Update src/api/agent.py
Browse files- src/api/agent.py +50 -15
src/api/agent.py
CHANGED
|
@@ -1399,31 +1399,63 @@ async def _semantic_scholar_search(topic: str, limit: int = 8) -> List[dict]:
|
|
| 1399 |
|
| 1400 |
async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
|
| 1401 |
"""
|
| 1402 |
-
Search OpenAlex
|
| 1403 |
-
|
| 1404 |
-
|
|
|
|
| 1405 |
"""
|
| 1406 |
try:
|
| 1407 |
-
from urllib.parse import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1408 |
params = {
|
| 1409 |
-
"
|
| 1410 |
-
|
| 1411 |
-
|
| 1412 |
-
|
| 1413 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1414 |
}
|
| 1415 |
-
url = "https://api.openalex.org/works?" +
|
| 1416 |
headers = {"User-Agent": "LibBee/3.2 (mailto:libse@ku.ac.ae)"}
|
| 1417 |
|
| 1418 |
async with httpx.AsyncClient(timeout=10) as client:
|
| 1419 |
r = await client.get(url, headers=headers)
|
| 1420 |
if r.status_code != 200:
|
| 1421 |
-
logger.warning(f"OpenAlex returned {r.status_code}")
|
| 1422 |
return []
|
| 1423 |
|
| 1424 |
data = r.json()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1425 |
results = []
|
| 1426 |
-
for work in
|
| 1427 |
title = work.get("title") or "Untitled"
|
| 1428 |
|
| 1429 |
# Reconstruct abstract from inverted index
|
|
@@ -1450,7 +1482,7 @@ async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
|
|
| 1450 |
authors_str = ", ".join(author_names[:3])
|
| 1451 |
if len(author_names) > 3:
|
| 1452 |
authors_str += " et al."
|
| 1453 |
-
first_last =
|
| 1454 |
|
| 1455 |
year = str(work.get("publication_year") or "n.d.")
|
| 1456 |
doi = (work.get("doi") or "").replace("https://doi.org/", "")
|
|
@@ -1460,7 +1492,10 @@ async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
|
|
| 1460 |
else (f"https://doi.org/{doi}" if doi else "")
|
| 1461 |
or work.get("id", "https://openalex.org")
|
| 1462 |
)
|
| 1463 |
-
venue = (
|
|
|
|
|
|
|
|
|
|
| 1464 |
|
| 1465 |
results.append({
|
| 1466 |
"title": title,
|
|
@@ -1474,7 +1509,7 @@ async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
|
|
| 1474 |
"_source": "OpenAlex",
|
| 1475 |
})
|
| 1476 |
|
| 1477 |
-
logger.info(f"OpenAlex: {len(results)} papers for '{
|
| 1478 |
return results
|
| 1479 |
|
| 1480 |
except Exception as e:
|
|
|
|
| 1399 |
|
| 1400 |
async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
|
| 1401 |
"""
|
| 1402 |
+
Search OpenAlex using title.search + abstract.search filters for precision.
|
| 1403 |
+
This targets only the title and abstract fields — far more relevant than the
|
| 1404 |
+
general `search` parameter which also hits author names, journals, institutions.
|
| 1405 |
+
Free, no API key, typically < 2s.
|
| 1406 |
"""
|
| 1407 |
try:
|
| 1408 |
+
from urllib.parse import urlencode
|
| 1409 |
+
|
| 1410 |
+
# Build a clean search term — strip boolean operators, keep the core phrase
|
| 1411 |
+
clean = re.sub(r'\b(AND|OR|NOT)\b|[()"\']', ' ', topic, flags=re.IGNORECASE)
|
| 1412 |
+
clean = re.sub(r'\s+', ' ', clean).strip()[:200]
|
| 1413 |
+
|
| 1414 |
+
# Use title.search OR abstract.search — papers that match in either field
|
| 1415 |
params = {
|
| 1416 |
+
"filter": (
|
| 1417 |
+
f"title.search:{clean}|abstract.search:{clean},"
|
| 1418 |
+
f"has_abstract:true,type:article,is_paratext:false"
|
| 1419 |
+
),
|
| 1420 |
+
"sort": "relevance_score:desc",
|
| 1421 |
+
"per_page": str(limit),
|
| 1422 |
+
"select": ",".join([
|
| 1423 |
+
"id", "title", "authorships", "publication_year",
|
| 1424 |
+
"primary_location", "abstract_inverted_index",
|
| 1425 |
+
"doi", "cited_by_count", "open_access",
|
| 1426 |
+
]),
|
| 1427 |
}
|
| 1428 |
+
url = "https://api.openalex.org/works?" + urlencode(params)
|
| 1429 |
headers = {"User-Agent": "LibBee/3.2 (mailto:libse@ku.ac.ae)"}
|
| 1430 |
|
| 1431 |
async with httpx.AsyncClient(timeout=10) as client:
|
| 1432 |
r = await client.get(url, headers=headers)
|
| 1433 |
if r.status_code != 200:
|
| 1434 |
+
logger.warning(f"OpenAlex returned {r.status_code} for topic '{clean[:40]}'")
|
| 1435 |
return []
|
| 1436 |
|
| 1437 |
data = r.json()
|
| 1438 |
+
works = data.get("results", [])
|
| 1439 |
+
|
| 1440 |
+
# If title+abstract search returns nothing, fall back to general search
|
| 1441 |
+
if not works:
|
| 1442 |
+
logger.info(f"OpenAlex title+abstract search returned 0 — retrying with general search")
|
| 1443 |
+
params_fallback = {
|
| 1444 |
+
"search": clean,
|
| 1445 |
+
"filter": "has_abstract:true,type:article,is_paratext:false",
|
| 1446 |
+
"sort": "relevance_score:desc",
|
| 1447 |
+
"per_page": str(limit),
|
| 1448 |
+
"select": params["select"],
|
| 1449 |
+
}
|
| 1450 |
+
r2 = await client.get(
|
| 1451 |
+
"https://api.openalex.org/works?" + urlencode(params_fallback),
|
| 1452 |
+
headers=headers,
|
| 1453 |
+
)
|
| 1454 |
+
if r2.status_code == 200:
|
| 1455 |
+
works = r2.json().get("results", [])
|
| 1456 |
+
|
| 1457 |
results = []
|
| 1458 |
+
for work in works:
|
| 1459 |
title = work.get("title") or "Untitled"
|
| 1460 |
|
| 1461 |
# Reconstruct abstract from inverted index
|
|
|
|
| 1482 |
authors_str = ", ".join(author_names[:3])
|
| 1483 |
if len(author_names) > 3:
|
| 1484 |
authors_str += " et al."
|
| 1485 |
+
first_last = author_names[0].split()[-1] if author_names else "Unknown"
|
| 1486 |
|
| 1487 |
year = str(work.get("publication_year") or "n.d.")
|
| 1488 |
doi = (work.get("doi") or "").replace("https://doi.org/", "")
|
|
|
|
| 1492 |
else (f"https://doi.org/{doi}" if doi else "")
|
| 1493 |
or work.get("id", "https://openalex.org")
|
| 1494 |
)
|
| 1495 |
+
venue = (
|
| 1496 |
+
(work.get("primary_location") or {})
|
| 1497 |
+
.get("source", {}) or {}
|
| 1498 |
+
).get("display_name") or ""
|
| 1499 |
|
| 1500 |
results.append({
|
| 1501 |
"title": title,
|
|
|
|
| 1509 |
"_source": "OpenAlex",
|
| 1510 |
})
|
| 1511 |
|
| 1512 |
+
logger.info(f"OpenAlex: {len(results)} papers for '{clean[:50]}'")
|
| 1513 |
return results
|
| 1514 |
|
| 1515 |
except Exception as e:
|