File size: 3,496 Bytes
9060529 | 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 | """Week 2 demo: topic -> search -> fetch pages -> Reader extracts structured Claims.
Usage:
uv run python scripts/week2_reader.py "solid-state batteries" --pages 3
Writes outputs/week2-<topic>.json with the validated claims. This is still a flat
pipeline (LangGraph orchestration arrives in Week 4); it exists to exercise the
ReaderAgent end to end and to be reused by eval/run_eval.py.
"""
from __future__ import annotations
import argparse
import asyncio
import json
import logging
import re
import sys
from datetime import datetime, timezone
from pathlib import Path
import structlog
sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "src"))
from polymath.agents.reader import ReaderAgent, ReaderResult # noqa: E402
from polymath.config import settings # noqa: E402
from polymath.tools.page_fetch import page_fetch # noqa: E402
from polymath.tools.web_search import web_search # noqa: E402
log = structlog.get_logger("week2")
OUTPUT_DIR = Path(__file__).resolve().parents[1] / "outputs"
async def gather_claims(
topic: str, n_pages: int = 3, max_results: int = 6
) -> list[ReaderResult]:
"""Search the topic, fetch the top pages, run the Reader on each.
Returns one ReaderResult per successfully-fetched page (so callers can
inspect per-page attempts/validated for metrics).
"""
results = await web_search(topic, max_results=max_results)
reader = ReaderAgent()
out: list[ReaderResult] = []
fetched = 0
for r in results:
if fetched >= n_pages:
break
url = r.get("url")
if not url:
continue
text = await page_fetch(url)
if not text:
continue
fetched += 1
out.append(await reader.extract(text=text, source_url=url))
return out
def _slugify(text: str) -> str:
slug = re.sub(r"[^a-z0-9]+", "-", text.lower()).strip("-")
return slug[:60] or "topic"
def _configure_logging() -> None:
level = getattr(logging, settings.log_level.upper(), logging.INFO)
structlog.configure(
wrapper_class=structlog.make_filtering_bound_logger(level),
processors=[
structlog.processors.add_log_level,
structlog.processors.TimeStamper(fmt="iso"),
structlog.processors.JSONRenderer(),
],
)
def main() -> None:
parser = argparse.ArgumentParser(description="Polymath Week 2 Reader demo.")
parser.add_argument("topic")
parser.add_argument("--pages", type=int, default=3, help="Pages to read (default 3).")
args = parser.parse_args()
_configure_logging()
results = asyncio.run(gather_claims(args.topic, n_pages=args.pages))
all_claims = [c for res in results for c in res.claims]
OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
ts = datetime.now(timezone.utc).strftime("%Y%m%d-%H%M%S")
out_path = OUTPUT_DIR / f"week2-{_slugify(args.topic)}-{ts}.json"
out_path.write_text(
json.dumps(
{
"topic": args.topic,
"generated_utc": ts,
"claims": [c.model_dump() for c in all_claims],
},
indent=2,
ensure_ascii=False,
),
encoding="utf-8",
)
n_pages = len(results)
n_valid_pages = sum(1 for r in results if r.validated)
print(f"\nPages read: {n_pages} | pages validated: {n_valid_pages} | claims: {len(all_claims)}")
print(f"Claims written to: {out_path}")
if __name__ == "__main__":
main()
|