fix(chat): empty assistant bubble for plain LLM responses (#290)

@cacheplane/partial-markdown@0.3 does not flush trailing text on
finish() unless the buffer ends with a newline. Plain LLM responses
typically omit the trailing newline, so the parser produced a document
with zero children — the assistant bubble shell rendered but the
message text was missing entirely.

Reproduce on main: send 'Say hello in one sentence.' to gpt-5-mini.
Backend logs show success (1.8s); agent.messages() carries 'Hello — nice
to meet you!'; the DOM shows the bubble structure (checkpoint marker,
action buttons) but no text.

Fix: push a sentinel newline before calling finish() in both branches
of the root() computed (content-changed-then-flush and streaming-flipped-
without-new-content). The newline only ever exists inside the parser
buffer; the original content() signal is unchanged.

Two new regression tests cover plain text without trailing newline in
both branches.

Found while walking the smoke checklist post-#288.

Author: blove

libs/chat/src/lib/streaming/streaming-markdown.component.spec.ts

@@ -73,4 +73,32 @@ describe('ChatStreamingMdComponent', () => {
     expect(fixture.nativeElement.querySelector('p')).toBeNull();
     expect(fixture.nativeElement.querySelector('h1')).toBeNull();
   });
+
+  it('renders a paragraph for plain text WITHOUT a trailing newline (LLM-response shape)', () => {
+    // Regression: @cacheplane/partial-markdown@0.3 does not flush trailing
+    // text on finish() unless the buffer ends with '\n'. LLM responses
+    // typically omit the trailing newline. The component must push a
+    // sentinel newline before finish() so the message renders.
+    const fixture = TestBed.createComponent(HostComponent);
+    fixture.componentInstance.content.set('Hello — nice to meet you!');
+    fixture.componentInstance.streaming.set(false);
+    fixture.detectChanges();
+    const p = fixture.nativeElement.querySelector('p');
+    expect(p).toBeTruthy();
+    expect(p.textContent?.trim()).toBe('Hello — nice to meet you!');
+  });
+
+  it('renders plain text when streaming flips to false (mirrored else-branch)', () => {
+    // The else-if branch (no content change, streaming flipped to false)
+    // must also push a sentinel newline before finish().
+    const fixture = TestBed.createComponent(HostComponent);
+    fixture.componentInstance.content.set('Plain answer.');
+    fixture.componentInstance.streaming.set(true);
+    fixture.detectChanges();
+    fixture.componentInstance.streaming.set(false);
+    fixture.detectChanges();
+    const p = fixture.nativeElement.querySelector('p');
+    expect(p).toBeTruthy();
+    expect(p.textContent?.trim()).toBe('Plain answer.');
+  });
 });

libs/chat/src/lib/streaming/streaming-markdown.component.ts

@@ -99,12 +99,20 @@ export class ChatStreamingMdComponent {
         if (c.length > 0) this.parser.push(c);
       }
       if (!isStreaming && !this.finished) {
+        // @cacheplane/partial-markdown@0.3 does not flush trailing text on
+        // finish() unless the buffer ends with a newline. Plain LLM
+        // responses often omit the trailing newline, which causes the
+        // parser to emit a document with zero children — i.e. the message
+        // renders empty. Push a sentinel newline first to force the open
+        // paragraph closed before we finalize.
+        if (!c.endsWith('\n')) this.parser.push('\n');
         this.parser.finish();
         this.finished = true;
       }
       this.prior = c;
     } else if (!isStreaming && !this.finished) {
       // Streaming flipped to false without new content; ensure parser is finalized.
+      if (!this.prior.endsWith('\n')) this.parser.push('\n');
       this.parser.finish();
       this.finished = true;
     }