                :this.model,
+                :model,
             choices:[
                 {
                     index:0,
@@ -4620,6 +4622,7 @@ class AI {
 
     async streamRequest({
         messages=[],
+        model,
         structuredOutput=false,
         localOnly=false,
         onChunk=function ignoreStreamChunk(){},
@@ -4676,6 +4679,7 @@ class AI {
         if(this.#shouldUseProviderRuntime('llm',this.llmService,localOnly)){
             const request={
                 messages,
+                ...(model!==undefined?{model}:{}),
                 structuredOutput,
                 tools,
                 toolChoice,
@@ -4801,7 +4805,8 @@ class AI {
                 onDataResult,
                 normalizedReasoningEffort,
                 observeToolText,
-                onRetry
+                onRetry,
+                model
             );
             const structuralToolCalls=normalizeAICompletionToolCalls(
                 completion
@@ -4910,7 +4915,8 @@ class AI {
         dataResultHandler=function ignoreBuiltInStreamDataResult(){},
         reasoningEffort,
         observeToolText=null,
-        onRetry = null
+        onRetry = null,
+        model=this.model
     ){
         let speechTurnCompleted=false;
 
@@ -4934,7 +4940,7 @@ class AI {
             reasoningEffort===undefined?this.reasoningEffort:reasoningEffort
         );
         const request={
-            model:this.model,
+            model,
             messages:messages, 
             stream:true
         }
@@ -4981,7 +4987,7 @@ class AI {
             const ollamaMessages=this.#ollamaMessages(messages,tool_choice);
             const adaptNativeToolCalls=value=>this.#openAICompatibleOllamaToolCalls(value,id);
             const ollamaRequest={
-                model:this.model,
+                model,
                 messages:ollamaMessages,
                 stream:true,
                 ...(normalizedReasoningEffort
@@ -5091,7 +5097,8 @@ class AI {
                             :nativeContent
                     }
                 },
-                id
+                id,
+                model
             );
             const structuralToolCalls=normalizeAICompletionToolCalls(
                 nativeCompletion
@@ -5583,7 +5590,7 @@ class AI {
             created:Object.hasOwn(streamMetadata,'created')
                 ?streamMetadata.created
                 :Math.floor(Date.now()/1000),
-            model:Object.hasOwn(streamMetadata,'model')?streamMetadata.model:this.model,
+            model:Object.hasOwn(streamMetadata,'model')?streamMetadata.model:model,
             choices:completionChoices
         };
         const terminalToolCalls=normalizeAICompletionToolCalls(completion);
@@ -5628,6 +5635,7 @@ class AI {
 
     async fetchRequest({
         messages=[],
+        model,
         structuredOutput=false,
         localOnly=false,
         tools=[],
@@ -5670,6 +5678,7 @@ class AI {
             }
             const request={
                 messages,
+                ...(model!==undefined?{model}:{}),
                 structuredOutput,
                 tools,
                 toolChoice,
@@ -5726,7 +5735,8 @@ class AI {
             onRequest,
             signal,
             normalizedReasoningEffort,
-            onRetry
+            onRetry,
+            model
         );
     }
 
@@ -5780,7 +5790,8 @@ class AI {
         requestHandler=function ignoreFetchRequest(){},
         signal=null,
         reasoningEffort,
-        onRetry = null
+        onRetry = null,
+        model=this.model
     ){
         validateAIStructuralRequest(messages,tools,parallel_tool_calls);
         this.#assertServiceConfigured(this.llmService);
@@ -5799,7 +5810,7 @@ class AI {
             reasoningEffort===undefined?this.reasoningEffort:reasoningEffort
         );
         const request={
-            model:this.model,
+            model,
             messages:messages, 
             stream:false
         }
@@ -5838,7 +5849,7 @@ class AI {
             const ollamaTools=this.#ollamaTools(tools,tool_choice);
             const ollamaMessages=this.#ollamaMessages(messages,tool_choice);
             const nativeRequest={
-                model:this.model,
+                model,
                 messages:ollamaMessages,
                 stream:false,
                 ...(normalizedReasoningEffort
@@ -5869,7 +5880,8 @@ class AI {
             }
             const responseJSON=this.#openAICompatibleOllamaResponse(
                 nativeResponse,
-                id
+                id,
+                model
             );
             this.#assertRequiredOllamaToolCall(
                 normalizeAICompletionToolCalls(responseJSON),