feat: re-land vocab demo on main via PyPI latincy-vocab

Restores pages/14_vocab_demo.py (removed from main in 1d0c8f9 pending the
PyPI publish) and depends on latincy-vocab>=0.1.0 from PyPI instead of the
stale file://../latincy-vocab local ref that lived on the vocab-demo branch.
Cherry-picked the demo file only; the old vocab-demo branch is a stale
ancestor of main and is not merged. Verified the PyPI wheel ships the
vocabbuilder package the demo imports; app.py already links the page.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

pages/14_vocab_demo.py

@@ -0,0 +1,167 @@
+import streamlit as st
+
+from dcc_helpers import is_dcc_core
+from lexicon_helpers import load_lexicon_pipeline
+from vocabbuilder.core.config import PipelineConfig
+from vocabbuilder.core.models import VocabList
+from vocabbuilder.data.gloss_provider import GlossProvider
+from vocabbuilder.processors.vocab_core import build_vocab_list
+from vocabbuilder.utils.normalization import to_u_form
+
+st.set_page_config(page_title="Vocab Builder Demo", layout="wide")
+st.sidebar.header("Vocab Builder Demo")
+
+st.title("Latin Vocabulary Builder")
+st.markdown(
+    """
+Generate a textbook-style vocabulary list from any Latin passage using
+[latincy-vocab](https://github.com/latincy/latincy-vocab) —
+lemmatized, glossed, and sortable.
+"""
+)
+
+DEFAULT_TEXT = (
+    "Haec narrantur a poetis de Perseo. Perseus filius erat Iovis, maximi "
+    "deorum; avus eius Acrisius appellabatur. Acrisius volebat Perseum nepotem "
+    "suum necare; nam propter oraculum puerum timebat."
+)
+
+
+@st.cache_resource(show_spinner="Loading vocabulary pipeline…")
+def load_resources():
+    # Use the shared lexicon pipeline so token._.lexicon is populated,
+    # giving us citation forms (principal parts, gen+gender, -a,-um).
+    nlp = load_lexicon_pipeline("la_core_web_lg")
+    config = PipelineConfig(spacy_model="la_core_web_lg", spacy_disable=[])
+    config.resolve_data_paths()
+    gloss_provider = GlossProvider(config.glosses_path) if config.glosses_path else None
+    return nlp, config, gloss_provider
+
+
+nlp, _config, _gloss_provider = load_resources()
+
+
+def process(text: str) -> VocabList:
+    doc = nlp(text)
+    vocab = build_vocab_list(doc, _config)
+    if _gloss_provider:
+        for entry in vocab:
+            result = _gloss_provider.lookup(entry.lemma, entry.pos, _config.max_glosses)
+            if result:
+                entry.glosses = result.glosses
+                entry.display_lemma = result.display_lemma
+            else:
+                entry.display_lemma = _gloss_provider.get_display_lemma(entry.lemma)
+    return vocab
+
+
+tab1, tab2 = st.tabs(["Vocab List", "About"])
+
+with tab1:
+    col_input, col_vocab = st.columns([2, 3])
+
+    with col_input:
+        text = st.text_area(
+            "Latin text:",
+            value=DEFAULT_TEXT,
+            height=280,
+            key="vocab_text",
+        )
+        run = st.button("Build Vocabulary List", key="vocab_run")
+
+    if run:
+        with st.spinner("Processing…"):
+            vocab: VocabList = process(text)
+        st.session_state["vocab_list"] = vocab
+        st.session_state["vocab_sort"] = "alpha"
+
+    with col_vocab:
+        if "vocab_list" in st.session_state:
+            vocab: VocabList = st.session_state["vocab_list"]
+
+            s1, s2, s3 = st.columns(3)
+            with s1:
+                if st.button("Alphabetical", key="sort_alpha", use_container_width=True):
+                    st.session_state["vocab_sort"] = "alpha"
+            with s2:
+                if st.button("First Occurrence", key="sort_occ", use_container_width=True):
+                    st.session_state["vocab_sort"] = "occurrence"
+            with s3:
+                if st.button("Frequency", key="sort_freq", use_container_width=True):
+                    st.session_state["vocab_sort"] = "freq"
+
+            hide_dcc = st.checkbox(
+                "Hide DCC Core words",
+                value=False,
+                key="hide_dcc",
+                help="Remove high-frequency words from the DCC Core Latin Vocabulary",
+            )
+
+            sort_key = st.session_state.get("vocab_sort", "alpha")
+            if sort_key == "alpha":
+                sorted_vocab = vocab.by_alpha()
+            elif sort_key == "occurrence":
+                sorted_vocab = vocab.by_first_occurrence()
+            else:
+                sorted_vocab = vocab.by_frequency()
+
+            visible = [e for e in sorted_vocab if e.headword and e.headword[0].isalpha()]
+            dcc_count = sum(1 for e in visible if is_dcc_core(to_u_form(e.lemma)))
+            new_count = len(visible) - dcc_count
+
+            if hide_dcc:
+                visible = [e for e in visible if not is_dcc_core(to_u_form(e.lemma))]
+
+            st.caption(
+                f"{len(visible)} entries shown · {new_count} new, {dcc_count} DCC core · "
+                "assembled from probabilistic LatinCy pipeline annotations — verify before use"
+            )
+
+            lines = []
+            for entry in visible:
+                in_dcc = is_dcc_core(to_u_form(entry.lemma))
+                hw_text = entry.headword.lower()
+                hw = f"<strong>{hw_text}</strong>"
+                pm = f" <em>{entry.pos_marker}</em>" if entry.pos_marker else ""
+                gloss = f", {entry.short_gloss}" if entry.short_gloss else ""
+                freq = f" <em>×{entry.frequency}</em>" if entry.frequency > 1 else ""
+                dcc_tag = (
+                    " <span style='color:#09a3d5;font-size:0.75em'>dcc</span>"
+                    if in_dcc else ""
+                )
+                lines.append(
+                    f"<p style='margin:0 0 2px 0;line-height:1.4'>"
+                    f"{hw}{pm}{gloss}{freq}{dcc_tag}</p>"
+                )
+
+            st.markdown("".join(lines), unsafe_allow_html=True)
+
+with tab2:
+    st.markdown(
+        """
+### About
+
+This demo uses **latincy-vocab** (`vocabbuilder.VocabPipeline`) to process a Latin
+passage end-to-end:
+
+1. **spaCy** tokenizes and annotates (lemma, POS, morphology)
+2. **latincy-lexicon** supplies Whitaker's Words glosses, display lemmas, and
+   citation forms (principal parts for verbs; nominative, genitive, and gender
+   for nouns; nominative with -a, -um endings for adjectives)
+3. `VocabList` deduplicates by lemma+POS and exposes three orderings
+
+The **Hide DCC Core words** checkbox filters out the ~1,000 high-frequency lemmas
+from the [DCC Core Latin Vocabulary](https://dcc.dickinson.edu/vocab/core-vocabulary),
+leaving only words a student would need to look up.
+
+> **Note:** vocabulary entries are assembled from probabilistic LatinCy pipeline
+> annotations and should be verified before use in publication or teaching.
+
+The default text is §1 of Ritchie's *Fabulae Faciles* (1902).
+
+**Sort modes**
+- *Alphabetical* — traditional glossary order
+- *First Occurrence* — passage reading order (as in textbook footnotes)
+- *Frequency* — most common lemmas first
+"""
+    )

requirements.txt

@@ -10,5 +10,6 @@ spacy-streamlit==1.0.6
 latincy-diacritics @ https://huggingface.co/latincy/latincy-diacritics/resolve/main/latincy_diacritics-0.1.0-py3-none-any.whl
 latincy-preprocess>=0.2.0
 latincy-lexicon>=0.5.0
+latincy-vocab>=0.1.0
 streamlit==1.45.1
 watchdog==6.0.0