Spaces:
Sleeping
Sleeping
File size: 7,762 Bytes
399944f | 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 | from __future__ import annotations
"""
Minimal entry point to test the Extractor component in isolation.
Pipeline:
1. Load documents from a text or PDF corpus & chunk them (Chunker).
2. Decompose each document into atomic sentences/qualities (Decomposer).
3. Extract S-P-O triplets for each atomic sentence (Triplet Extractor).
4. Print or optionally write results to a JSON file.
Configuration is driven by environment variables (see ExtractorConfig.from_env).
"""
import json
import os
from dataclasses import dataclass
from typing import Literal, Tuple, List
from rich.console import Console
from rich.table import Table
from rich.panel import Panel
from rich.text import Text
console = Console()
from kbdebugger.utils.warnings_config import install_warning_filters
from kbdebugger.compat.langchain import Document
from kbdebugger.extraction.text_to_sentences import extract_txt_sentences
# from kbdebugger.extraction.pdf_to_sentences import extract_pdf_sentences
from kbdebugger.extraction.pdf_to_chunks import extract_pdf_chunks
from kbdebugger.extraction.decompose import decompose, DecomposeMode
from kbdebugger.extraction.triplet_extraction_batch import extract_triplets_batch
from kbdebugger.types import ExtractionResult, SourceKind
from kbdebugger.extraction.types import Qualities
# ---------------------------------------------------------------------------
# Configuration
# ---------------------------------------------------------------------------
@dataclass(frozen=True)
class ExtractorConfig:
"""
Runtime configuration for the Extractor main script.
Environment variables:
- KB_SOURCE_KIND: one of
* "TEXT" → load plain-text file via `text_to_sentences`
* "PDF_SENTENCES" → load PDF and split directly into sentences
* "PDF_CHUNKS" → load PDF and split into larger chunks
Default: "TEXT"
- KB_TEXT_PATH: path to txt file (used when KB_SOURCE_KIND == "TEXT")
Default: "data/DSA/DSA_knowledge.txt"
- KB_PDF_PATH: path to pdf file (used when KB_SOURCE_KIND starts with "PDF")
Default: "data/SDS/example.pdf"
- KB_BATCH_SIZE: number of atomic sentences per LLM batch for triplet extraction.
Default: 5
- KB_OUTPUT_PATH: optional path to write JSON results.
If empty, results are only printed to stdout.
"""
source_kind: SourceKind
text_path: str
pdf_path: str
batch_size: int
output_path: str | None
@classmethod
def from_env(cls) -> "ExtractorConfig":
source_raw = os.getenv("KB_SOURCE_KIND", "TEXT").upper().strip()
if source_raw not in {"TEXT", "PDF_SENTENCES", "PDF_CHUNKS"}:
raise ValueError(f"Invalid KB_SOURCE_KIND={source_raw!r}")
text_path = os.getenv("KB_TEXT_PATH", "data/DSA/DSA_knowledge.txt").strip()
pdf_path = os.getenv("KB_PDF_PATH", "data/SDS/InstructCIR.pdf").strip()
batch_size_str = os.getenv("KB_BATCH_SIZE", "5").strip()
try:
batch_size = max(1, int(batch_size_str))
except ValueError:
batch_size = 5
output_raw = os.getenv("KB_OUTPUT_PATH", "").strip()
output_path = output_raw or None
return cls(
source_kind=source_raw, # type: ignore[arg-type]
text_path=text_path,
pdf_path=pdf_path,
batch_size=batch_size,
output_path=output_path,
)
# ---------------------------------------------------------------------------
# High-level steps
# ---------------------------------------------------------------------------
def load_documents(cfg: ExtractorConfig) -> Tuple[List[Document], DecomposeMode]:
"""
Use the Chunker submodule to load a corpus into LangChain Documents.
Returns (documents, decompose_mode).
"""
match cfg.source_kind:
case "TEXT":
docs = extract_txt_sentences(cfg.text_path)
mode = DecomposeMode.SENTENCES
# case "PDF_SENTENCES":
# docs = extract_pdf_sentences(cfg.pdf_path)
# mode = DecomposeMode.SENTENCES
case "PDF_CHUNKS":
docs = extract_pdf_chunks(cfg.pdf_path)
mode = DecomposeMode.CHUNKS
if not docs:
raise ValueError("No documents loaded. Check your paths and source kind.")
return (docs, mode)
def run_extractor(cfg: ExtractorConfig) -> List[ExtractionResult]:
"""
End-to-end run of the Extractor component:
Corpus -> Chunked Documents -> Atomic sentences/Qualities -> Triplets
"""
# 1. Load corpus & chunk into LangChain Documents
docs, mode = load_documents(cfg)
# 2. Decompose each document into atomic sentences/qualities
all_qualities: Qualities = []
for doc in docs:
qualities = decompose(text=doc.page_content, mode=mode)
all_qualities.extend(qualities)
if not all_qualities:
raise ValueError("Decomposition produced no qualities.")
# 3. Triplet extraction in batches
results = extract_triplets_batch(all_qualities, batch_size=cfg.batch_size)
return results
def print_results(results: List[ExtractionResult]) -> None:
"""
Pretty-print extraction results to stdout using rich.
"""
console.rule("[bold blue]Triplet Extraction Results")
for i, item in enumerate(results, start=1):
sentence: str = item.get("sentence", "")
triplets: List[Tuple[str, str, str]] = item.get("triplets", [])
# Header panel for each atomic sentence
console.print(
Panel.fit(
Text(sentence, style="bold white"),
title=f"[cyan]Sentence {i}",
border_style="cyan",
)
)
if not triplets:
console.print("[yellow]No triplets extracted.[/yellow]\n")
continue
# Table of triplets
table = Table(
show_header=True,
header_style="bold magenta",
show_lines=True,
border_style="dim",
)
table.add_column("Subject", style="green", ratio=2)
table.add_column("Relation", style="white", ratio=3)
table.add_column("Object", style="green", ratio=2)
for subj, obj, rel in triplets:
table.add_row(subj, rel, obj)
console.print(table)
console.print() # extra newline
def save_results_json(results: List[ExtractionResult], path: str) -> None:
"""
Write extraction results to a JSON file.
"""
data = {
"results": results,
}
os.makedirs(os.path.dirname(path), exist_ok=True)
with open(path, "w", encoding="utf-8") as f:
json.dump(data, f, ensure_ascii=False, indent=2)
print(f"\n[INFO] Wrote JSON results to {path}")
# ---------------------------------------------------------------------------
# Main
# ---------------------------------------------------------------------------
def main() -> None:
install_warning_filters()
cfg = ExtractorConfig.from_env()
console.rule("[bold green]KBDEBUGGER Extractor")
console.print(f"[bold]Source kind:[/bold] [cyan]{cfg.source_kind}[/cyan]")
if cfg.source_kind == "TEXT":
console.print(f"[bold]Text path:[/bold] [white]{cfg.text_path}[/white]")
else:
console.print(f"[bold]PDF path:[/bold] [white]{cfg.pdf_path}[/white]")
console.print(f"[bold]Batch size:[/bold] [yellow]{cfg.batch_size}[/yellow]")
if cfg.output_path:
console.print(f"[bold]Output file:[/bold] [green]{cfg.output_path}[/green]")
console.print() # spacing
results = run_extractor(cfg)
if cfg.output_path:
save_results_json(results, cfg.output_path)
print_results(results)
if __name__ == "__main__":
main() |