Accept pasted images in inbox Other and Reply answers

internal/agents/inbox.go

@@ -326,7 +326,11 @@ func (m *Manager) Deliver(ctx context.Context, sessionID string) error {
 		return err
 	}
 
-	after, err := m.turn(ctx, sess, task, pending[0].Response.User, deliveryMessage(pending), nil)
+	images := []string{}
+	for _, p := range pending {
+		images = append(images, p.Response.Images...)
+	}
+	after, err := m.turn(ctx, sess, task, pending[0].Response.User, deliveryMessage(pending), images)
 	if errors.Is(err, ErrConflict) {
 		slog.Info("deliver held: session changed state", "session", sessionID, "err", err, "responses", ids)
 		return nil
@@ -433,6 +437,12 @@ func (m *Manager) responsesMidTurn(ctx context.Context, l *live) (string, []stri
 		return "", nil
 	}
 	ids := responseIDs(pending)
+	for _, p := range pending {
+		if len(p.Response.Images) > 0 {
+			slog.Info("deliver held mid-turn: response has images", "session", l.id, "responses", ids)
+			return "", nil
+		}
+	}
 
 	waiting, err := m.store.AwaitingHuman(ctx, l.id)
 	if err != nil {
@@ -498,14 +508,22 @@ func deliveryMessage(pending []store.Pending) string {
 				fmt.Fprintf(b, "%s chose: %s\n", r.User, (*r.Options)[0])
 			}
 		case api.Other:
+			text := ""
+			if r.Text != nil {
+				text = *r.Text
+			}
 			if p.Item.Kind == api.Question {
-				fmt.Fprintf(b, "%s answered (none of your options): %s\n", r.User, *r.Text)
+				fmt.Fprintf(b, "%s answered (none of your options): %s\n", r.User, text)
 				fmt.Fprintf(b, "If this asks you something back, answer by calling ask_human with thread_id %q.\n", p.Item.ThreadId)
 			} else {
-				fmt.Fprintf(b, "%s replied: %s\n", r.User, *r.Text)
+				fmt.Fprintf(b, "%s replied: %s\n", r.User, text)
 			}
 		}
 
+		if len(r.Images) > 0 {
+			fmt.Fprintf(b, "%s attached %d image(s) to this response; they come with this message.\n", r.User, len(r.Images))
+		}
+
 		for _, a := range r.Attachments {
 			j, err := json.Marshal(a)
 			if err != nil {

internal/agents/respond.go

@@ -42,6 +42,10 @@ func (m *Manager) Respond(ctx context.Context, user string, creates []api.InboxR
 		if create.Attachments != nil {
 			attachments = *create.Attachments
 		}
+		images := []string{}
+		if create.Images != nil {
+			images = *create.Images
+		}
 
 		resolutions = append(resolutions, store.Resolution{
 			State: state,
@@ -52,6 +56,7 @@ func (m *Manager) Respond(ctx context.Context, user string, creates []api.InboxR
 				Type:        create.Type,
 				Options:     create.Options,
 				Text:        create.Text,
+				Images:      images,
 				Attachments: attachments,
 				CreatedAt:   time.Now().UTC(),
 			},
@@ -102,6 +107,7 @@ func CheckResponse(item api.InboxItem, c api.InboxResponseCreate) (api.InboxStat
 	if c.Text != nil {
 		text = strings.TrimSpace(*c.Text)
 	}
+	images := c.Images != nil && len(*c.Images) > 0
 	question := item.Kind == api.Question
 	bad := func(format string, args ...any) (api.InboxState, error) {
 		return "", fmt.Errorf("%w: item %s: "+format, append([]any{ErrInvalid, item.Id}, args...)...)
@@ -128,14 +134,14 @@ func CheckResponse(item api.InboxItem, c api.InboxResponseCreate) (api.InboxStat
 			}
 			seen[o] = true
 		}
-		if text != "" && item.Payload["multiple"] != true {
-			return bad("a pick-one option takes no text; use other")
+		if (text != "" || images) && item.Payload["multiple"] != true {
+			return bad("a pick-one option takes no text or images; use other")
 		}
 		return api.Answered, nil
 
 	case api.Other:
-		if text == "" {
-			return bad("other needs text")
+		if text == "" && !images {
+			return bad("other needs text or images")
 		}
 		return api.Answered, nil
 
@@ -143,6 +149,9 @@ func CheckResponse(item api.InboxItem, c api.InboxResponseCreate) (api.InboxStat
 		if question {
 			return bad("questions need an option or other, not a dismissal")
 		}
+		if images {
+			return bad("a dismissal takes no images")
+		}
 		return api.Dismissed, nil
 	}