Skip to content

VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.

18.2 KB
Raw Blame
"""Text / EPUB import → chapter-delimited script.

Pure helpers, tested without a server. The EPUB case builds a minimal valid
EPUB zip in memory (no fixture file, no new dep).
"""
from __future__ import annotations

import asyncio
import io
import zipfile

import pytest
from fastapi import UploadFile

from services.longform_import import (
    chapterize_plaintext,
    epub_to_chapter_script,
    pdf_to_chapter_script,
)
from services.audiobook import parse_audiobook_script
from api.routers.audiobook import audiobook_import


# ── PDF fixture builder ───────────────────────────────────────────────────
# A minimal hand-built single-page PDF with a Helvetica text layer, so the PDF
# tests need no PDF-authoring dependency (mirrors the in-memory-EPUB approach).

def _make_pdf(lines: list[str], *, content_override: bytes | None = None) -> bytes:
    if content_override is not None:
        content = content_override
    else:
        show = "BT /F1 12 Tf 72 720 Td 16 TL\n"
        for ln in lines:
            esc = ln.replace("\\", "\\\\").replace("(", r"\(").replace(")", r"\)")
            show += f"({esc}) Tj T*\n"
        show += "ET"
        content = show.encode("latin-1")

    objs = [
        b"<< /Type /Catalog /Pages 2 0 R >>",
        b"<< /Type /Pages /Kids [3 0 R] /Count 1 >>",
        b"<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] "
        b"/Resources << /Font << /F1 4 0 R >> >> /Contents 5 0 R >>",
        b"<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>",
        b"<< /Length " + str(len(content)).encode() + b" >>\nstream\n" + content + b"\nendstream",
    ]
    out = io.BytesIO()
    out.write(b"%PDF-1.4\n")
    offsets = []
    for i, body in enumerate(objs, 1):
        offsets.append(out.tell())
        out.write(f"{i} 0 obj\n".encode() + body + b"\nendobj\n")
    xref_pos = out.tell()
    n = len(objs) + 1
    out.write(f"xref\n0 {n}\n".encode())
    out.write(b"0000000000 65535 f \n")
    for off in offsets:
        out.write(f"{off:010d} 00000 n \n".encode())
    out.write(f"trailer\n<< /Size {n} /Root 1 0 R >>\nstartxref\n{xref_pos}\n%%EOF".encode())
    return out.getvalue()


# ── plain text ──────────────────────────────────────────────────────────────

def test_plaintext_leaves_existing_h1_untouched():
    src = "# One\nhello\n\n# Two\nworld"
    assert chapterize_plaintext(src) == src


def test_plaintext_promotes_chapter_lines():
    src = "Chapter 1\nOnce upon a time.\n\nChapter 2\nThe end."
    out = chapterize_plaintext(src)
    assert "# Chapter 1" in out
    assert "# Chapter 2" in out
    # And it parses into two chapters.
    assert len(parse_audiobook_script(out).chapters) == 2


def test_plaintext_ignores_sentences_starting_with_keyword():
    # A long line beginning with "Chapter" is prose, not a heading.
    src = "Chapter books were her favorite thing in the whole wide world to read."
    out = chapterize_plaintext(src)
    assert not out.startswith("# ")


def test_plaintext_promotes_prologue_and_part():
    out = chapterize_plaintext("Prologue\nhi\n\nPart One\nthere")
    assert "# Prologue" in out and "# Part One" in out


def test_plaintext_no_breaks_is_single_chapter():
    out = chapterize_plaintext("just a flat blob of narration with no headings")
    assert len(parse_audiobook_script(out).chapters) == 1


@pytest.mark.parametrize("newline", ["\n", "\r\n", "\r"])
def test_import_endpoint_chapter_titles_across_line_endings(newline):
    from fastapi import FastAPI
    from fastapi.testclient import TestClient
    from api.routers.audiobook import router

    lines = ["Chapter 1", "Once upon a time, the story began.", "", "Chapter 2", "The story ended."]
    app = FastAPI()
    app.include_router(router)
    client = TestClient(app)
    response = client.post(
        "/audiobook/import",
        files={"file": ("book.txt", newline.join(lines).encode(), "text/plain")},
    )
    assert response.status_code == 200
    result = response.json()
    assert result["chapters"] == 2
    assert result["text"] == "# Chapter 1\nOnce upon a time, the story began.\n\n# Chapter 2\nThe story ended."


@pytest.mark.parametrize("newline", ["\n", "\r\n", "\r"])
def test_existing_markdown_headings_after_preamble_stay_verbatim(newline):
    from services.longform_import import chapterize_plaintext

    manuscript = newline.join(["Prologue", "# First", "Body."])
    assert chapterize_plaintext(manuscript) == manuscript


# ── EPUB ────────────────────────────────────────────────────────────────────

def _make_epub_raw(documents: list[bytes]) -> bytes:
    """Build a minimal EPUB: container.xml → content.opf (manifest+spine) →
    one chapter document per entry, written verbatim so a test can control the
    bytes (and therefore the encoding) the importer receives."""
    buf = io.BytesIO()
    with zipfile.ZipFile(buf, "w", zipfile.ZIP_DEFLATED) as z:
        z.writestr("mimetype", "application/epub+zip")
        z.writestr(
            "META-INF/container.xml",
            '<?xml version="1.0"?>'
            '<container version="1.0" xmlns="urn:oasis:names:tc:opendocument:xmlns:container">'
            '<rootfiles><rootfile full-path="OEBPS/content.opf" '
            'media-type="application/oebps-package+xml"/></rootfiles></container>',
        )
        items, refs = [], []
        for i in range(len(documents)):
            items.append(f'<item id="c{i}" href="ch{i}.xhtml" media-type="application/xhtml+xml"/>')
            refs.append(f'<itemref idref="c{i}"/>')
        opf = (
            '<?xml version="1.0"?>'
            '<package xmlns="http://www.idpf.org/2007/opf" version="3.0">'
            f'<manifest>{"".join(items)}</manifest>'
            f'<spine>{"".join(refs)}</spine></package>'
        )
        z.writestr("OEBPS/content.opf", opf)
        for i, document in enumerate(documents):
            z.writestr(f"OEBPS/ch{i}.xhtml", document)
    return buf.getvalue()


def _chapter_html(title: str, body: str) -> str:
    return (
        f"<html><head><title>{title}</title></head><body>"
        f"<h1>{title}</h1><p>{body}</p></body></html>"
    )


def _make_epub(chapters: list[tuple[str, str]]) -> bytes:
    """The common case: UTF-8 chapter documents, no encoding declaration."""
    return _make_epub_raw(
        [_chapter_html(title, body).encode("utf-8") for title, body in chapters]
    )


def test_epub_extracts_chapters_in_spine_order():
    data = _make_epub([("Intro", "Welcome aboard."), ("Finale", "Goodbye now.")])
    script = epub_to_chapter_script(data)
    assert "# Intro" in script and "# Finale" in script
    assert "Welcome aboard." in script and "Goodbye now." in script
    assert script.index("Intro") < script.index("Finale")
    plan = parse_audiobook_script(script)
    assert len(plan.chapters) == 2


def test_epub_chapter_title_keeps_all_of_a_styled_heading():
    # A heading holding inline markup or a <br> reaches the parser in pieces, and
    # only the first piece named the chapter: "Chapter" for "Chapter <em>One</em>",
    # "3" for "<span>3</span> The Calm". The rest was neither title nor body.
    data = _make_epub_raw([
        b"<html><body><h1>Chapter <em>One</em></h1><p>It began.</p></body></html>",
        b"<html><body><h1>CHAPTER II<br/>THE STORM</h1><p>Rain fell.</p></body></html>",
        b'<html><body><h2><span class="num">3</span> The Calm</h2><p>Quiet.</p></body></html>',
        b"<html><body><h1>Chapter&#160;<a href=\"#n\">Four</a></h1><p>Last.</p></body></html>",
    ])
    script = epub_to_chapter_script(data)
    assert "# Chapter One\n" in script
    assert "# CHAPTER II THE STORM\n" in script
    assert "# 3 The Calm\n" in script
    assert "# Chapter Four\n" in script
    assert [c.title for c in parse_audiobook_script(script).chapters] == [
        "Chapter One", "CHAPTER II THE STORM", "3 The Calm", "Chapter Four"]


def test_epub_skips_empty_documents():
    data = _make_epub([("Real", "Has text."), ("Blank", "")])
    plan = parse_audiobook_script(epub_to_chapter_script(data))
    assert len(plan.chapters) == 1


def test_epub_strips_html_tags():
    buf = io.BytesIO()
    with zipfile.ZipFile(buf, "w") as z:
        z.writestr("META-INF/container.xml",
                   '<?xml version="1.0"?><container version="1.0" '
                   'xmlns="urn:oasis:names:tc:opendocument:xmlns:container"><rootfiles>'
                   '<rootfile full-path="content.opf" media-type="application/oebps-package+xml"/>'
                   '</rootfiles></container>')
        z.writestr("content.opf",
                   '<?xml version="1.0"?><package xmlns="http://www.idpf.org/2007/opf" version="3.0">'
                   '<manifest><item id="a" href="a.xhtml" media-type="application/xhtml+xml"/></manifest>'
                   '<spine><itemref idref="a"/></spine></package>')
        z.writestr("a.xhtml",
                   "<html><body><h1>T</h1><p>Hello <b>bold</b> "
                   "<script>ignore()</script>world.</p></body></html>")
    script = epub_to_chapter_script(buf.getvalue())
    assert "ignore()" not in script
    assert "<b>" not in script
    assert "Hello" in script and "world." in script


def test_epub_bad_zip_raises_valueerror():
    with pytest.raises(ValueError):
        epub_to_chapter_script(b"not a zip at all")


def test_epub_total_size_cap_truncates():
    # Cap sits between one and two chapter docs (~1.1 KB uncompressed each), so
    # the first is read and the rest skipped. Caps passed directly (no
    # monkeypatch) so the bound holds regardless of module import path.
    data = _make_epub([("One", "x" * 1000), ("Two", "y" * 1000), ("Three", "z" * 1000)])
    with zipfile.ZipFile(io.BytesIO(data)) as archive:
        overhead = sum(info.file_size for info in archive.infolist() if info.filename.endswith(('container.xml', '.opf')))
    plan = parse_audiobook_script(epub_to_chapter_script(data, max_total_bytes=overhead + 1500))
    assert 1 <= len(plan.chapters) < 3  # capped before reading all three


def test_epub_oversize_entry_skipped():
    data = _make_epub([("Big", "x" * 500)])
    with pytest.raises(ValueError):  # the one entry exceeds the cap → all skipped
        epub_to_chapter_script(data, max_entry_bytes=50)


# ── EPUB: the encoding the document declares ────────────────────────────────
# UTF-8 is only the *default* for an XML document. EPUB 2 books and Calibre
# conversions of older HTML routinely declare something else, and decoding
# those as UTF-8 with errors="ignore" deleted every accent, dash and curly
# quote instead of reading them.

_ACCENTED = "Le café était fermé — hélas."


def test_epub_reads_a_declared_latin1_document():
    document = (
        '<?xml version="1.0" encoding="ISO-8859-1"?>'
        + _chapter_html("Un", "Le caf\xe9 \xe9tait ferm\xe9.")
    ).encode("iso-8859-1")
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert "Le café était fermé." in script


def test_epub_reads_a_declared_windows1252_meta_charset():
    # Written as bytes: 0xE9 is cp1252's "é" and 0x97 its em dash, and neither
    # is valid UTF-8, so errors="ignore" used to drop both.
    document = (
        b'<html><head><meta charset="windows-1252"/><title>Un</title></head>'
        b"<body><h1>Un</h1><p>Le caf\xe9 \x97 h\xe9las.</p></body></html>"
    )
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert "Le café — hélas." in script


def test_epub_reads_a_declared_cjk_document():
    body = "こんにちは世界"  # "hello world", ja
    document = (
        '<?xml version="1.0" encoding="Shift_JIS"?>' + _chapter_html("Ch", body)
    ).encode("shift_jis")
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert body in script


def test_epub_reads_a_utf16_document_by_its_bom():
    document = _chapter_html("Un", _ACCENTED).encode("utf-16-le")
    document = b"\xff\xfe" + document
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert _ACCENTED in script


def test_epub_undeclared_utf8_still_reads():
    """UTF-8 stays the default when nothing is declared."""
    script = epub_to_chapter_script(_make_epub([("Un", _ACCENTED)]))
    assert _ACCENTED in script


def test_epub_reads_a_utf32_document_by_its_bom():
    """The UTF-32 LE mark starts with the UTF-16 LE one, so a BOM table that
    checks UTF-16 first strips two bytes and reads the chapter as NUL-
    interleaved UTF-16."""
    document = b"\xff\xfe\x00\x00" + _chapter_html("Un", _ACCENTED).encode("utf-32-le")
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert _ACCENTED in script


@pytest.mark.parametrize(
    "wide", ["utf-16-le", "utf-16-be", "utf-32-le", "utf-32-be"]
)
def test_epub_reads_a_wide_document_with_no_bom(wide):
    """UTF-16 without a mark is out of spec, but real EPUBs carry it — and the
    declaration is unreadable there, since the ASCII patterns never match
    NUL-interleaved bytes. XML 1.0 §F reads the opening "<" instead."""
    document = _chapter_html("Un", _ACCENTED).encode(wide)
    assert not document.startswith(b"\xff\xfe") and not document.startswith(b"\xfe\xff")
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert _ACCENTED in script


@pytest.mark.parametrize("declared", ["x-not-a-real-charset", "hex_codec", "idna"])
def test_epub_undecodable_declared_encoding_falls_back_instead_of_failing(declared):
    """An encoding Python doesn't have, and a bytes-to-bytes codec that
    resolves but refuses to produce text, both guess rather than fail a book."""
    document = (
        f'<?xml version="1.0" encoding="{declared}"?>'
        + _chapter_html("Un", "A pause — preserved.")
    ).encode("windows-1252")
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert "A pause — preserved." in script


# ── PDF ──────────────────────────────────────────────────────────────────

def test_pdf_extracts_and_chapterizes():
    data = _make_pdf(["Chapter 1", "Once upon a time.", "Chapter 2", "The end."])
    script = pdf_to_chapter_script(data)
    # Chapter-keyword lines from the extracted text become headings …
    assert "# Chapter 1" in script
    assert "# Chapter 2" in script
    # … and the body survives.
    assert "Once upon a time." in script
    # And it parses into two chapters via the shared grammar.
    assert len(parse_audiobook_script(script).chapters) == 2


def test_pdf_without_chapter_markers_is_single_chapter():
    data = _make_pdf(["Just some flowing prose.", "With no chapter headings at all."])
    script = pdf_to_chapter_script(data)
    assert "With no chapter headings" in script
    assert len(parse_audiobook_script(script).chapters) == 1


def test_pdf_corrupt_raises_valueerror():
    with pytest.raises(ValueError):
        pdf_to_chapter_script(b"this is definitely not a pdf")


def test_pdf_image_only_raises_actionable_error():
    # A valid PDF whose page has no text-showing operators → nothing to extract.
    data = _make_pdf([], content_override=b"q Q")  # graphics-only, no BT/Tj
    with pytest.raises(ValueError, match="scanned or image-only"):
        pdf_to_chapter_script(data)


def test_pdf_too_many_pages_guard():
    data = _make_pdf(["Chapter 1", "Hi."])
    with pytest.raises(ValueError, match="too many pages"):
        pdf_to_chapter_script(data, max_pages=0)


# ── import endpoint ────────────────────────────────────────────────────────
# Calls the handler directly (no TestClient → no main+torch import), mirroring
# tests/test_audiobook_cover.py.

def _upload(name: str, data: bytes) -> UploadFile:
    return UploadFile(io.BytesIO(data), filename=name)


def test_import_endpoint_returns_chapter_count():
    # Regression for #543: parsing succeeded but the endpoint then read
    # plan.chapter_count, which didn't exist → 500 AttributeError. Covers the
    # whole class — every import format hits this same return path.
    pdf = _make_pdf(["Chapter 1", "Once upon a time.", "Chapter 2", "The end."])
    for name, data in [
        ("book.pdf", pdf),
        ("book.md", b"# One\nhello\n\n# Two\nworld"),
        ("book.txt", b"just a flat blob of narration with no headings"),
    ]:
        res = asyncio.run(audiobook_import(_upload(name, data)))
        assert isinstance(res["chapters"], int) and res["chapters"] >= 1
        assert res["text"].strip()
    # The two-chapter inputs parse to exactly two chapters.
    assert asyncio.run(audiobook_import(_upload("book.pdf", pdf)))["chapters"] == 2


@pytest.mark.parametrize("wide", ["utf-16-le", "utf-16-be", "utf-32-le", "utf-32-be"])
@pytest.mark.parametrize("whitespace", [" ", "\t\r\n "])
def test_epub_wide_document_with_leading_whitespace(wide, whitespace):
    document = (whitespace + _chapter_html("Un", _ACCENTED)).encode(wide)
    script = epub_to_chapter_script(_make_epub_raw([document]))
    assert _ACCENTED in script
    assert "\x00" not in script


def test_text_after_block_end_tags_starts_a_new_line():
    """#2630: text following a closing block tag must not join the previous one."""
    from services.longform_import import _html_to_title_body

    _, body = _html_to_title_body("<body><p>One.</p>Two.<ul><li>Three</li>Four</ul></body>")
    lines = [ln for ln in body.split("\n") if ln]
    assert lines == ["One.", "Two.", "Three", "Four"]

    _, table = _html_to_title_body("<table><tr><td>A1</td><td>B1</td></tr><tr><th>A2</th>B2</tr></table>")
    assert [ln for ln in table.split("\n") if ln] == ["A1", "B1", "A2", "B2"]

    _, blocks = _html_to_title_body(
        "<div>x<blockquote>quote</blockquote>y</div><section>s</section>z<h4>h</h4>w"
    )
    assert [ln for ln in blocks.split("\n") if ln] == ["x", "quote", "y", "s", "z", "h", "w"]

    title, body = _html_to_title_body("<h1>Chapter <p>One</p></h1><p>Body</p>after")
    assert title == "Chapter One"
    assert [ln for ln in body.split("\n") if ln] == ["Body", "after"]