From 434f717ec6240e6fbf480123197faccce543845f Mon Sep 17 00:00:00 2001
From: abrichr <richard.abrich@mldsai.com>
Date: Fri, 28 Aug 2026 11:55:19 -0400
Subject: [PATCH 1/2] fix(benchmark): let a legacy capture directory report its
 migration command

`generate_benchmark_html` wrapped `load_real_capture_data` in
`except (FileNotFoundError, ValueError, KeyError)` to sniff which of two
formats `data_path` holds. `LegacyCaptureError` is a subclass of
`FileNotFoundError`, so the clause caught it too and fell through to
`load_benchmark_data`, which reads a recording directory as a benchmark
result directory with no results in it.

The user pointed `openadapt-viewer benchmark --data DIR` at a pre-2026-07-17
capture, read `Generated: ...`, and opened a viewer holding zero tasks. The
conversion command `recording_db` raises was in the exception nobody saw.

`LegacyCaptureError` is re-raised ahead of the broad clause. The inheritance
stays: the absence of a readable recording.db is a FileNotFoundError, and the
CLI's top-level handler and the scanner's per-directory handler both report
it correctly through that base class without an extra except clause. Its
docstring now states the obligation this defect broke -- a caller catching
FileNotFoundError to mean "try another format" must re-raise this class first.

`_legacy_capture` moves from tests/test_scanner_recording_db.py to
tests/capture_examples.py, so the two test modules that need a legacy
directory build the same one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---
 src/openadapt_viewer/recording_db.py          |  9 ++++-
 .../viewers/benchmark/generator.py            | 13 ++++++
 tests/capture_examples.py                     | 36 +++++++++++++++++
 tests/test_generator.py                       | 35 ++++++++++++++++
 tests/test_scanner_recording_db.py            | 40 +++++--------------
 5 files changed, 103 insertions(+), 30 deletions(-)

diff --git a/src/openadapt_viewer/recording_db.py b/src/openadapt_viewer/recording_db.py
index 699c20f..d108322 100644
--- a/src/openadapt_viewer/recording_db.py
+++ b/src/openadapt_viewer/recording_db.py
@@ -55,7 +55,14 @@ class LegacyCaptureError(FileNotFoundError):
 
     A subclass of ``FileNotFoundError`` because that is what the absence of a
     readable ``recording.db`` is, and because callers that only want to report
-    "this directory is not loadable" need no new except clause.
+    "this directory is not loadable" need no new except clause. The CLI's
+    top-level handler and the scanner's per-directory handler both rely on
+    that.
+
+    The cost of the inheritance is that a caller which catches
+    ``FileNotFoundError`` to mean "try a different format" swallows this too,
+    and the message it carries is the only place the conversion command
+    appears. Such a caller must re-raise this class ahead of the broad clause.
     """
 
 
diff --git a/src/openadapt_viewer/viewers/benchmark/generator.py b/src/openadapt_viewer/viewers/benchmark/generator.py
index f164a27..d2fdf4f 100644
--- a/src/openadapt_viewer/viewers/benchmark/generator.py
+++ b/src/openadapt_viewer/viewers/benchmark/generator.py
@@ -8,6 +8,7 @@
 
 from openadapt_viewer.core.html_builder import HTMLBuilder
 from openadapt_viewer.core.types import BenchmarkRun
+from openadapt_viewer.recording_db import LegacyCaptureError
 from openadapt_viewer.viewers.benchmark.data import create_sample_data, load_benchmark_data
 from openadapt_viewer.viewers.benchmark.real_data_loader import load_real_capture_data
 
@@ -31,6 +32,10 @@ def generate_benchmark_html(
     Returns:
         Path to the generated HTML file
 
+    Raises:
+        LegacyCaptureError: If data_path holds a pre-2026-07-17 ``capture.db``.
+            The message names the conversion command.
+
     POLICY: ALWAYS defaults to real data, from $OPENADAPT_CAPTURE_RECORDING.
     Set use_real_data=False ONLY for unit tests with sample data.
     """
@@ -41,6 +46,14 @@ def generate_benchmark_html(
         # Try to load as capture directory first, fall back to benchmark data
         try:
             run = load_real_capture_data(data_path)
+        except LegacyCaptureError:
+            # Re-raised ahead of the fallback because it is a subclass of
+            # FileNotFoundError and the fallback would otherwise swallow it.
+            # This directory is a recording, not a benchmark result directory,
+            # and load_benchmark_data reads it as an empty one: the user got
+            # "Generated: ..." and a viewer holding zero tasks, while the
+            # conversion command sat unread in this exception.
+            raise
         except (FileNotFoundError, ValueError, KeyError):
             # Fall back to benchmark data format
             run = load_benchmark_data(data_path)
diff --git a/tests/capture_examples.py b/tests/capture_examples.py
index f740050..e08406a 100644
--- a/tests/capture_examples.py
+++ b/tests/capture_examples.py
@@ -9,11 +9,17 @@
 
 Two readers depend on this now -- the catalog scanner and the benchmark
 loader -- so the lookup lives here rather than in either test module.
+
+The legacy format has no committed example, because openadapt-capture stopped
+writing it. ``write_legacy_capture`` below builds one, which is sound for the
+opposite reason: the only thing read from a legacy directory is that it holds a
+``capture.db`` this viewer refuses.
 """
 
 from __future__ import annotations
 
 import os
+import sqlite3
 from pathlib import Path
 
 import pytest
@@ -70,3 +76,33 @@ def require_examples() -> Path:
     if missing:
         pytest.fail(f"{directory} is missing recording.db for: {', '.join(missing)}")
     return directory
+
+
+def write_legacy_capture(directory: Path) -> Path:
+    """Write a pre-#28 capture directory, in the format this viewer refuses.
+
+    The shape is the one openadapt-capture's ``migrate_legacy_capture.py``
+    reads: one ``capture`` row and a generic ``events`` table.
+
+    Args:
+        directory: Where to write it. Created if it does not exist.
+
+    Returns:
+        The directory, now holding a ``capture.db`` and no ``recording.db``.
+    """
+    directory.mkdir(parents=True, exist_ok=True)
+    with sqlite3.connect(directory / "capture.db") as conn:
+        conn.execute(
+            "CREATE TABLE capture (id INTEGER PRIMARY KEY, started_at REAL, "
+            "ended_at REAL, platform TEXT, screen_width INTEGER, "
+            "screen_height INTEGER, pixel_ratio REAL, task_description TEXT)"
+        )
+        conn.execute(
+            "INSERT INTO capture VALUES (1, 1000.0, 1012.0, 'darwin', 1920, 1080, 2.0, 'old')"
+        )
+        conn.execute(
+            "CREATE TABLE events (id INTEGER PRIMARY KEY, timestamp REAL, "
+            "type TEXT, data TEXT, parent_id INTEGER)"
+        )
+        conn.execute("INSERT INTO events VALUES (1, 1000.5, 'mouse.move', '{}', NULL)")
+    return directory
diff --git a/tests/test_generator.py b/tests/test_generator.py
index e1cdcad..0871b7a 100644
--- a/tests/test_generator.py
+++ b/tests/test_generator.py
@@ -1,11 +1,17 @@
 """Tests for HTML generation functionality."""
 
+import json
 from pathlib import Path
 
+import pytest
+
 from openadapt_viewer.core.html_builder import HTMLBuilder
 from openadapt_viewer.core.types import BenchmarkRun
+from openadapt_viewer.recording_db import LegacyCaptureError
 from openadapt_viewer.viewers.benchmark import create_sample_data, generate_benchmark_html
 
+from .capture_examples import write_legacy_capture
+
 
 class TestGenerateBenchmarkHtml:
     """Tests for the generate_benchmark_html function."""
@@ -452,3 +458,32 @@ def test_html_no_xss_vulnerability(self, temp_dir):
         # or other methods - what matters is the dangerous code isn't executable
         assert "<script>alert('xss')</script>" not in html_content or "&lt;script&gt;" in html_content
         assert "<script>evil()" not in html_content or "&lt;script&gt;" in html_content
+
+
+class TestLegacyCaptureDirectoryIsReported:
+    """A legacy capture directory must report the conversion command.
+
+    ``recording_db`` refuses the pre-2026-07-17 ``capture.db`` and raises
+    ``LegacyCaptureError`` naming the migration script. The benchmark generator
+    caught that error as part of a format-sniffing fallback and wrote a viewer
+    holding zero tasks, so the user read ``Generated: ...`` and opened an empty
+    page. The command they needed was in the exception nobody saw.
+    """
+
+    def test_a_legacy_directory_raises_rather_than_generating(self, tmp_path):
+        directory = write_legacy_capture(tmp_path / "old-recording")
+        (directory / "episodes.json").write_text(json.dumps({"episodes": []}))
+        output_path = tmp_path / "benchmark.html"
+
+        with pytest.raises(LegacyCaptureError, match="migrate_legacy_capture.py"):
+            generate_benchmark_html(data_path=directory, output_path=output_path)
+
+        assert not output_path.exists()
+
+    def test_a_directory_of_neither_format_still_falls_back(self, benchmark_data_dir, tmp_path):
+        """The fallback itself stays: benchmark result directories use it."""
+        output_path = tmp_path / "benchmark.html"
+
+        generate_benchmark_html(data_path=benchmark_data_dir, output_path=output_path)
+
+        assert "Test Benchmark" in output_path.read_text()
diff --git a/tests/test_scanner_recording_db.py b/tests/test_scanner_recording_db.py
index 9650473..c57d71b 100644
--- a/tests/test_scanner_recording_db.py
+++ b/tests/test_scanner_recording_db.py
@@ -31,7 +31,13 @@
 from openadapt_viewer.cli import main
 from openadapt_viewer.scanner import RecordingScanner
 
-from .capture_examples import EXAMPLES_ENV, REAL_RECORDINGS, examples_dir, require_examples
+from .capture_examples import (
+    EXAMPLES_ENV,
+    REAL_RECORDINGS,
+    examples_dir,
+    require_examples,
+    write_legacy_capture,
+)
 
 
 @pytest.fixture
@@ -39,30 +45,6 @@ def scanner(tmp_path):
     return RecordingScanner(RecordingCatalog(db_path=str(tmp_path / "catalog.db")))
 
 
-def _legacy_capture(directory: Path) -> Path:
-    """Write a pre-#28 capture directory, in the format this viewer refuses.
-
-    The shape is the one openadapt-capture's ``migrate_legacy_capture.py``
-    reads: one ``capture`` row and a generic ``events`` table.
-    """
-    directory.mkdir(parents=True, exist_ok=True)
-    with sqlite3.connect(directory / "capture.db") as conn:
-        conn.execute(
-            "CREATE TABLE capture (id INTEGER PRIMARY KEY, started_at REAL, "
-            "ended_at REAL, platform TEXT, screen_width INTEGER, "
-            "screen_height INTEGER, pixel_ratio REAL, task_description TEXT)"
-        )
-        conn.execute(
-            "INSERT INTO capture VALUES (1, 1000.0, 1012.0, 'darwin', 1920, 1080, 2.0, 'old')"
-        )
-        conn.execute(
-            "CREATE TABLE events (id INTEGER PRIMARY KEY, timestamp REAL, "
-            "type TEXT, data TEXT, parent_id INTEGER)"
-        )
-        conn.execute("INSERT INTO events VALUES (1, 1000.5, 'mouse.move', '{}', NULL)")
-    return directory
-
-
 class TestCommittedRecordingsAreFound:
     """The defect, stated as the user sees it: scan finds nothing."""
 
@@ -174,7 +156,7 @@ class TestLegacyCapturesAreReportedNotIndexed:
     def test_scan_skips_a_legacy_directory_and_names_the_migration(
         self, scanner, tmp_path, capsys
     ):
-        _legacy_capture(tmp_path / "old-recording")
+        write_legacy_capture(tmp_path / "old-recording")
 
         found = scanner.scan_recording_directory(str(tmp_path))
 
@@ -185,7 +167,7 @@ def test_scan_skips_a_legacy_directory_and_names_the_migration(
         assert "migrate_legacy_capture.py" in out
 
     def test_extract_refuses_a_legacy_directory(self, scanner, tmp_path):
-        directory = _legacy_capture(tmp_path / "old-recording")
+        directory = write_legacy_capture(tmp_path / "old-recording")
 
         with pytest.raises(FileNotFoundError, match="migrate_legacy_capture.py"):
             scanner._extract_recording_info(directory, "old-recording")
@@ -194,7 +176,7 @@ def test_a_converted_directory_is_indexed_without_a_legacy_warning(
         self, scanner, tmp_path, capsys
     ):
         examples = require_examples()
-        directory = _legacy_capture(tmp_path / "demo_new")
+        directory = write_legacy_capture(tmp_path / "demo_new")
         (directory / "recording.db").write_bytes(
             (examples / "demo_new" / "recording.db").read_bytes()
         )
@@ -252,7 +234,7 @@ def test_register_indexes_a_real_recording(self, tmp_path, monkeypatch, capsys):
         assert "Events: 14" in out
 
     def test_register_rejects_a_legacy_directory(self, tmp_path, monkeypatch, capsys):
-        directory = _legacy_capture(tmp_path / "captures" / "old-recording")
+        directory = write_legacy_capture(tmp_path / "captures" / "old-recording")
         _isolate_catalog_home(monkeypatch, tmp_path / "home")
         monkeypatch.setattr(
             sys, "argv", ["openadapt-viewer", "catalog", "register", str(directory)]

From 8cc647650efc08fe3fcff48a025204147fac9c01 Mon Sep 17 00:00:00 2001
From: abrichr <richard.abrich@mldsai.com>
Date: Fri, 28 Aug 2026 11:55:36 -0400
Subject: [PATCH 2/2] fix(capture): inline the episode timeline assets instead
 of linking them

Every generated capture page carried:

    <link rel="stylesheet" href="src/openadapt_viewer/styles/episode_timeline.css">
    <script src="src/openadapt_viewer/components/episode_timeline.js"></script>

Both resolve against the directory the output HTML lands in, so they loaded
only when that file happened to sit at the root of a source checkout. The two
files ship inside the installed package, so anyone who ran
`pip install openadapt-viewer` got two 404s and an episode timeline that was
unstyled and did nothing.

They are now read with `importlib.resources` and inlined, which is how the
page already carries core.css and how PageBuilder already carries its
JavaScript. `importlib.resources` rather than a path built from `__file__`
because the lookup has to answer for an installed distribution, which is the
case that was broken. The script keeps its place in `<head>` without `defer`:
it defines the `EpisodeTimeline` class the page constructs during Alpine's
`init`.

An asset that cannot be read prints a warning and is omitted, rather than
failing the whole page or degrading in silence.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---
 .../viewers/capture/generator.py              | 52 ++++++++++++++++++-
 tests/test_generator.py                       | 44 ++++++++++++++++
 2 files changed, 94 insertions(+), 2 deletions(-)

diff --git a/src/openadapt_viewer/viewers/capture/generator.py b/src/openadapt_viewer/viewers/capture/generator.py
index 6ffb0f8..6b5d28e 100644
--- a/src/openadapt_viewer/viewers/capture/generator.py
+++ b/src/openadapt_viewer/viewers/capture/generator.py
@@ -7,6 +7,7 @@
 from __future__ import annotations
 
 import json
+from importlib import resources
 from pathlib import Path
 from typing import Any
 
@@ -89,12 +90,15 @@ def _generate_viewer_html(
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <title>Capture Viewer - {capture_id}</title>
     <script defer src="https://cdn.jsdelivr.net/npm/alpinejs@3.x.x/dist/cdn.min.js"></script>
-    <link rel="stylesheet" href="src/openadapt_viewer/styles/episode_timeline.css">
-    <script src="src/openadapt_viewer/components/episode_timeline.js"></script>
 
     <style>
         {_get_core_css()}
+        {_get_episode_timeline_css()}
     </style>
+
+    <script>
+        {_get_episode_timeline_js()}
+    </script>
 </head>
 <body style="background: var(--oa-bg-primary); color: var(--oa-text-primary); font-family: var(--oa-font-sans); min-height: 100vh; margin: 0;">
 
@@ -401,6 +405,50 @@ def _generate_viewer_html(
 </html>"""
 
 
+def _read_package_asset(package: str, name: str) -> str:
+    """Return the text of a file that ships inside the installed package.
+
+    The generated page is one file the user can move anywhere, so its stylesheet
+    and its script are inlined rather than linked. A ``<link>`` or ``<script
+    src>`` resolves against the directory the output HTML lands in, which is
+    not a source checkout for anyone who installed this package.
+
+    ``importlib.resources`` is used rather than a path built from ``__file__``
+    so the lookup answers for an installed distribution as well as this tree.
+
+    Args:
+        package: Dotted name of the package holding the file.
+        name: The file's name within that package.
+
+    Returns:
+        The file's text, or an empty string when it cannot be read. An
+        unreadable asset degrades one component of the page rather than
+        failing the whole generation, and says so on stdout.
+    """
+    try:
+        return resources.files(package).joinpath(name).read_text(encoding="utf-8")
+    except (OSError, UnicodeDecodeError, ModuleNotFoundError) as e:
+        # Silence here would render a page whose episode timeline is unstyled
+        # and inert, with nothing anywhere saying why.
+        print(f"Warning: Could not read {package}/{name}, omitting it: {e}")
+        return ""
+
+
+def _get_episode_timeline_css() -> str:
+    """Return the episode timeline's stylesheet, for inlining in a <style>."""
+    return _read_package_asset("openadapt_viewer.styles", "episode_timeline.css")
+
+
+def _get_episode_timeline_js() -> str:
+    """Return the episode timeline's script, for inlining in a <script>.
+
+    It defines the ``EpisodeTimeline`` class the page constructs during Alpine's
+    ``init``, so it must run before that: the caller places it in ``<head>``
+    without ``defer``.
+    """
+    return _read_package_asset("openadapt_viewer.components", "episode_timeline.js")
+
+
 def _get_core_css() -> str:
     """Return core CSS with variables and component styles."""
     # Try to read from core.css file
diff --git a/tests/test_generator.py b/tests/test_generator.py
index 0871b7a..a654788 100644
--- a/tests/test_generator.py
+++ b/tests/test_generator.py
@@ -9,6 +9,7 @@
 from openadapt_viewer.core.types import BenchmarkRun
 from openadapt_viewer.recording_db import LegacyCaptureError
 from openadapt_viewer.viewers.benchmark import create_sample_data, generate_benchmark_html
+from openadapt_viewer.viewers.capture import generate_capture_html
 
 from .capture_examples import write_legacy_capture
 
@@ -487,3 +488,46 @@ def test_a_directory_of_neither_format_still_falls_back(self, benchmark_data_dir
         generate_benchmark_html(data_path=benchmark_data_dir, output_path=output_path)
 
         assert "Test Benchmark" in output_path.read_text()
+
+
+class TestGeneratedPagesAreSelfContained:
+    """A generated page must not link to paths inside a source checkout.
+
+    ``episode_timeline.css`` and ``episode_timeline.js`` were referenced as
+    ``src/openadapt_viewer/...``, which resolves against the directory the
+    output HTML lands in. Both files ship inside the installed package, so for
+    anyone who installed openadapt-viewer the two requests were always 404 and
+    the episode timeline rendered unstyled and inert.
+    """
+
+    def test_the_capture_page_references_no_repository_path(self, tmp_path):
+        output_path = tmp_path / "capture.html"
+
+        generate_capture_html(
+            steps=[{"timestamp": 0.0, "duration": 1.0, "action": {"type": "click"}}],
+            episodes=[{"episode_id": "e1", "name": "Episode", "start": 0.0, "end": 1.0}],
+            output_path=output_path,
+        )
+
+        assert "src/openadapt_viewer/" not in output_path.read_text()
+
+    def test_the_capture_page_carries_the_timeline_css_and_js(self, tmp_path):
+        output_path = tmp_path / "capture.html"
+
+        generate_capture_html(
+            steps=[{"timestamp": 0.0, "duration": 1.0, "action": {"type": "click"}}],
+            output_path=output_path,
+        )
+
+        html = output_path.read_text()
+        assert ".oa-episode-timeline" in html
+        assert "class EpisodeTimeline" in html
+
+    def test_the_benchmark_page_references_no_repository_path(
+        self, sample_benchmark_run, tmp_path
+    ):
+        output_path = tmp_path / "benchmark.html"
+
+        generate_benchmark_html(run_data=sample_benchmark_run, output_path=output_path)
+
+        assert "src/openadapt_viewer/" not in output_path.read_text()