diff --git a/apps/activities/routes.py b/apps/activities/routes.py
index 6a015b394..c3dcb1701 100644
--- a/apps/activities/routes.py
+++ b/apps/activities/routes.py
@@ -266,7 +266,7 @@ def _dev_activities_screens_detail(
         return "", 404
 
     # Check if the screen.jsonl file exists in segment
-    segment_dir = str(get_segment_path(day, timestamp, stream))
+    segment_dir = str(get_segment_path(day, timestamp, stream, create=False))
     jsonl_path = os.path.join(segment_dir, filename)
     if not os.path.isfile(jsonl_path):
         return "", 404
@@ -354,7 +354,7 @@ def _dev_screen_frames(
     if not filename.endswith("screen.jsonl"):
         return "", 404
 
-    segment_dir = str(get_segment_path(day, timestamp, stream))
+    segment_dir = str(get_segment_path(day, timestamp, stream, create=False))
     jsonl_path = os.path.join(segment_dir, filename)
 
     if not os.path.isfile(jsonl_path):
diff --git a/apps/import/ingest.py b/apps/import/ingest.py
index 7cafe14e0..9887911a2 100644
--- a/apps/import/ingest.py
+++ b/apps/import/ingest.py
@@ -31,7 +31,7 @@ from think.entities.journal import (
     save_journal_entity,
 )
 from think.entities.matching import find_matching_entity
-from think.utils import DEFAULT_STREAM, day_path
+from think.utils import DEFAULT_STREAM, STREAM_RE, day_path
 
 from .journal_sources import (
     get_state_directory,
@@ -43,7 +43,6 @@ logger = logging.getLogger(__name__)
 
 _DAY_RE = re.compile(r"^\d{8}$")
 _SEGMENT_RE = re.compile(r"^\d{6}_\d+$")
-_STREAM_RE = re.compile(r"^[a-z0-9][a-z0-9._-]*$")
 _FACET_NAME_RE = re.compile(r"^[a-z0-9][a-z0-9_-]*$")
 _IMPORT_ID_RE = re.compile(r"^\d{8}_\d{6}$")
 
@@ -150,7 +149,7 @@ def register_ingest_routes(bp) -> None:
 
                 if not _DAY_RE.match(day):
                     raise ValueError("Invalid day format")
-                if stream != DEFAULT_STREAM and not _STREAM_RE.match(stream):
+                if stream != DEFAULT_STREAM and not STREAM_RE.fullmatch(stream):
                     raise ValueError("Invalid stream format")
                 if not _SEGMENT_RE.match(segment_key):
                     raise ValueError("Invalid segment_key format")
diff --git a/apps/speakers/attribution.py b/apps/speakers/attribution.py
index 280ca05bd..b44973a18 100644
--- a/apps/speakers/attribution.py
+++ b/apps/speakers/attribution.py
@@ -564,7 +564,7 @@ def accumulate_voiceprints(
         return {}
     owner_centroid, owner_threshold = centroid_data
 
-    seg_dir = segment_path(day, segment_key, stream)
+    seg_dir = segment_path(day, segment_key, stream, create=False)
     emb_data = load_embeddings_file(seg_dir / f"{source}.npz")
     if emb_data is None:
         return {}
diff --git a/apps/speakers/discovery.py b/apps/speakers/discovery.py
index 818c7a9ee..8d7ce531b 100644
--- a/apps/speakers/discovery.py
+++ b/apps/speakers/discovery.py
@@ -113,7 +113,7 @@ def discover_unknown_speakers() -> dict[str, Any]:
         for segment in scan_segment_embeddings(day):
             stream = segment["stream"]
             seg_key = segment["key"]
-            seg_dir = segment_path(day, seg_key, stream)
+            seg_dir = segment_path(day, seg_key, stream, create=False)
 
             labels_data = load_speaker_labels(seg_dir)
             attributed_sids: set[int] = set()
@@ -221,7 +221,7 @@ def discover_unknown_speakers() -> dict[str, Any]:
                 continue
             seen_segments.add(seg_triplet)
             seg_dir = segment_path(
-                record["day"], record["segment_key"], record["stream"]
+                record["day"], record["segment_key"], record["stream"], create=False
             )
             samples.append(
                 {
@@ -247,7 +247,7 @@ def discover_unknown_speakers() -> dict[str, Any]:
             for pos in sorted_positions:
                 record = provenance[int(cluster_indices[int(pos)])]
                 seg_dir = segment_path(
-                    record["day"], record["segment_key"], record["stream"]
+                    record["day"], record["segment_key"], record["stream"], create=False
                 )
                 sample = {
                     **record,
diff --git a/apps/speakers/owner.py b/apps/speakers/owner.py
index 41aa8f750..3f303fa5b 100644
--- a/apps/speakers/owner.py
+++ b/apps/speakers/owner.py
@@ -342,7 +342,7 @@ def classify_sentences(
 
     centroid, threshold = centroid_data
     emb_data = load_embeddings_file(
-        segment_path(day, segment_key, stream) / f"{source}.npz"
+        segment_path(day, segment_key, stream, create=False) / f"{source}.npz"
     )
     if emb_data is None:
         return []
diff --git a/apps/speakers/routes.py b/apps/speakers/routes.py
index 59952426a..8538977cb 100644
--- a/apps/speakers/routes.py
+++ b/apps/speakers/routes.py
@@ -433,7 +433,7 @@ def _load_sentences(
         - emb_data: Tuple of (embeddings, statement_ids) or None if no embeddings
     """
     if stream:
-        segment_dir = get_segment_path(day, segment_key, stream)
+        segment_dir = get_segment_path(day, segment_key, stream, create=False)
     else:
         segment_dir = day_path(day) / segment_key
 
@@ -493,7 +493,7 @@ def _get_sentence_embedding(
 ) -> np.ndarray | None:
     """Get a specific sentence's embedding, normalized."""
     if stream:
-        segment_dir = get_segment_path(day, segment_key, stream)
+        segment_dir = get_segment_path(day, segment_key, stream, create=False)
     else:
         segment_dir = day_path(day) / segment_key
     npz_path = segment_dir / f"{source}.npz"
@@ -571,7 +571,7 @@ def api_segments(day: str) -> Any:
     principal = get_journal_principal()
     principal_id = principal["id"] if principal else None
     for seg in segments:
-        seg_dir = get_segment_path(day, seg["key"], seg["stream"])
+        seg_dir = get_segment_path(day, seg["key"], seg["stream"], create=False)
         labels_data = _load_speaker_labels(seg_dir)
         if labels_data:
             labels = labels_data.get("labels", [])
@@ -612,7 +612,7 @@ def api_segment_speakers(day: str, stream: str, segment_key: str) -> Any:
         return error_response("Invalid segment key", 400)
 
     # Load speakers from speakers.json
-    segment_dir = get_segment_path(day, segment_key, stream)
+    segment_dir = get_segment_path(day, segment_key, stream, create=False)
     speakers = _load_segment_speakers(segment_dir)
     if not speakers:
         return jsonify({"matched": [], "unmatched": []})
@@ -658,7 +658,7 @@ def api_review(day: str, stream: str, segment_key: str, source: str) -> Any:
     if not sentences:
         return error_response("No transcript found", 404)
 
-    segment_dir = get_segment_path(day, segment_key, stream)
+    segment_dir = get_segment_path(day, segment_key, stream, create=False)
     labels_data = _load_speaker_labels(segment_dir)
     label_map: dict[int, dict] = {}
     if labels_data:
diff --git a/apps/transcripts/routes.py b/apps/transcripts/routes.py
index c1f406459..f13389656 100644
--- a/apps/transcripts/routes.py
+++ b/apps/transcripts/routes.py
@@ -34,7 +34,8 @@ from observe.utils import AUDIO_EXTENSIONS, VIDEO_EXTENSIONS
 from think.cluster import cluster_scan, cluster_segments, scan_day
 from think.entities.journal import get_journal_principal, load_journal_entity
 from think.models import get_usage_cost
-from think.utils import day_dirs, day_path, segment_path
+from think.supervisor import is_supervisor_up
+from think.utils import STREAM_RE, day_dirs, day_path, segment_path
 from think.utils import segment_key as validate_segment_key
 
 logger = logging.getLogger(__name__)
@@ -248,10 +249,13 @@ def segment_content(day: str, stream: str, segment_key: str) -> Any:
     if not DATE_RE.fullmatch(day):
         return error_response("Invalid day format", 404)
 
+    if not STREAM_RE.fullmatch(stream):
+        return error_response("Invalid stream format", 404)
+
     if not validate_segment_key(segment_key):
         return error_response("Invalid segment key format", 404)
 
-    segment_dir = str(segment_path(day, segment_key, stream))
+    segment_dir = str(segment_path(day, segment_key, stream, create=False))
     if not os.path.isdir(segment_dir):
         return error_response("Segment directory not found", 404)
 
@@ -505,8 +509,11 @@ def delete_segment(day: str, stream: str, segment_key: str) -> Any:
     if not validate_segment_key(segment_key):
         return error_response("Invalid segment key format", 400)
 
-    day_dir = str(day_path(day))
-    segment_dir = str(segment_path(day, segment_key, stream))
+    if not STREAM_RE.fullmatch(stream):
+        return error_response("Invalid stream format", 400)
+
+    day_dir = str(day_path(day, create=False))
+    segment_dir = str(segment_path(day, segment_key, stream, create=False))
 
     # Verify segment exists
     if not os.path.isdir(segment_dir):
@@ -529,6 +536,10 @@ def delete_segment(day: str, stream: str, segment_key: str) -> Any:
             day=day,
         )
 
+        payload = {"deleted": segment_key}
+        if not is_supervisor_up():
+            payload["search_index_warning"] = True
+
         # Trigger indexer rescan to remove deleted segment from search index
         # Supervisor queues by command name, serializing concurrent indexer requests
         emit(
@@ -537,7 +548,7 @@ def delete_segment(day: str, stream: str, segment_key: str) -> Any:
             cmd=["sol", "indexer", "--rescan-full"],
         )
 
-        return success_response({"deleted": segment_key})
+        return success_response(payload)
 
     except OSError as e:
         return error_response(f"Failed to delete segment: {e}", 500)
diff --git a/apps/transcripts/tests/conftest.py b/apps/transcripts/tests/conftest.py
index 5d6c79520..5b79a1f02 100644
--- a/apps/transcripts/tests/conftest.py
+++ b/apps/transcripts/tests/conftest.py
@@ -4,14 +4,45 @@
 """Fixtures for transcripts app tests."""
 
 import os
+import sys
+from pathlib import Path
 
 import pytest
 
+from convey import create_app
+
+ROOT = Path(__file__).resolve().parents[3]
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+
+from tests._baseline_harness import copytree_tracked
+
 
 @pytest.fixture(autouse=True)
-def _journal_env(monkeypatch):
-    """Point _SOLSTONE_JOURNAL_OVERRIDE at the test fixtures."""
+def _journal_env(request, monkeypatch):
+    """Point tests at a copied journal when needed, otherwise the tracked fixture."""
+    if "journal_copy" in request.fixturenames:
+        journal_copy = request.getfixturevalue("journal_copy")
+        monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(journal_copy))
+        return
+
     monkeypatch.setenv(
         "_SOLSTONE_JOURNAL_OVERRIDE",
         os.path.join(os.getcwd(), "tests", "fixtures", "journal"),
     )
+
+
+@pytest.fixture
+def client(journal_copy, monkeypatch):
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(journal_copy))
+    app = create_app(str(journal_copy))
+    return app.test_client()
+
+
+@pytest.fixture
+def journal_copy(tmp_path, monkeypatch):
+    src = Path(__file__).resolve().parents[3] / "tests" / "fixtures" / "journal"
+    dst = tmp_path / "journal"
+    copytree_tracked(src, dst)
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(dst.resolve()))
+    return dst
diff --git a/apps/transcripts/tests/test_segment_routes.py b/apps/transcripts/tests/test_segment_routes.py
new file mode 100644
index 000000000..b2477ba19
--- /dev/null
+++ b/apps/transcripts/tests/test_segment_routes.py
@@ -0,0 +1,107 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright (c) 2026 sol pbc
+
+import pytest
+
+FIXTURE_DAY = "20260304"
+FIXTURE_STREAM = "default"
+FIXTURE_SEGMENT = "090000_300"
+
+
+@pytest.mark.parametrize("stream", ["-bad", "Upper", "..bad"])
+def test_segment_content_rejects_invalid_stream(client, stream):
+    response = client.get(
+        f"/app/transcripts/api/segment/{FIXTURE_DAY}/{stream}/{FIXTURE_SEGMENT}"
+    )
+
+    assert response.status_code == 404
+    assert response.get_json() == {"error": "Invalid stream format"}
+
+
+@pytest.mark.parametrize("stream", ["-bad", "Upper", "..bad"])
+def test_delete_segment_rejects_invalid_stream(client, stream):
+    response = client.delete(
+        f"/app/transcripts/api/segment/{FIXTURE_DAY}/{stream}/{FIXTURE_SEGMENT}"
+    )
+
+    assert response.status_code == 400
+    assert response.get_json() == {"error": "Invalid stream format"}
+
+
+def test_segment_content_missing_segment_does_not_create_phantom_directory(
+    client, journal_copy
+):
+    response = client.get("/app/transcripts/api/segment/29990101/default/090000_300")
+
+    assert response.status_code == 404
+    assert response.get_json() == {"error": "Segment directory not found"}
+    assert not (journal_copy / "chronicle" / "29990101").exists()
+    assert not (
+        journal_copy / "chronicle" / "29990101" / "default" / "090000_300"
+    ).exists()
+
+
+def test_delete_missing_segment_does_not_create_phantom_directory(client, journal_copy):
+    response = client.delete("/app/transcripts/api/segment/29990101/default/090000_300")
+
+    assert response.status_code == 404
+    assert response.get_json() == {"error": "Segment not found"}
+    assert not (journal_copy / "chronicle" / "29990101").exists()
+    assert not (
+        journal_copy / "chronicle" / "29990101" / "default" / "090000_300"
+    ).exists()
+
+
+def test_segment_content_happy_path_returns_segment_payload(client):
+    response = client.get(
+        f"/app/transcripts/api/segment/{FIXTURE_DAY}/{FIXTURE_STREAM}/{FIXTURE_SEGMENT}"
+    )
+
+    assert response.status_code == 200
+    data = response.get_json()
+    assert data["segment_key"] == FIXTURE_SEGMENT
+    assert data["chunks"]
+    assert "media_sizes" in data
+
+
+def test_delete_segment_happy_path_removes_segment_directory(
+    client, journal_copy, monkeypatch
+):
+    monkeypatch.setattr("apps.transcripts.routes.is_supervisor_up", lambda: True)
+    segment_dir = (
+        journal_copy / "chronicle" / FIXTURE_DAY / FIXTURE_STREAM / FIXTURE_SEGMENT
+    )
+
+    response = client.delete(
+        f"/app/transcripts/api/segment/{FIXTURE_DAY}/{FIXTURE_STREAM}/{FIXTURE_SEGMENT}"
+    )
+
+    assert response.status_code == 200
+    assert response.get_json() == {"success": True, "deleted": FIXTURE_SEGMENT}
+    assert not segment_dir.exists()
+
+
+def test_delete_segment_includes_search_index_warning_when_supervisor_is_down(client):
+    response = client.delete(
+        f"/app/transcripts/api/segment/{FIXTURE_DAY}/{FIXTURE_STREAM}/{FIXTURE_SEGMENT}"
+    )
+
+    assert response.status_code == 200
+    assert response.get_json() == {
+        "success": True,
+        "deleted": FIXTURE_SEGMENT,
+        "search_index_warning": True,
+    }
+
+
+def test_delete_segment_omits_search_index_warning_when_supervisor_is_up(
+    client, monkeypatch
+):
+    monkeypatch.setattr("apps.transcripts.routes.is_supervisor_up", lambda: True)
+
+    response = client.delete(
+        f"/app/transcripts/api/segment/{FIXTURE_DAY}/{FIXTURE_STREAM}/{FIXTURE_SEGMENT}"
+    )
+
+    assert response.status_code == 200
+    assert response.get_json() == {"success": True, "deleted": FIXTURE_SEGMENT}
diff --git a/apps/transcripts/workspace.html b/apps/transcripts/workspace.html
index 193c8cea5..61bfa3500 100644
--- a/apps/transcripts/workspace.html
+++ b/apps/transcripts/workspace.html
@@ -787,6 +787,143 @@ body.has-date-nav .workspace:has(.tr-wrap) {
   flex-shrink: 0;
 }
 
+#trDeleteSegmentModal {
+  position: fixed;
+  inset: 0;
+  display: none;
+  align-items: center;
+  justify-content: center;
+  padding: 24px;
+  background: rgba(15, 23, 42, 0.55);
+  z-index: 1100;
+}
+
+#trDeleteSegmentModal .modal-content {
+  width: min(520px, 100%);
+  background: #ffffff;
+  border-radius: 16px;
+  box-shadow: 0 24px 48px rgba(15, 23, 42, 0.22);
+  overflow: hidden;
+}
+
+#trDeleteSegmentModal .modal-header {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  padding: 20px 24px 16px;
+  border-bottom: 1px solid #e5e7eb;
+}
+
+#trDeleteSegmentModal .modal-header h3 {
+  margin: 0;
+  font-size: 20px;
+  font-weight: 700;
+  color: #111827;
+}
+
+#trDeleteSegmentModal .modal-header.danger h3 {
+  color: #991b1b;
+}
+
+#trDeleteSegmentModal .modal-body {
+  padding: 20px 24px;
+}
+
+#trDeleteSegmentModal .modal-footer {
+  display: flex;
+  justify-content: flex-end;
+  gap: 12px;
+  padding: 0 24px 24px;
+}
+
+#trDeleteSegmentModal .close {
+  color: #9ca3af;
+  font-size: 28px;
+  line-height: 1;
+  cursor: pointer;
+  transition: color 0.15s ease;
+}
+
+#trDeleteSegmentModal .close:hover {
+  color: #4b5563;
+}
+
+#trDeleteSegmentModal .close:active {
+  color: #111827;
+}
+
+#trDeleteSegmentModal .btn-secondary,
+#trDeleteSegmentModal .btn-danger {
+  border: none;
+  border-radius: 10px;
+  padding: 10px 18px;
+  font-size: 14px;
+  font-weight: 600;
+  cursor: pointer;
+  transition: background 0.15s ease, color 0.15s ease, transform 0.15s ease;
+}
+
+#trDeleteSegmentModal .btn-secondary {
+  background: #f3f4f6;
+  color: #374151;
+}
+
+#trDeleteSegmentModal .btn-secondary:hover {
+  background: #e5e7eb;
+}
+
+#trDeleteSegmentModal .btn-secondary:active {
+  transform: translateY(1px);
+}
+
+#trDeleteSegmentModal .btn-danger {
+  background: #dc2626;
+  color: #ffffff;
+}
+
+#trDeleteSegmentModal .btn-danger:hover {
+  background: #b91c1c;
+}
+
+#trDeleteSegmentModal .btn-danger:active {
+  transform: translateY(1px);
+}
+
+#trDeleteSegmentModal .delete-warning {
+  padding: 16px 18px;
+  border-radius: 12px;
+  background: #fef2f2;
+  border-left: 4px solid #dc2626;
+  color: #7f1d1d;
+}
+
+#trDeleteSegmentModal .delete-warning p {
+  margin: 0 0 12px;
+}
+
+#trDeleteSegmentModal .delete-warning p:last-child {
+  margin-bottom: 0;
+}
+
+#trDeleteSegmentModal .tr-delete-segment-meta {
+  font-size: 14px;
+  color: #991b1b;
+}
+
+#trDeleteSegmentModal .tr-delete-segment-list {
+  margin: 0 0 12px;
+  padding-left: 20px;
+}
+
+#trDeleteSegmentModal .tr-delete-segment-list li + li {
+  margin-top: 6px;
+}
+
+#trDeleteSegmentModal .tr-delete-segment-size {
+  font-size: 13px;
+  color: #7f1d1d;
+}
+
 /* Unified timeline view */
 .tr-unified {
   display: flex;
@@ -1290,6 +1427,22 @@ body.has-date-nav .workspace:has(.tr-wrap) {
   </div>
 </div>
 
+<div id="trDeleteSegmentModal" class="modal" style="display: none;" role="dialog" aria-modal="true" aria-labelledby="trDeleteSegmentModalTitle">
+  <div class="modal-content">
+    <div class="modal-header danger">
+      <h3 id="trDeleteSegmentModalTitle">Delete segment?</h3>
+      <span class="close" id="trDeleteSegmentModalClose" onclick="closeDeleteSegmentModal()">&times;</span>
+    </div>
+    <div class="modal-body">
+      <div id="trDeleteSegmentModalBody" class="delete-warning"></div>
+    </div>
+    <div class="modal-footer">
+      <button type="button" class="btn-secondary" onclick="closeDeleteSegmentModal()">Cancel</button>
+      <button type="button" id="trDeleteSegmentModalConfirm" class="btn-danger" onclick="confirmDeleteSegment()">Delete</button>
+    </div>
+  </div>
+</div>
+
 <script src="{{ vendor_lib('marked') }}"></script>
 <script>
 (() => {
@@ -1340,6 +1493,9 @@ body.has-date-nav .workspace:has(.tr-wrap) {
   });
   const panel = document.getElementById('trPanel');
   const deleteBtn = document.getElementById('trDeleteBtn');
+  const deleteSegmentModal = document.getElementById('trDeleteSegmentModal');
+  const deleteSegmentModalBody = document.getElementById('trDeleteSegmentModalBody');
+  const deleteSegmentModalConfirm = document.getElementById('trDeleteSegmentModalConfirm');
   const emptyIcons = {
     day: '<svg viewBox="0 0 24 24"><circle cx="12" cy="12" r="10"></circle><polyline points="12 6 12 12 16 14"></polyline></svg>',
     nothing: '<svg viewBox="0 0 24 24"><rect x="3" y="4" width="18" height="18" rx="2" ry="2"></rect><line x1="16" y1="2" x2="16" y2="6"></line><line x1="8" y1="2" x2="8" y2="6"></line><line x1="3" y1="10" x2="21" y2="10"></line></svg>',
@@ -1366,6 +1522,8 @@ body.has-date-nav .workspace:has(.tr-wrap) {
   let allSegments = [];
   let selectedSegment = null;
   let updateNowPosition = null;
+  let deleteModalReturnFocus = null;
+  let deleteModalPrevOverflow = '';
 
   // Zoom state
   let zoomHeight = zoom.clientHeight;
@@ -3066,13 +3224,25 @@ body.has-date-nav .workspace:has(.tr-wrap) {
 
   // Keyboard navigation for segment stepping
   document.addEventListener('keydown', (e) => {
+    if (e.key === 'Escape' && deleteSegmentModal.style.display !== 'none') {
+      e.preventDefault();
+      closeDeleteSegmentModal();
+      return;
+    }
     const tag = e.target.tagName;
     if (tag === 'INPUT' || tag === 'TEXTAREA' || e.target.isContentEditable) return;
     if (document.getElementById('trImageModal')) return;
+    if (deleteSegmentModal.style.display !== 'none') return;
     if (e.key === ']') { e.preventDefault(); navigateSegment(1); }
     else if (e.key === '[') { e.preventDefault(); navigateSegment(-1); }
   });
 
+  window.addEventListener('click', (event) => {
+    if (event.target === deleteSegmentModal) {
+      closeDeleteSegmentModal();
+    }
+  });
+
   // Click on timeline to set range
   timeline.addEventListener('click', (e) => {
     if (e.target.closest('.tr-sel')) return;
@@ -3143,53 +3313,57 @@ body.has-date-nav .workspace:has(.tr-wrap) {
     return pos.charAt(0).toUpperCase() + pos.slice(1);
   }
 
-  // Clear selection and reset UI state
-  function clearSegmentSelection() {
-    selectedSegment = null;
-    segmentData = null;
-    currentVideoFiles = {};
-    activeTab = null;
-    tabPanes = {};
-    screenDecoded = false;
-    frameCapture.clear();
-    allScreenFrames = [];
-    currentFrameIndex = -1;
-    groupEntriesByIdx.clear();
-
-    // Stop and clear audio player reference
-    panel.querySelectorAll('audio').forEach(audio => audio.pause());
-
-    // Hide delete button
-    deleteBtn.classList.remove('visible');
+  function showDeleteSegmentModal() {
+    if (!selectedSegment) return;
 
-    // Clear URL hash
-    history.replaceState(null, '', window.location.pathname);
+    const seg = selectedSegment;
+    const totalBytes = (segmentData?.media_sizes?.audio || 0) + (segmentData?.media_sizes?.screen || 0);
+    const sizeLine = totalBytes > 0
+      ? `<p class="tr-delete-segment-size">Current raw media size: ${formatSize(totalBytes)}</p>`
+      : '';
+
+    deleteSegmentModalBody.innerHTML = `
+      <p class="tr-delete-segment-meta"><strong>${escapeHtml(seg.stream)}</strong> · <strong>${escapeHtml(day)}</strong> · <strong>${escapeHtml(seg.start)} - ${escapeHtml(seg.end)}</strong></p>
+      <p>You'll have ~30 seconds to undo this after confirming.</p>
+      <p>This removes:</p>
+      <ul class="tr-delete-segment-list">
+        <li>audio</li>
+        <li>screen recording</li>
+        <li>transcripts</li>
+        <li>derived insights</li>
+      </ul>
+      ${sizeLine}
+    `;
+
+    deleteModalReturnFocus = document.activeElement;
+    deleteModalPrevOverflow = document.body.style.overflow;
+    deleteSegmentModal.style.display = 'flex';
+    document.body.style.overflow = 'hidden';
+    deleteSegmentModalConfirm.focus();
+  }
 
-    // Reset UI
-    titleEl.textContent = 'Transcripts';
-    rangeText.textContent = '';
-    tabsContainer.innerHTML = '';
-    tabsContainer.classList.remove('visible');
-    document.getElementById('trWarningNotice').classList.remove('visible');
-    panel.innerHTML = emptyStateHTML(emptyIcons.day, 'your day at a glance', 'select a segment from the timeline to view its transcript');
+  function closeDeleteSegmentModal(restoreFocus = true) {
+    if (deleteSegmentModal.style.display === 'none') return;
 
-    // Clear active state in zoom view
-    zoomSegments.querySelectorAll('.tr-zoom-pill').forEach(pill => {
-      pill.classList.remove('tr-active');
-    });
+    deleteSegmentModal.style.display = 'none';
+    deleteSegmentModalBody.innerHTML = '';
+    document.body.style.overflow = deleteModalPrevOverflow;
+    deleteModalPrevOverflow = '';
+    if (restoreFocus && deleteModalReturnFocus && document.contains(deleteModalReturnFocus)) {
+      deleteModalReturnFocus.focus();
+    }
+    deleteModalReturnFocus = null;
   }
 
-  // Delete segment handler
-  deleteBtn.addEventListener('click', async () => {
+  async function confirmDeleteSegment() {
     if (!selectedSegment) return;
 
     const seg = selectedSegment;
-    const confirmMsg = `Delete segment ${seg.start} - ${seg.end}?\n\n` +
-      `This will permanently remove all audio, screen recordings, and transcripts for this segment.\n\n` +
-      `This cannot be undone.`;
-
-    if (!confirm(confirmMsg)) return;
+    closeDeleteSegmentModal(false);
+    await performDeleteSegment(seg);
+  }
 
+  async function performDeleteSegment(seg) {
     try {
       const response = await fetch(`/app/transcripts/api/segment/${day}/${seg.stream}/${seg.key}`, {
         method: 'DELETE'
@@ -3200,6 +3374,8 @@ body.has-date-nav .workspace:has(.tr-wrap) {
         throw new Error(data.error || 'Failed to delete segment');
       }
 
+      const data = await response.json();
+
       // Remove segment from local state
       allSegments = allSegments.filter(s => s.key !== seg.key);
 
@@ -3228,6 +3404,17 @@ body.has-date-nav .workspace:has(.tr-wrap) {
           // Range indicators may be stale, but segment is deleted
         });
 
+      if (data.search_index_warning === true) {
+        window.AppServices.notifications.show({
+          app: 'transcripts',
+          icon: '⚠️',
+          title: 'Search index may be stale',
+          message: 'Search results may be briefly stale until the supervisor is restarted and the index is rebuilt.',
+          dismissible: true,
+          autoDismiss: 10000
+        });
+      }
+
     } catch (err) {
       const notice = document.createElement('div');
       notice.className = 'tr-warning-notice visible';
@@ -3241,6 +3428,50 @@ body.has-date-nav .workspace:has(.tr-wrap) {
         setTimeout(() => notice.remove(), 300);
       }, 5000);
     }
+  }
+
+  window.closeDeleteSegmentModal = closeDeleteSegmentModal;
+  window.confirmDeleteSegment = confirmDeleteSegment;
+
+  // Clear selection and reset UI state
+  function clearSegmentSelection() {
+    selectedSegment = null;
+    segmentData = null;
+    currentVideoFiles = {};
+    activeTab = null;
+    tabPanes = {};
+    screenDecoded = false;
+    frameCapture.clear();
+    allScreenFrames = [];
+    currentFrameIndex = -1;
+    groupEntriesByIdx.clear();
+
+    // Stop and clear audio player reference
+    panel.querySelectorAll('audio').forEach(audio => audio.pause());
+
+    // Hide delete button
+    deleteBtn.classList.remove('visible');
+
+    // Clear URL hash
+    history.replaceState(null, '', window.location.pathname);
+
+    // Reset UI
+    titleEl.textContent = 'Transcripts';
+    rangeText.textContent = '';
+    tabsContainer.innerHTML = '';
+    tabsContainer.classList.remove('visible');
+    document.getElementById('trWarningNotice').classList.remove('visible');
+    panel.innerHTML = emptyStateHTML(emptyIcons.day, 'your day at a glance', 'select a segment from the timeline to view its transcript');
+
+    // Clear active state in zoom view
+    zoomSegments.querySelectorAll('.tr-zoom-pill').forEach(pill => {
+      pill.classList.remove('tr-active');
+    });
+  }
+
+  // Delete segment handler
+  deleteBtn.addEventListener('click', () => {
+    showDeleteSegmentModal();
   });
 
 })();
diff --git a/docs/design/voice-server.md b/docs/design/voice-server.md
index df58d85ac..a5b5e1e6a 100644
--- a/docs/design/voice-server.md
+++ b/docs/design/voice-server.md
@@ -470,4 +470,4 @@ Journal-data rule:
 - Fixture briefing date drift: without the `_today()` seam, `briefing.get` and date-window tools would fail when run against the real system clock.
 - `commitments.list.resolution` diminishment is accepted as a Wave 2 known-limit. The ledger surface stores resolution nuance in close-note edits rather than a typed field, so `commitments.list` surfaces `"dropped"` only for dropped items and omits the field otherwise. Full resolution carry-through remains available for a follow-up if post-ship live validation reveals a real need.
 - `sources` leakage: `LedgerItem.sources` is provenance, not model-facing data. The tool layer must strip it before returning commitments to OpenAI.
-- `segment_path()` is write-creating: `journal.get_day` must not call `think.utils.segment_path()` when reading segment summaries because that helper creates directories.
+- `segment_path()` still creates directories by default for write paths; read-only callers (for example segment-summary readers) must pass `create=False`.
diff --git a/talent/activities.py b/talent/activities.py
index 7a72a896f..6a90d87b3 100644
--- a/talent/activities.py
+++ b/talent/activities.py
@@ -46,7 +46,7 @@ logger = logging.getLogger(__name__)
 
 def _list_facets_with_activity_state(day: str, segment: str, stream: str) -> list[str]:
     """Find all facets that have activity_state.json in a segment."""
-    agents_dir = segment_path(day, segment, stream) / "talents"
+    agents_dir = segment_path(day, segment, stream, create=False) / "talents"
     if not agents_dir.is_dir():
         return []
 
@@ -61,7 +61,10 @@ def _list_facets_with_activity_state(day: str, segment: str, stream: str) -> lis
 def _load_activity_state(day: str, segment: str, facet: str, stream: str) -> list[dict]:
     """Load activity_state.json for a facet in a segment. Returns [] on failure."""
     state_path = (
-        segment_path(day, segment, stream) / "talents" / facet / "activity_state.json"
+        segment_path(day, segment, stream, create=False)
+        / "talents"
+        / facet
+        / "activity_state.json"
     )
     if not state_path.exists():
         return []
diff --git a/talent/activity_state.py b/talent/activity_state.py
index 2146f4edc..2df6cb499 100644
--- a/talent/activity_state.py
+++ b/talent/activity_state.py
@@ -124,7 +124,10 @@ def load_previous_state(
     parsed JSON array or None if not found/invalid.
     """
     state_path = (
-        segment_path(day, segment, stream) / "talents" / facet / "activity_state.json"
+        segment_path(day, segment, stream, create=False)
+        / "talents"
+        / facet
+        / "activity_state.json"
     )
     if not state_path.exists():
         return None, None
diff --git a/tests/test_cluster.py b/tests/test_cluster.py
index e048b7fda..6f4de8e52 100644
--- a/tests/test_cluster.py
+++ b/tests/test_cluster.py
@@ -635,3 +635,15 @@ def test_day_path_create_false(tmp_path, monkeypatch):
 
     created = day_path("29990101")
     assert created.exists()
+
+
+def test_find_segment_dir_missing_streamed_segment_does_not_create_directory(
+    tmp_path, monkeypatch
+):
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(tmp_path))
+
+    mod = importlib.import_module("think.cluster")
+    result = mod._find_segment_dir("29990101", "090000_300", "default")
+
+    assert result is None
+    assert not (tmp_path / "chronicle" / "29990101").exists()
diff --git a/tests/test_supervisor.py b/tests/test_supervisor.py
index 019635085..1748d4a04 100644
--- a/tests/test_supervisor.py
+++ b/tests/test_supervisor.py
@@ -8,6 +8,7 @@ import subprocess
 import sys
 from unittest.mock import MagicMock
 
+import psutil
 import pytest
 
 
@@ -610,6 +611,10 @@ def test_supervisor_singleton_lock_acquired(tmp_path, monkeypatch):
     assert (tmp_path / "health" / "supervisor.pid").read_text().strip() == str(
         os.getpid()
     )
+    start_time = float(
+        (tmp_path / "health" / "supervisor.start_time").read_text().strip()
+    )
+    assert start_time > 0
 
 
 def test_supervisor_singleton_lock_blocked(tmp_path, monkeypatch, capsys):
@@ -672,3 +677,64 @@ def test_supervisor_singleton_lock_blocked_with_health(tmp_path, monkeypatch, ca
     assert "PID 12345" in output
     health_mock.assert_called_once_with()
     start_mock.assert_not_called()
+
+
+def test_is_supervisor_up_without_pid_file(tmp_path, monkeypatch):
+    mod = importlib.reload(importlib.import_module("think.supervisor"))
+
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(tmp_path))
+    (tmp_path / "health").mkdir(parents=True, exist_ok=True)
+
+    assert mod.is_supervisor_up() is False
+
+
+def test_is_supervisor_up_with_dead_pid(tmp_path, monkeypatch):
+    mod = importlib.reload(importlib.import_module("think.supervisor"))
+
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(tmp_path))
+    health_dir = tmp_path / "health"
+    health_dir.mkdir(parents=True, exist_ok=True)
+
+    proc = subprocess.Popen(["true"])
+    proc.wait()
+    (health_dir / "supervisor.pid").write_text(str(proc.pid))
+
+    assert mod.is_supervisor_up() is False
+
+
+def test_is_supervisor_up_with_live_pid_missing_start_time(tmp_path, monkeypatch):
+    mod = importlib.reload(importlib.import_module("think.supervisor"))
+
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(tmp_path))
+    health_dir = tmp_path / "health"
+    health_dir.mkdir(parents=True, exist_ok=True)
+    (health_dir / "supervisor.pid").write_text(str(os.getpid()))
+
+    assert mod.is_supervisor_up() is False
+
+
+def test_is_supervisor_up_with_live_pid_mismatched_start_time(tmp_path, monkeypatch):
+    mod = importlib.reload(importlib.import_module("think.supervisor"))
+
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(tmp_path))
+    health_dir = tmp_path / "health"
+    health_dir.mkdir(parents=True, exist_ok=True)
+    (health_dir / "supervisor.pid").write_text(str(os.getpid()))
+    create_time = psutil.Process(os.getpid()).create_time()
+    (health_dir / "supervisor.start_time").write_text(str(create_time + 60))
+
+    assert mod.is_supervisor_up() is False
+
+
+def test_is_supervisor_up_with_matching_process_identity(tmp_path, monkeypatch):
+    mod = importlib.reload(importlib.import_module("think.supervisor"))
+
+    monkeypatch.setenv("_SOLSTONE_JOURNAL_OVERRIDE", str(tmp_path))
+    health_dir = tmp_path / "health"
+    health_dir.mkdir(parents=True, exist_ok=True)
+    (health_dir / "supervisor.pid").write_text(str(os.getpid()))
+    (health_dir / "supervisor.start_time").write_text(
+        str(psutil.Process(os.getpid()).create_time())
+    )
+
+    assert mod.is_supervisor_up() is True
diff --git a/think/cluster.py b/think/cluster.py
index 57ed56a60..0addd5ee8 100644
--- a/think/cluster.py
+++ b/think/cluster.py
@@ -519,7 +519,7 @@ def _find_segment_dir(day: str, segment: str, stream: str | None) -> Path | None
     from think.utils import segment_path as _segment_path
 
     if stream:
-        path = _segment_path(day, segment, stream)
+        path = _segment_path(day, segment, stream, create=False)
         return path if path.is_dir() else None
 
     # Search all streams for this segment
diff --git a/think/supervisor.py b/think/supervisor.py
index 225ad52b3..7addebe11 100644
--- a/think/supervisor.py
+++ b/think/supervisor.py
@@ -17,6 +17,7 @@ from dataclasses import dataclass, field
 from datetime import datetime
 from pathlib import Path
 
+import psutil
 from desktop_notifier import DesktopNotifier, Urgency
 
 from think import routines, scheduler
@@ -513,6 +514,45 @@ def _get_journal_path() -> Path:
     return Path(get_journal())
 
 
+def is_supervisor_up() -> bool:
+    """Return True when supervisor.pid and supervisor.start_time identify a live supervisor process for the current journal."""
+    health_dir = Path(get_journal()) / "health"
+    pid_path = health_dir / "supervisor.pid"
+    try:
+        pid = int(pid_path.read_text().strip())
+    except FileNotFoundError:
+        return False
+    except (OSError, ValueError):
+        return False
+
+    try:
+        os.kill(pid, 0)
+    except ProcessLookupError:
+        return False
+    except PermissionError:
+        return False
+    except OSError:
+        return False
+
+    start_time_path = health_dir / "supervisor.start_time"
+    try:
+        recorded_start = float(start_time_path.read_text().strip())
+    except FileNotFoundError:
+        return False
+    except (OSError, ValueError):
+        return False
+
+    try:
+        create_time = psutil.Process(pid).create_time()
+    except psutil.NoSuchProcess:
+        return False
+    except psutil.Error:
+        return False
+
+    tolerance = 1.5  # drift between time.time() and psutil create_time()
+    return abs(recorded_start - create_time) <= tolerance
+
+
 class RestartPolicy:
     """Track restart attempts and compute backoff delays."""
 
@@ -1510,6 +1550,9 @@ def main() -> None:
             print(f"Supervisor already running{pid_msg}")
         sys.exit(1)
     pid_path.write_text(str(os.getpid()))
+    start_time_path = health_dir / "supervisor.start_time"
+    # Written here, not at _supervisor_start, to minimize drift from psutil create_time().
+    start_time_path.write_text(str(time.time()))
     logging.info("Singleton lock acquired (PID %d)", os.getpid())
 
     # Set up signal handlers
diff --git a/think/utils.py b/think/utils.py
index cea673414..bd9315843 100644
--- a/think/utils.py
+++ b/think/utils.py
@@ -29,6 +29,7 @@ from timefhuman import timefhuman
 from media import MIME_TYPES
 
 DATE_RE = re.compile(r"\d{8}")
+STREAM_RE = re.compile(r"^[a-z0-9][a-z0-9._-]*$")
 CHRONICLE_DIR = "chronicle"
 DEFAULT_STREAM = "_default"
 EXIT_TEMPFAIL = 75
@@ -245,7 +246,7 @@ def updated_days(exclude: set[str] | None = None) -> list[str]:
     return updated
 
 
-def segment_path(day: str, segment: str, stream: str) -> Path:
+def segment_path(day: str, segment: str, stream: str, *, create: bool = True) -> Path:
     """Return absolute path for a segment directory within a stream.
 
     Parameters
@@ -256,14 +257,17 @@ def segment_path(day: str, segment: str, stream: str) -> Path:
         Segment key in HHMMSS_LEN format.
     stream : str
         Stream name (e.g., "archon", "import.apple").
+    create : bool, optional
+        Create the segment directory if it does not exist. Defaults to True.
 
     Returns
     -------
     Path
-        Absolute path to the segment directory (created if it doesn't exist).
+        Absolute path to the segment directory.
     """
-    path = day_path(day) / stream / segment
-    path.mkdir(parents=True, exist_ok=True)
+    path = day_path(day, create=create) / stream / segment
+    if create:
+        path.mkdir(parents=True, exist_ok=True)
     return path