feat(learn): choosing how much an assistant thinks, and the assistants by provider in the field

Halkyon Learn's ADR 0021, D60 and D63:

- A chip after the scope's offers the efforts the operator allows for the
  assistant asked, its own marked; the one picked is kept per assistant on
  the device (DeviceSettings.askEfforts) and asked at, also by a
  conversation continued from the history. One the operator no longer
  allows counts as none.
- Every question asks for its effort's ceiling on output, as the thinking
  counts towards it (8,192 to 65,536 tokens; 16,384 at the model's own),
  where it asked for 4,096; the flashcards proposed, Explain and Why? ask
  at the assistant's own effort, a person's own key at none and 4,096.
- The field's picker lists the assistants under their providers' names,
  each in its colour.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Tizian 「ティツィアーン」 <raisondetredev@pm.me>

apps/learn/halkyon_learn_flutter/test/feature/assistant/bloc/card_proposals_cubit_test.dart

@@ -80,6 +80,12 @@ void main() {
     expect(assistant, 'claude');
     expect(request.tools, [proposeCardsTool]);
     expect(request.system, proposalPrompt());
+    expect(
+      request.effort,
+      isNull,
+      reason: "the assistant's own effort: nobody chose one here (D63)",
+    );
+    expect(request.maxOutputTokens, LlmEffort.defaultOutputCeiling);
     expect(
       (request.messages.single.parts.single as LlmText).text,
       'Verbs: 食べる is to eat.',

apps/learn/halkyon_learn_flutter/test/feature/assistant/bloc/conversation_cubit_test.dart

@@ -131,6 +131,7 @@ void main() {
     AssistantGateway? asking,
     AssistantTask task = AssistantTask.ask,
     String? title,
+    LlmEffort? effort,
   }) => ConversationCubit(
     assistants: AssistantRepository(asking ?? gateway),
     lookups: lookups,
@@ -141,6 +142,7 @@ void main() {
     conversationId: conversationId,
     task: task,
     title: title,
+    effort: effort,
     answerIn: 'en',
   );
 
@@ -182,7 +184,12 @@ void main() {
 
     final (assistant, request) = gateway.asked.single;
     expect(assistant, 'claude');
-    expect(request.maxOutputTokens, ConversationCubit.maxOutputTokens);
+    expect(
+      request.maxOutputTokens,
+      LlmEffort.defaultOutputCeiling,
+      reason: "the model's own effort thinks within the default ceiling",
+    );
+    expect(request.effort, isNull);
     expect(request.tools, isEmpty, reason: 'only what is ticked goes (D50)');
     expect(request.system, contains('nothing to look up'));
 
@@ -195,6 +202,60 @@ void main() {
     await conversation.close();
   });
 
+  test('asks every question at the effort the person chose, with its '
+      'ceiling (D63)', () async {
+    const thinking = OfferedAssistant(
+      id: 'claude',
+      name: 'Claude',
+      protocol: 'anthropic',
+      host: 'api.anthropic.com',
+      model: 'claude-opus',
+      effort: LlmEffort.low,
+      efforts: [LlmEffort.low, LlmEffort.medium, LlmEffort.high],
+    );
+    final conversation = cubit(
+      choice: const GrantedAssistant(thinking),
+      effort: LlmEffort.high,
+    );
+    await conversation.open();
+    gateway.answers
+      ..add(_says('Because it joins.'))
+      ..add(_says('Yes.'));
+
+    await conversation.ask('Why て-form?', const []);
+    await conversation.ask('Always?', const []);
+
+    for (final (_, request) in gateway.asked) {
+      expect(request.effort, LlmEffort.high);
+      expect(request.maxOutputTokens, LlmEffort.high.outputCeiling);
+    }
+    expect(gateway.asked, hasLength(2));
+    await conversation.close();
+  });
+
+  test("without an effort chosen, it asks at the assistant's, within its "
+      'ceiling', () async {
+    const thinking = OfferedAssistant(
+      id: 'claude',
+      name: 'Claude',
+      protocol: 'anthropic',
+      host: 'api.anthropic.com',
+      model: 'claude-opus',
+      effort: LlmEffort.low,
+      efforts: [LlmEffort.low, LlmEffort.medium],
+    );
+    final conversation = cubit(choice: const GrantedAssistant(thinking));
+    await conversation.open();
+    gateway.answers.add(_says('Because it joins.'));
+
+    await conversation.ask('Why て-form?', const []);
+
+    final (_, request) = gateway.asked.single;
+    expect(request.effort, isNull, reason: 'the instance applies its own');
+    expect(request.maxOutputTokens, LlmEffort.low.outputCeiling);
+    await conversation.close();
+  });
+
   test('sends the conversation so far with a follow-up, and keeps it in the '
       'same conversation', () async {
     final conversation = cubit();