Spaces:
Running
Running
File size: 18,545 Bytes
c1708ae | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 | #!/usr/bin/env python3
"""Chunk an already-parsed RICS report (markdown/text) or a PDF/DOCX with the
production segmenters.
Same mechanism as reference ingest:
LLM markers (llm_segmenter) -> else regex (reference_chunker)
Accepts ``.md`` / ``.txt`` (e.g. LlamaParse output) or ``.pdf`` / ``.docx``
(text extracted via ``backend.ingest.doc_extractor``).
Usage (from repo root):
python scripts/chunk_rics_text.py report.llamaparse.md -o ./chunks_out --regex-only --one-chunk-per-section
python scripts/chunk_rics_text.py "E:\\path\\report.pdf" -o ./chunks_out --regex-only --one-chunk-per-section
"""
"""
python scripts/chunk_rics_text.py "E:\my report ai\20b Harvist Road, London, NW6 6SD_llamaparse\20b Harvist Road, London, NW6 6SD.llamaparse.md" -o "E:\my report ai\20b Harvist Road, London, NW6 6SD_llamaparse\chunks" --regex-only --one-chunk-per-section
"""
from __future__ import annotations
import argparse
import json
import re
import sys
from datetime import datetime, timezone
from pathlib import Path
from typing import Any
_REPO_ROOT = Path(__file__).resolve().parents[1]
if str(_REPO_ROOT) not in sys.path:
sys.path.insert(0, str(_REPO_ROOT))
_TEXT_SUFFIXES = {".md", ".markdown", ".txt"}
_DOC_SUFFIXES = {".pdf", ".docx", ".docm"}
_SUPPORTED_SUFFIXES = _TEXT_SUFFIXES | _DOC_SUFFIXES
def _load_dotenv() -> None:
import os
for path in (_REPO_ROOT / ".env", Path.cwd() / ".env"):
if not path.is_file():
continue
for raw in path.read_text(encoding="utf-8").splitlines():
line = raw.strip()
if not line or line.startswith("#") or "=" not in line:
continue
key, _, val = line.partition("=")
key, val = key.strip(), val.strip().strip('"').strip("'")
if key and key not in os.environ:
os.environ[key] = val
def load_input_text(path: Path) -> str:
"""Load report text from markdown/plain text or PDF/DOCX."""
suffix = path.suffix.lower()
if suffix in _TEXT_SUFFIXES:
text = path.read_text(encoding="utf-8")
if suffix in {".md", ".markdown"}:
return normalize_markdown_for_rics(text)
return text
if suffix in _DOC_SUFFIXES:
from backend.ingest.doc_extractor import extract_text
return extract_text(path)
raise ValueError(
f"Unsupported file type {suffix!r}. "
f"Use one of: {', '.join(sorted(_SUPPORTED_SUFFIXES))}"
)
def _chunks_to_rows(chunks: list[Any], *, scrubbed: bool) -> list[dict[str, Any]]:
rows: list[dict[str, Any]] = []
for c in chunks:
text = getattr(c, "text", "") or ""
rows.append(
{
"chunk_id": getattr(c, "chunk_id", "") or "",
"section_id": getattr(c, "section_id", "") or "",
"paragraph_index": getattr(c, "paragraph_index", 1) or 1,
"content_role": getattr(c, "content_role", "body") or "body",
"parent_id": getattr(c, "parent_id", "") or "",
"document_type": getattr(c, "document_type", "reference_report")
or "reference_report",
"is_scrubbed": scrubbed,
"text": text,
}
)
return rows
def normalize_markdown_for_rics(text: str) -> str:
"""Delegate to the production normalizer so CLI and ingest never drift.
Kept as a thin wrapper (same name/signature) for callers of this script.
"""
from backend.rag.reference_chunker import normalize_reference_markdown
return normalize_reference_markdown(text)
def _legacy_normalize_markdown_for_rics(text: str) -> str:
"""Original CLI-local implementation, retained for reference only."""
from backend.domain.section_scope import parent_letter_for_title
raw_lines = (text or "").splitlines()
lines: list[str] = []
for line in raw_lines:
stripped = line.strip()
# Bold leaf: **J1 Insulation**
bold_leaf = re.match(
r"^\*\*\s*([A-N]\d{1,2})\b(?:\s*[:.\-\u2013\u2014]\s*|\s+)([^*]+?)\s*\*\*\s*$",
stripped,
re.IGNORECASE,
)
if bold_leaf:
code = bold_leaf.group(1).upper()
title = bold_leaf.group(2).strip()
lines.append(f"{code} {title}".rstrip() if title else code)
continue
# ATX headings: "# D1 …" / "## Outside the property" / "# J"
m = re.match(r"^(#{1,6})\s+(.*)$", line)
if m:
body = m.group(2).strip()
leaf = re.match(
r"^([A-N]\d{1,2})\b(?:\s*[:.\-\u2013\u2014]\s*|\s+)(.*)$",
body,
re.IGNORECASE,
)
if leaf:
code = leaf.group(1).upper()
title = (leaf.group(2) or "").strip()
lines.append(f"{code} {title}".rstrip() if title else code)
continue
# Lone parent letter: "# J"
if re.fullmatch(r"[A-N]", body, re.IGNORECASE):
lines.append(body.upper())
continue
# "D Outside the property" — but not "D icon Full detail…" chrome.
parent = re.match(r"^([A-N])\s+(.+)$", body, re.IGNORECASE)
if parent:
rest = parent.group(2).strip()
rest_l = rest.lower()
if rest_l.startswith("icon") or rest_l.startswith("logo"):
continue
lines.append(parent.group(1).upper())
lines.append(rest)
continue
lines.append(body)
continue
lines.append(line)
# Second pass: title-only parent lines -> Letter\nTitle (unless already preceded
# by that letter). Prevents "Inside the property" from sticking to D9.
out: list[str] = []
for i, line in enumerate(lines):
letter = parent_letter_for_title(line.strip())
if letter:
prev = out[-1].strip().upper() if out else ""
if prev != letter:
out.append(letter)
out.append(line.strip())
continue
out.append(line)
return "\n".join(out)
def segment_text(
text: str,
*,
source_filename: str,
prefer_llm: bool,
regex_only: bool,
one_chunk_per_section: bool,
) -> tuple[list[Any], str]:
from backend.domain.rics_level3_schema import build_canonical_template_schema
from backend.rag.reference_chunker import build_reference_chunks
schema = build_canonical_template_schema(source_filename="RICS_L3_CANONICAL")
valid_ids = set(schema.section_ids())
# Parent letters are also valid storage keys for A/B/C/K/L/M/N.
valid_ids |= {s.id[0].upper() for s in schema.sections if s.id}
if not regex_only and prefer_llm:
from backend.ingest import llm_segmenter
llm_chunks = llm_segmenter.llm_segment_reference_text(
text,
source_filename=source_filename,
one_chunk_per_section=one_chunk_per_section,
)
if llm_chunks:
return llm_chunks, "llm"
chunks = build_reference_chunks(
text,
source_filename=source_filename,
valid_section_ids=valid_ids,
one_chunk_per_section=one_chunk_per_section,
include_section_headings=True,
)
return chunks, "regex"
def maybe_scrub(chunks: list[Any], *, enabled: bool) -> tuple[list[Any], bool]:
if not enabled:
return chunks, False
from backend.pii import scrubber as pii_scrubber
scrubbed: list[Any] = []
for c in chunks:
text = getattr(c, "text", "") or ""
cleaned, _hits = pii_scrubber.scrub_text(text)
# Chunk is a dataclass-like object; rebuild via replace if available.
replace = getattr(c, "model_copy", None) or getattr(c, "replace", None)
if callable(replace):
try:
scrubbed.append(replace(text=cleaned, is_scrubbed=True))
continue
except TypeError:
pass
try:
from dataclasses import replace as dc_replace
scrubbed.append(dc_replace(c, text=cleaned, is_scrubbed=True))
except Exception:
c.text = cleaned # type: ignore[attr-defined]
if hasattr(c, "is_scrubbed"):
c.is_scrubbed = True # type: ignore[attr-defined]
scrubbed.append(c)
return scrubbed, True
def write_section_files(out_dir: Path, stem: str, rows: list[dict[str, Any]]) -> Path:
sections_dir = out_dir / "sections"
sections_dir.mkdir(parents=True, exist_ok=True)
by_key: dict[str, list[str]] = {}
for row in rows:
if row.get("content_role") == "parent_intro":
key = f"parent_intro_{row.get('parent_id') or 'unknown'}"
else:
key = row.get("section_id") or "unscoped"
by_key.setdefault(key, []).append(row.get("text") or "")
for key, texts in by_key.items():
(sections_dir / f"{stem}__{key}.txt").write_text(
"\n\n".join(t for t in texts if t.strip()),
encoding="utf-8",
)
return sections_dir
def process_one(
path: Path,
out_dir: Path,
*,
prefer_llm: bool,
regex_only: bool,
scrub: bool,
source_name: str | None,
one_chunk_per_section: bool,
) -> dict[str, Any]:
text = load_input_text(path)
source_filename = source_name or path.name
# Optional sidecar of extracted plain text for PDF/DOCX runs.
if path.suffix.lower() in _DOC_SUFFIXES:
out_dir.mkdir(parents=True, exist_ok=True)
(out_dir / f"{path.stem}_extracted.txt").write_text(text, encoding="utf-8")
chunks, method = segment_text(
text,
source_filename=source_filename,
prefer_llm=prefer_llm,
regex_only=regex_only,
one_chunk_per_section=one_chunk_per_section,
)
chunks, was_scrubbed = maybe_scrub(chunks, enabled=scrub)
rows = _chunks_to_rows(chunks, scrubbed=was_scrubbed)
body_sections = sorted(
{
r["section_id"]
for r in rows
if r.get("content_role") != "parent_intro" and r.get("section_id")
}
)
parent_intros = sorted(
{
r["parent_id"]
for r in rows
if r.get("content_role") == "parent_intro" and r.get("parent_id")
}
)
out_dir.mkdir(parents=True, exist_ok=True)
stem = path.stem
file_entry = {
"document_id": "",
"status": "chunked",
"file_size": path.stat().st_size,
"content_hash": "",
"created_at": datetime.now(timezone.utc).isoformat(),
"source_path": str(path.resolve()),
"segmentation_method": method,
"one_chunk_per_section": one_chunk_per_section,
"chunk_count": len(rows),
"sections": body_sections,
"parent_intro_sections": parent_intros,
"chunks": rows,
}
# Tenant-manifest shape: keyed by source filename
extracted = {source_filename: file_entry}
extracted_path = out_dir / "extracted_chunks.json"
extracted_path.write_text(
json.dumps(extracted, indent=2, ensure_ascii=False), encoding="utf-8"
)
# Flat list sidecar (easier to inspect)
chunks_only = {
"source_filename": source_filename,
"segmentation_method": method,
"chunk_count": len(rows),
"chunks": rows,
}
chunks_only_path = out_dir / "chunks_only.json"
chunks_only_path.write_text(
json.dumps(chunks_only, indent=2, ensure_ascii=False), encoding="utf-8"
)
manifest = {
"source_file": str(path.resolve()),
"source_filename": source_filename,
"extracted_at": file_entry["created_at"],
"segmentation_method": method,
"scrubbed": was_scrubbed,
"chunk_count": len(rows),
"sections": body_sections,
"parent_intro_sections": parent_intros,
"outputs": {
"extracted_chunks": str(extracted_path.resolve()),
"chunks_only": str(chunks_only_path.resolve()),
},
}
manifest_path = out_dir / f"{stem}_chunk_manifest.json"
manifest_path.write_text(
json.dumps(manifest, indent=2, ensure_ascii=False), encoding="utf-8"
)
sections_dir = write_section_files(out_dir, stem, rows)
manifest["outputs"]["sections_dir"] = str(sections_dir.resolve())
manifest_path.write_text(
json.dumps(manifest, indent=2, ensure_ascii=False), encoding="utf-8"
)
return {
"method": method,
"chunk_count": len(rows),
"sections": body_sections,
"out_dir": str(out_dir.resolve()),
"extracted_chunks": str(extracted_path.resolve()),
}
def parse_args(argv: list[str] | None = None) -> argparse.Namespace:
p = argparse.ArgumentParser(
description=(
"Create RICS L3 chunks from an existing .md/.txt using "
"llm_segmenter + reference_chunker (same as reference ingest)."
)
)
p.add_argument(
"inputs",
nargs="+",
type=Path,
help="Report path(s): .md/.txt or .pdf/.docx",
)
p.add_argument(
"-o",
"--output",
type=Path,
default=None,
help=(
"Output root (writes under <output>/<source_stem>/ so the folder "
"names the PDF/MD; default: <source_dir>/<stem>_chunks)"
),
)
p.add_argument(
"--regex-only",
action="store_true",
help="Skip LLM; use reference_chunker regex only (default: LLM then regex)",
)
p.add_argument(
"--one-chunk-per-section",
action="store_true",
help=(
"Emit exactly one chunk per subsection / parent-intro / parent body "
"(no paragraph or max-char splitting)"
),
)
p.add_argument(
"--multi-chunk",
action="store_true",
help=(
"Allow multiple chunks per subsection when bodies exceed "
"reference_paragraph_max_chars (default unless --one-chunk-per-section "
"or REFERENCE_ONE_CHUNK_PER_SECTION=true)"
),
)
p.add_argument(
"--scrub",
action="store_true",
help="Run PII scrubber on chunk text before save",
)
p.add_argument(
"--source-name",
default=None,
help="Override source_filename used in chunk_ids (default: input basename)",
)
return p.parse_args(argv)
def main(argv: list[str] | None = None) -> int:
_load_dotenv()
args = parse_args(argv)
paths: list[Path] = []
for p in args.inputs:
if not p.is_file():
print(f"Error: not found: {p}", file=sys.stderr)
continue
if p.suffix.lower() not in _SUPPORTED_SUFFIXES:
print(
f"Error: unsupported type {p.suffix!r} for {p.name} "
f"(use {', '.join(sorted(_SUPPORTED_SUFFIXES))})",
file=sys.stderr,
)
continue
paths.append(p)
if not paths:
return 1
prefer_llm = not args.regex_only
from backend.config import settings
if args.one_chunk_per_section and args.multi_chunk:
print(
"Error: pass only one of --one-chunk-per-section / --multi-chunk",
file=sys.stderr,
)
return 1
if args.one_chunk_per_section:
one_chunk = True
elif args.multi_chunk:
one_chunk = False
else:
one_chunk = bool(settings.reference_one_chunk_per_section)
ok = 0
for path in paths:
# Always keep the source stem in the output path so PDF vs MD runs
# do not overwrite each other and the folder names the source file.
if args.output is None:
out_dir = path.parent / f"{path.stem}_chunks"
else:
out_dir = args.output / path.stem
print(
f"Chunking {path.name} "
f"(one_chunk_per_section={one_chunk}) ..."
)
try:
result = process_one(
path,
out_dir,
prefer_llm=prefer_llm,
regex_only=args.regex_only,
scrub=args.scrub,
source_name=args.source_name,
one_chunk_per_section=one_chunk,
)
except Exception as exc:
print(f" Error: {exc}", file=sys.stderr)
continue
ok += 1
print(
f" method={result['method']} chunks={result['chunk_count']} "
f"sections={len(result['sections'])}"
)
print(f" saved: {result['extracted_chunks']}")
return 0 if ok else 1
if __name__ == "__main__":
raise SystemExit(main())
# ---------------------------------------------------------------------------
# Run examples (from Report-genius-ai repo root):
#
# Chunk LlamaParse markdown:
# python scripts/chunk_rics_text.py "E:\my report ai\out_llamaparse\5 Hillcrest Avenue, Pinner, HA5 1AJ.llamaparse.md" -o "E:\my report ai\out_llamaparse" --regex-only --one-chunk-per-section
#
# python scripts/chunk_rics_text.py "E:\my report ai\out_llamaparse\1a Woodland Hill London SE19 1PB.llamaparse.md" -o "E:\my report ai\out_llamaparse" --regex-only --one-chunk-per-section
#
# Chunk Textract markdown:
# python scripts/chunk_rics_text.py "E:\my report ai\out_textract\1a Woodland Hill London SE19 1PB.textract.md" -o "E:\my report ai\out_textract" --regex-only --one-chunk-per-section
#
# Chunk a PDF directly (local extract, NOT LlamaParse/Textract):
# python scripts/chunk_rics_text.py "E:\my report ai\5 Hillcrest Avenue, Pinner, HA5 1AJ.pdf" -o "E:\my report ai\out_llamaparse" --regex-only --one-chunk-per-section
#
# One-shot PDF -> LlamaParse MD -> chunks:
# python scripts/rics_pdf_to_chunks.py "E:\my report ai\5 Hillcrest Avenue, Pinner, HA5 1AJ.pdf" -o "E:\my report ai\out_llamaparse" --regex-only --one-chunk-per-section
#
# One-shot PDF -> Textract MD -> chunks:
# python scripts/rics_pdf_textract_to_chunks.py "E:\my report ai\1a Woodland Hill London SE19 1PB.pdf" -o "E:\my report ai\out_textract" --regex-only --one-chunk-per-section
#
# Chunk existing Textract MD via one-shot helper:
# python scripts/rics_pdf_textract_to_chunks.py --from-md "E:\my report ai\out_textract\1a Woodland Hill London SE19 1PB.textract.md" -o "E:\my report ai\out_textract" --regex-only --one-chunk-per-section
# ---------------------------------------------------------------------------
|