TIKAZI/TIKAZ-AI-Skills / suites/context-economy
context-economy
FlaggedAutomatically prepare fidelity-first, bounded context when users paste text, attach files or folders, provide webpages or PDFs for Markdown conversion, continue long conversations, or supply documents containing images and complex tables. Route each input through Text, Hybrid, or Source; reduce exact or formatting-only repetition; protect facts and source anchors; queue only useful visual evidence; preserve uncertain originals; and benchmark efficiency and fidelity separately.
benchmarks/pdf/run_fidelity.py
View on GitHub#!/usr/bin/env python3
"""Convert generated PDFs page-by-page and publish literal fidelity metrics."""
import argparse
import json
import subprocess
import tempfile
from pathlib import Path
import pdfplumber
from pypdf import PdfReader, PdfWriter
def convert_page(converter: Path, page_pdf: Path, output_md: Path) -> None:
completed = subprocess.run(
[str(converter), str(page_pdf), "-o", str(output_md)],
check=False, capture_output=True, text=True,
)
if completed.returncode != 0 or not output_md.is_file():
raise RuntimeError(f"converter failed for {page_pdf.name}: {completed.stderr.strip()}")
def convert_page_pdfplumber(page_pdf: Path, output_md: Path) -> None:
"""Extract literal text and tables with the installed pdfplumber adapter."""
blocks = []
with pdfplumber.open(page_pdf) as document:
page = document.pages[0]
text = (page.extract_text() or "").strip()
if text:
blocks.append(text)
for table in page.extract_tables() or []:
rows = [["" if cell is None else str(cell).replace("\n", " ") for cell in row] for row in table]
if not rows:
continue
width = max(len(row) for row in rows)
rows = [row + [""] * (width - len(row)) for row in rows]
blocks.append("| " + " | ".join(rows[0]) + " |")
blocks.append("| " + " | ".join(["---"] * width) + " |")
blocks.extend("| " + " | ".join(row) + " |" for row in rows[1:])
output_md.write_text("\n\n".join(blocks).strip() + "\n", encoding="utf-8", newline="\n")
def score(expected: dict, markdown: str) -> dict:
categories = {name: [str(value) for value in expected.get(name, [])] for name in ("required_text", "numeric_facts", "table_cells")}
missing = {name: [value for value in values if value not in markdown] for name, values in categories.items()}
pages = int(expected["pages"])
page_anchors = [f"<!-- page: {page} -->" for page in range(1, pages + 1)]
missing["page_anchors"] = [str(page) for page, anchor in enumerate(page_anchors, 1) if anchor not in markdown]
def recall(name):
return 1.0 if not categories[name] else round((len(categories[name]) - len(missing[name])) / len(categories[name]), 4)
return {
"id": expected["id"], "required_text_recall": recall("required_text"),
"numeric_fact_recall": recall("numeric_facts"), "table_cell_recall": recall("table_cells"),
"page_anchor_coverage": round((pages - len(missing["page_anchors"])) / pages, 4),
"missing": missing, "visual_claim_scored": False,
}
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("--adapter", choices=("markitdown", "pdfplumber"), default="markitdown")
parser.add_argument("--converter", type=Path)
parser.add_argument("--ground-truth", required=True, type=Path)
parser.add_argument("--output", required=True, type=Path)
args = parser.parse_args()
converter = args.converter.resolve() if args.converter else None
if args.adapter == "markitdown" and converter is None:
parser.error("--converter is required for the markitdown adapter")
truth_file = args.ground_truth.resolve()
truth = json.loads(truth_file.read_text(encoding="utf-8"))
output = args.output.resolve()
converted = output / "converted"
converted.mkdir(parents=True, exist_ok=True)
results = []
with tempfile.TemporaryDirectory() as temp_dir:
temp = Path(temp_dir)
for document in truth["documents"]:
pdf = (truth_file.parent / document["pdf"]).resolve()
reader = PdfReader(str(pdf))
parts = []
for page_number, page in enumerate(reader.pages, 1):
writer = PdfWriter()
writer.add_page(page)
page_pdf = temp / f"{document['id']}-page-{page_number}.pdf"
page_md = temp / f"{document['id']}-page-{page_number}.md"
with page_pdf.open("wb") as stream:
writer.write(stream)
if args.adapter == "pdfplumber":
convert_page_pdfplumber(page_pdf, page_md)
else:
convert_page(converter, page_pdf, page_md)
parts.append(f"<!-- page: {page_number} -->\n\n{page_md.read_text(encoding='utf-8').strip()}\n")
markdown = "\n".join(parts)
(converted / f"{document['id']}.md").write_text(markdown, encoding="utf-8", newline="\n")
results.append(score(document, markdown))
def aggregate(field):
values = [float(item[field]) for item in results]
return round(sum(values) / len(values), 4)
summary = {
"schema_version": 1, "adapter": args.adapter,
"converter": str(converter) if converter else "installed pdfplumber",
"documents": len(results), "results": results,
"required_text_recall": aggregate("required_text_recall"),
"numeric_fact_recall": aggregate("numeric_fact_recall"),
"table_cell_recall": aggregate("table_cell_recall"),
"page_anchor_coverage": aggregate("page_anchor_coverage"),
"visual_description_accuracy": None,
"claim_boundary": "Generated text/table PDFs; literal recall only. Visual description remains pending.",
}
(output / "metrics.json").write_text(json.dumps(summary, ensure_ascii=False, indent=2, sort_keys=True) + "\n", encoding="utf-8")
print(json.dumps(summary, ensure_ascii=False, sort_keys=True))
if __name__ == "__main__":
main()
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |