feat(learn): an assistant reads PDFs' pages, and sees images the person chooses or approves

ADR 0021, D12: text first. Lookups find a studyset's PDFs and read a page's
text as PDFium extracts it. An assistant that reads images is offered a
card's images as chips, and may ask to see a card's image or a PDF's page
drawn whole; the conversation asks the person each time (Show it, Don't
show) and sends nothing meanwhile. Images go only as png, jpeg, gif or webp
of at most 5 MB, are never kept in a turn, and a follow-up goes without them.

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/model/assistant_prompt_test.dart

@@ -1,3 +1,5 @@
+import 'dart:typed_data';
+
 import 'package:flutter_test/flutter_test.dart';
 import 'package:halkyon_learn_flutter/src/feature/assistant/model/assistant_conversation.dart';
 import 'package:halkyon_learn_flutter/src/feature/assistant/model/assistant_prompt.dart';
@@ -67,6 +69,59 @@ void main() {
     });
   });
 
+  group('an image chosen (D12)', () {
+    final bowl = LlmImage(
+      mediaType: 'image/png',
+      bytes: Uint8List.fromList([1, 2, 3]),
+    );
+    const image = AssistantPart(
+      kind: AssistantPartKind.image,
+      text: 'asset-bowl',
+      number: 2,
+    );
+
+    test('goes labelled, the image after its label', () {
+      final message = questionMessage(
+        'What is it?',
+        const [
+          image,
+        ],
+        images: {'asset-bowl': bowl},
+      );
+
+      expect(message.parts, hasLength(3));
+      expect((message.parts[1] as LlmText).text, "\n\nThe card's image 2:\n");
+      expect(message.parts[2], same(bowl));
+    });
+
+    test('that cannot be had is said to be missing', () {
+      final message = questionMessage('What is it?', const [image]);
+
+      expect(_texts(message).last, "\n\nThe card's image 2:\n$imageNotHad");
+    });
+
+    test('is left as a mark once answered, read back as the part it was', () {
+      final question = questionMessage(
+        'What is it?',
+        const [
+          image,
+        ],
+        images: {'asset-bowl': bowl},
+      );
+      final answer = LlmMessage.assistant(const [LlmText('A bowl.')]);
+
+      final kept = withoutImages([question, answer]);
+
+      expect(kept.last, same(answer));
+      expect(_texts(kept.first).last, imageNotKept);
+      final asked = askedIn(kept.first);
+      expect(asked.question, 'What is it?');
+      expect(asked.parts, [
+        const AssistantPart(kind: AssistantPartKind.image, text: '', number: 2),
+      ]);
+    });
+  });
+
   group('the system prompt', () {
     test('offers nothing to look up where only what is ticked goes', () {
       final prompt = systemPrompt(const AssistantScope());