feat: AI categorisation fallback, unsubscribe confidence scoring, one-line summaries

Heuristic classifier falls back to AI only when it can't determine a category;
unsubscribe confidence blends method reliability with an AI safety opinion and
surfaces it in the Unsubscribe Manager; emails get an on-demand AI one-line
summary in the detail pane. All AI calls degrade gracefully when AI is disabled
or the provider errors.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/InboxIntel.Api/Controllers/EmailController.cs

@@ -18,12 +18,14 @@ public class EmailController : ApiControllerBase
 {
     private readonly ICleanupService _cleanup;
     private readonly IUnsubscribeService _unsub;
+    private readonly IAiService _ai;
     private readonly AppDbContext _db;
 
-    public EmailController(ICleanupService cleanup, IUnsubscribeService unsub, AppDbContext db)
+    public EmailController(ICleanupService cleanup, IUnsubscribeService unsub, IAiService ai, AppDbContext db)
     {
         _cleanup = cleanup;
         _unsub = unsub;
+        _ai = ai;
         _db = db;
     }
 
@@ -42,6 +44,16 @@ public class EmailController : ApiControllerBase
         return e is null ? NotFound() : Ok(e);
     }
 
+    /// <summary>One-line AI summary, fetched on demand (returns the snippet verbatim if AI is disabled).</summary>
+    [HttpGet("{id:guid}/summary")]
+    public async Task<IActionResult> Summary(Guid id, CancellationToken ct)
+    {
+        var e = await _db.Emails.FirstOrDefaultAsync(e => e.Id == id && e.UserId == UserId, ct);
+        if (e is null) return NotFound();
+        var summary = await _ai.SummarizeEmailAsync(e.Subject, e.Snippet, e.BodyText, ct);
+        return Ok(new { summary });
+    }
+
     [HttpPost("{id:guid}/read")]
     public Task<IActionResult> MarkRead(Guid id, CancellationToken ct) => Act(id, CleanupActionType.MarkRead, ct);