CORE_SOURCE_GUIDE.md

CORE Source Guide

Use this guide when wiring CORE into scholar-rag-agent. The agent can route enrichment through GPT-5.5 / Claude Sonnet 4.6 / Gemini 2.5 / Kimi K2 when enabled, but the CORE connector itself is deterministic JSON — no LLM required to list matching open-access works.

Why CORE

CORE aggregates open-access research outputs from repositories and journals worldwide. Alongside OpenAIRE, Zenodo, and Figshare it covers OA copies that may never appear as the primary record in PubMed or Crossref.

Public keyword search:

GET https://api.core.ac.uk/v3/search/works?q=retrieval+augmented+generation&limit=5

limit is capped at 100. The response is a JSON object with a results array of work objects. An optional CORE API key may be passed as a Bearer token for higher rate limits.

What you get

Field	Source
title	title
text	Collapsed abstract, or a By authors (year) descriptor when absent
source	links entry with type=display, else downloadUrl, else https://doi.org/{doi}, else title
metadata.doi	doi
metadata.year	yearPublished as an integer, or the leading four digits of a date-shaped string
metadata.authors	Comma-joined authors[].name
metadata.source_type	"core"

Example

import asyncio

from ingestion.core import CoreConnector

documents = asyncio.run(CoreConnector().search("climate adaptation", max_results=5))
for document in documents:
    print(document.metadata["doi"], document.title)

Safety notes

• Blank queries and non-positive max_results short-circuit with no HTTP call.
• Works without a title are skipped rather than raising.
• Search metadata works without a key; configure a CORE API key only when you need higher quotas or authenticated full-text fields.