semantic_RAG/html_sectioning.py at main · semanticClimate/semantic_RAG · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
"""
HTML book outline parsing, decimal section numbering, and chunking for RAG.

Expects structured HTML (see docs/HTML_SECTION_NESTING.md): nested <section>
elements with data-outline-level (recommended) and a heading (h1–h6) per section.
"""

from __future__ import annotations

import re
from dataclasses import dataclass
from pathlib import Path
from typing import Iterable, List, Optional, Tuple

from bs4 import BeautifulSoup, NavigableString, Tag

HEADING_TAGS = tuple(f"h{i}" for i in range(1, 7))
MAX_OUTLINE_DEPTH = 6


@dataclass(frozen=True)
class SectionRecord:
    """One indexed section with extractable body text (no nested <section> content)."""

    section_number: str
    title: str
    body: str
    level: int


def load_html_file(path: Path | str) -> str:
    p = Path(path)
    assert p.is_file(), f"HTML book not found at {p.resolve()}"
    return p.read_text(encoding="utf-8", errors="replace")


def find_book_root(soup: BeautifulSoup) -> Tag:
    """Prefer <article id='climate-academy-book'>; fall back to <main> or <body>."""
    for sel in ("article#climate-academy-book", "article.book", "main", "body"):
        found = soup.select_one(sel)
        if found:
            return found
    return soup


def _direct_child_tags(tag: Tag) -> List[Tag]:
    return [c for c in tag.children if isinstance(c, Tag)]


def _section_level_from_attr(tag: Tag) -> Optional[int]:
    raw = tag.get("data-outline-level")
    if raw is None:
        return None
    try:
        n = int(str(raw).strip())
    except ValueError:
        return None
    if 1 <= n <= MAX_OUTLINE_DEPTH:
        return n
    return None


def _first_heading_title(tag: Tag) -> Tuple[Optional[str], Optional[int]]:
    """First h1–h6 in document order within this subtree; returns (title, level 1–6)."""
    for h in tag.find_all(HEADING_TAGS):
        text = h.get_text(separator=" ", strip=True)
        if not text:
            continue
        level = int(h.name[1])
        return text, level
    return None, None


def _heading_from_direct_content(section: Tag) -> Tuple[Optional[str], Optional[int]]:
    """
    Heading that belongs to this section only — not inside nested <section> children.
    """
    for child in _direct_child_tags(section):
        if child.name == "section":
            continue
        if child.name in HEADING_TAGS:
            text = child.get_text(separator=" ", strip=True)
            if text:
                return text, int(child.name[1])
        for h in child.find_all(HEADING_TAGS):
            parent_sec = h.find_parent("section")
            if parent_sec is section and h.get_text(strip=True):
                return h.get_text(separator=" ", strip=True), int(h.name[1])
    return None, None


def _section_title_and_level(tag: Tag, parent_depth: int, default_child_level: int) -> Tuple[str, int]:
    """
    Title from first heading in section. Level from data-outline-level, else heading level,
    else default_child_level (typically parent_depth + 1).
    """
    attr_level = _section_level_from_attr(tag)
    h_title, h_level = _heading_from_direct_content(tag)
    title = h_title or tag.get("aria-label") or ""
    title = re.sub(r"\s+", " ", title).strip()
    if attr_level is not None:
        level = attr_level
    elif h_level is not None:
        level = h_level
    else:
        level = default_child_level
    if level <= parent_depth:
        level = parent_depth + 1
    if level > MAX_OUTLINE_DEPTH:
        level = MAX_OUTLINE_DEPTH
    return title, level


def _split_intro_and_child_sections(section: Tag) -> Tuple[List[Tag], List[Tag]]:
    intro: List[Tag] = []
    children: List[Tag] = []
    for child in _direct_child_tags(section):
        if child.name == "section":
            children.append(child)
        else:
            intro.append(child)
    return intro, children


def _strip_nested_sections(tag: Tag) -> str:
    """Text content of tag with nested <section> subtrees removed (avoid double-counting)."""
    clone = BeautifulSoup(str(tag), "html.parser")
    root = clone.find() or clone
    for nested in root.find_all("section"):
        nested.decompose()
    return root.get_text(separator="\n", strip=True)


def _normalize_whitespace(text: str) -> str:
    text = text.replace("\u00a0", " ")
    text = re.sub(r"[ \t]+\n", "\n", text)
    text = re.sub(r"\n{3,}", "\n\n", text)
    text = re.sub(r" {2,}", " ", text)
    return text.strip()


def _bump_counters(counters: List[int], level: int) -> None:
    """1-based level; increment at depth and zero deeper slots."""
    idx = level - 1
    counters[idx] += 1
    for j in range(level, MAX_OUTLINE_DEPTH):
        counters[j] = 0


def _format_section_number(counters: List[int], level: int) -> str:
    return ".".join(str(counters[i]) for i in range(level))


def _parse_section_tree(section: Tag, counters: List[int], parent_depth: int) -> List[SectionRecord]:
    default_child = min(parent_depth + 1, MAX_OUTLINE_DEPTH)
    title, level = _section_title_and_level(section, parent_depth, default_child)
    _bump_counters(counters, level)
    number = _format_section_number(counters, level)

    intro_tags, child_sections = _split_intro_and_child_sections(section)
    if intro_tags:
        container = section
        body = _normalize_whitespace(
            BeautifulSoup("".join(str(t) for t in intro_tags), "html.parser").get_text(
                separator="\n", strip=True
            )
        )
    else:
        body = ""

    if not body:
        body = _normalize_whitespace(_strip_nested_sections(section))
        for nested in section.find_all("section"):
            nested_body = nested.get_text(separator="\n", strip=True)
            if nested_body and nested_body in body:
                body = body.replace(nested_body, "")
        body = _normalize_whitespace(body)

    out: List[SectionRecord] = []
    if body:
        out.append(
            SectionRecord(section_number=number, title=title, body=body, level=level)
        )

    child_parent_depth = level
    for child in child_sections:
        out.extend(_parse_section_tree(child, counters, child_parent_depth))
    return out


def parse_book_html(html: str) -> List[SectionRecord]:
    """
    Parse book HTML into section records with decimal section_number strings.

    Root: first match of article#climate-academy-book, article.book, main, or body.
    Only direct <section> children of the root are outline roots; if none, one synthetic
    section "1" is created from visible text (nested sections stripped).
    """
    soup = BeautifulSoup(html, "html.parser")
    root = find_book_root(soup)
    top_sections = [c for c in _direct_child_tags(root) if c.name == "section"]
    counters = [0] * MAX_OUTLINE_DEPTH
    records: List[SectionRecord] = []

    if top_sections:
        for sec in top_sections:
            records.extend(_parse_section_tree(sec, counters, parent_depth=0))
        return records

    title, _ = _first_heading_title(root)
    if not title:
        t = root.find(["h1", "h2"])
        title = t.get_text(strip=True) if t else "Book"
    body = _normalize_whitespace(_strip_nested_sections(root))
    if not body:
        body = _normalize_whitespace(root.get_text(separator="\n", strip=True))
    if body:
        counters[0] = 1
        records.append(
            SectionRecord(section_number="1", title=title, body=body, level=1)
        )
    return records


def word_chunks(text: str, chunk_size: int, overlap: int) -> List[str]:
    """Split on words; chunk_size and overlap are word counts."""
    words = text.split()
    if chunk_size <= 0:
        raise ValueError("chunk_size must be positive")
    if overlap < 0 or overlap >= chunk_size:
        raise ValueError("overlap must be in [0, chunk_size)")
    chunks: List[str] = []
    i = 0
    while i < len(words):
        chunks.append(" ".join(words[i : i + chunk_size]))
        i += chunk_size - overlap
    return chunks


@dataclass(frozen=True)
class IndexedChunk:
    """One embedding unit with section metadata."""

    document: str
    section_number: str
    section_title: str
    chunk_index: int


def records_to_indexed_chunks(
    records: Iterable[SectionRecord],
    chunk_size: int,
    chunk_overlap: int,
) -> List[IndexedChunk]:
    out: List[IndexedChunk] = []
    for rec in records:
        for idx, part in enumerate(word_chunks(rec.body, chunk_size, chunk_overlap)):
            header = f"[§ {rec.section_number}"
            if rec.title:
                header += f" — {rec.title}"
            header += "]"
            doc = f"{header}\n{part}"
            out.append(
                IndexedChunk(
                    document=doc,
                    section_number=rec.section_number,
                    section_title=rec.title,
                    chunk_index=idx,
                )
            )
    return out


def format_passage_for_prompt(section_number: str, section_title: str, body: str) -> str:
    """Format a retrieved chunk for the LLM (strip duplicate bracket line if present)."""
    t = body.strip()
    if t.startswith("[§"):
        return t
    line = f"[§ {section_number}"
    if section_title:
        line += f" — {section_title}"
    line += "]"
    return f"{line}\n{t}"


def parse_html_path_to_chunks(
    path: Path | str,
    chunk_size: int,
    chunk_overlap: int,
) -> List[IndexedChunk]:
    html = load_html_file(path)
    records = parse_book_html(html)
    return records_to_indexed_chunks(records, chunk_size, chunk_overlap)