tests/test_tts_text.py

from app.services import tts_text
from app.services.tts_text import StableSentenceStream, normalize_for_speech


def test_sentence_stream_waits_for_stable_boundary_across_tokens():
    stream = StableSentenceStream()

    assert stream.feed("The first sen") == []
    assert stream.feed("tence is ready. The second") == ["The first sentence is ready."]
    assert stream.finish() == ["The second"]


def test_sentence_stream_does_not_emit_fenced_code_or_table_rows():
    stream = StableSentenceStream()
    markdown = (
        "Read this sentence.\n\n"
        "```python\nprint('Do not read me.')\n```\n\n"
        "| Model | Score |\n| --- | --- |\n| A | 1.0 |\n\n"
        "Read the final sentence."
    )

    sentences = stream.feed(markdown) + stream.finish()

    assert sentences == ["Read this sentence.", "Read the final sentence."]


def test_normalization_reads_prose_and_removes_markup_urls_and_citations():
    text = (
        "## Result\n"
        "See **the main finding** at https://example.com/a and [the appendix](https://example.com/b). "
        "It is grounded [Source: chunk 4] [12]."
    )

    assert normalize_for_speech(text) == (
        "Result See the main finding at and the appendix. It is grounded."
    )


def test_normalization_skips_tables_and_code_without_summarizing_them():
    text = (
        "Before.\n\n"
        "```python\nprint('hidden')\n```\n\n"
        "| A | B |\n| - | - |\n| 1 | 2 |\n\n"
        "- After."
    )

    assert normalize_for_speech(text) == "Before. After."


def test_synthesis_notation_verbalizes_numbers_percentages_and_scientific_symbols():
    assert tts_text.normalize_synthesis_notation(
        "The value is 3.5%, the result is 42, x² + 2, and H₂O."
    ) == (
        "The value is three point five percent, the result is forty-two, "
        "x squared plus two, and H two O."
    )