Research Paper RAG - arXiv, PubMed, Open Access

🤖 What does the research paper RAG do?

A research paper RAG reads open research papers from arXiv, Europe PMC and OpenAlex. It can collect:

• open research papers from arXiv, Europe PMC and OpenAlex
• the title, source, license and text chunks
• A saved baseline when the source compares the next run
• JSON, CSV or Excel from the dataset
• The same run through the Apify API or MCP

💡 Why use the research paper RAG?

People use a research paper RAG when the same public page would otherwise be checked by hand.

• Watch a change. The Actor keeps a baseline and writes the next difference.
• Feed a sheet. The dataset opens as JSON, CSV or Excel.
• Call it from code. The Apify API runs the same input.
• Hand it to an agent. Apify MCP can call the Actor with that input.
• Stay on the sample. The prefilled input is the short, cheap run.

🚀 How do you run the research paper RAG?

1. Open this Actor and click Try for free.
2. Keep the prefilled sample, or set the prefilled fields.
3. Leave the other fields on their defaults unless you need a narrower run.
4. Click Run.
5. Open the Dataset tab and download JSON, CSV or Excel.

Sample input from the prefilled schema:

{
  "papers": [
    "PMC4792175",
    "arXiv:2405.13832"
  ],
  "sources": [
    "europepmc",
    "openalex",
    "arxiv"
  ],
  "maxPapers": 5,
  "watchNewPapers": false,
  "mode": "both",
  "chunkSize": 512,
  "chunkOverlap": 64,
  "includeMarkdown": true,
  "includeReferences": false,
  "maxChunksPerPaper": 400,
  "licenseFilter": "any-open"
}

⚠️ What limits does this Actor have?

The input schema is the limit. Defaults and prefills are unchanged by this copy update.

• papers: at most 25 items.
• maxPapers: maximum 25, minimum 1.
• chunkSize: maximum 4096, minimum 64.
• chunkOverlap: maximum 2048, minimum 0.
• maxChunksPerPaper: maximum 2000, minimum 1.
• A robots.txt disallow is not treated as a successful priced read when the source checks robots.txt.
• Actor Start is priced in the table. The other events are billed only when the source charges them and the charge is accepted.
• A spending limit that cannot pay for the next event stops that event. The source does not invent a result to replace it.

💳 How much does the research paper RAG cost?

The table is the live pay-per-event price list (pricingInfos) for this Actor. It is not a guessed price.

Event	What the price list calls it	Free	Bronze	Silver	Gold	Platinum	Diamond
apify-actor-start	Actor Start	$0.0005	$0.0005	$0.0005	$0.0005	$0.0005	$0.0005
research-paper	Research paper	$0.012	$0.0108	$0.0096	$0.0084	$0.0084	$0.0084

One Free-plan run that bills one research-paper plus Actor Start is $0.0005 + $0.012 = $0.0125. 1,000 research-paper events on the Free plan are $12.00, before Actor Start. These figures use only the live prices in the table.

apify-actor-start is in the live price list. Charged when the Actor starts running. Number of events charged depends on Actor memory (one event per GB, minimum one event). The source names research-paper and charges it with Actor.charge when that event happens. One open-access research paper whose full text was read and returned. Papers that are not open access, use a licence that does not allow redistribution, have no full text in these sources, or fail to download are not charged. Chunks are included and do not add a charge.

📝 What input does this Actor take?

Required fields are marked. Defaults and prefills are the quick run; this text does not change them.

Field	Required	What it means
papers (Papers)	No	DOIs, PMIDs, PMCIDs, arXiv ids, OpenAlex work ids, or links to those records. One per line, up to...
query (Search query)	No	Optional, unless New papers this week is on. Finds open-access papers with full text. Europe PMC ...
sources (Sources)	No	Which official sources to ask. Europe PMC serves PubMed open-access full text. arXiv serves HTML ...
maxPapers (Maximum papers)	No	Stop after this many papers, identifiers and search hits combined. A low number caps the cost of ...
datasetId (Read identifiers from a dataset)	No	Optional. An Apify dataset whose items contain doi, pmcid, pmid</...
watchNewPapers (New papers this week)	No	Keyword watch. With a search query, Europe PMC and arXiv return papers first published in the las...
mode (What to return)	No	Paper and chunks writes one paper row plus its chunk rows. Chunks only writes chunk...
chunkSize (Chunk size (tokens))	No	Token budget per chunk, counted with cl100k_base, the tokenizer OpenAI embedding and chat models ...
chunkOverlap (Chunk overlap (tokens))	No	How many tokens from the end of one chunk are repeated at the start of the next. At most half the...
includeMarkdown (Include Markdown)	No	Add the full markdown field on the paper row. Turn it off when you only want the chu...
includeReferences (Include the reference list)	No	Off by default. Reference lists are long and rarely useful in a retrieval index. Turn this on whe...
maxChunksPerPaper (Maximum chunks per paper)	No	Safety cap on chunk rows for one paper. The paper row is marked truncated when the cap is hit. Ch...
licenseFilter (Which licences to read)	No	Any open licence returns CC BY, CC BY-SA, CC BY-NC, CC BY-ND, the NC-ND combinations, CC0 ...
openAlexApiKey (OpenAlex API key (optional))	No	Your own free OpenAlex key. Needed only for OpenAlex keyword search and for OpenAlex's cached PDF...
contactEmail (Email for Unpaywall)	No	Your own email, sent only to the Unpaywall API when a DOI has no redistributable full text in Eur...

📊 What output does this Actor return?

The dataset schema names every stored field.

Field	What the schema says
recordType	Row kind: document (one per paper) or chunk (one per token-counted ch...
url	Stable link for the paper, used as the join key between the paper row...
finalUrl	The full-text URL that was actually fetched, after redirects.
status	Document rows: processed (full text returned and charged), skipped (n...
failureReason	Document rows: why a paper was skipped or failed. Null when it was pr...
message	Document rows: a sentence explaining a skip or a failure.
title	Paper title.
author	Authors as one string, for a spreadsheet column. Chunk rows repeat it...
authors	Document rows: author names, up to 40. Emails are not included.
authorCount	Document rows: how many authors the source listed, which can be more ...
year	Publication year.
venue	Journal or repository name.
doi	DOI, without the https://doi.org/ prefix.
pmid	PubMed id.
pmcid	PubMed Central id.
arxivId	arXiv id without a version suffix.
openAlexId	OpenAlex work id.
license	Licence name, for example CC BY 4.0. Null when no licence was found.
licenseUrl	Link to the licence.
commercialUseAllowed	False for NC licences. Null when no licence was found.
derivativesAllowed	False for ND licences. Chunks of an ND paper are still returned, with...
source	Where the full text was read. unpaywall is a licensed open-access PDF...
fullTextFormat	jats is Europe PMC XML, arxiv-html is the official arXiv HTML, pdf is...
fullTextUrl	URL of the full text that was read.
oaPdfUrl	Open-access PDF link reported by OpenAlex, whether or not this run do...
abstract	Abstract. For an arXiv paper whose licence does not allow redistribut...
citation	A ready citation: authors, year, title, DOI and licence. Repeated on ...
markdown	Document rows: the paper as Markdown with headings. Null when Include...
charCount	Characters in the Markdown (document) or the chunk text (chunk).
wordCount	Document rows: words in the text.
tokenCount	Tokens counted with cl100k_base. Document rows count the Markdown. Ch...
chunkCount	Document rows: how many chunk rows were written.
sectionCount	Document rows: sections in the full text. Null for a PDF.
pageCount	Document rows: pages in a PDF. Null for XML and HTML.
pageStart	First page this row came from. Null when the source has no pages (XML...
pageEnd	Last page this row came from. Null when the source has no pages.
truncated	Document rows: true when the Markdown, the PDF page cap or the chunk ...
durationMs	Document rows: how long resolving this paper took.
fetchedAt	Document rows: when the row was written, ISO 8601 UTC.
documentTitle	Chunk rows: the paper title, so a chunk can be cited without a join.
chunkIndex	Chunk rows: position in the paper, starting at 0.
text	Chunk rows: the chunk text. This is what an embedding model should re...
headings	Chunk rows: the heading the chunk falls under, and any headings insid...
isPartial	Chunk rows: true when a paragraph was split to stay inside the token ...

One record from assets/sample-run-data.json shipped with this Actor, shortened where a string is long.

{
  "chunkIndex": 0,
  "tokenCount": 431,
  "charCount": 2407,
  "pageStart": null,
  "pageEnd": null,
  "headings": [
    "Abstract",
    "Comment",
    "Supporting discovery through good data management"
  ],
  "isPartial": false,
  "license": "CC BY 4.0",
  "doi": "10.1038/sdata.2016.18",
  "text": "## Abstract There is an urgent need to improve the infrastructure supporting the reuse of scholarly data. A diverse set of stakeholders—representing academia, industry, funding ag…"
}

🔌 How do you call this Actor from the API?

Call mmdincer/research-paper-rag with the Apify API. The body is the input schema. MCP clients on Apify use the same input. Make and Zapier can start the Actor through Apify's integration. The input is the same JSON.

from apify_client import ApifyClient
client = ApifyClient("<APIFY_TOKEN>")
run = client.actor("mmdincer/research-paper-rag").call(run_input={"papers": ["PMC4792175", "arXiv:2405.13832"], "sources": ["europepmc", "openalex", "arxiv"], "maxPapers": 5, "watchNewPapers": false, "mode": "both", "chunkSize": 512, "chunkOverlap": 64, "includeMarkdown": true, "includeReferences": false, "maxChunksPerPaper": 400, "licenseFilter": "any-open"})
items = client.dataset(run["defaultDatasetId"]).list_items().items

import { ApifyClient } from "apify-client";
const client = new ApifyClient({ token: "<APIFY_TOKEN>" });
const run = await client.actor("mmdincer/research-paper-rag").call({"papers": ["PMC4792175", "arXiv:2405.13832"], "sources": ["europepmc", "openalex", "arxiv"], "maxPapers": 5, "watchNewPapers": false, "mode": "both", "chunkSize": 512, "chunkOverlap": 64, "includeMarkdown": true, "includeReferences": false, "maxChunksPerPaper": 400, "licenseFilter": "any-open"});
const { items } = await client.dataset(run.defaultDatasetId).listItems();

❓ What else should you know?

Does the research paper RAG need a login?

The sample run uses the public source and the prefilled input. A field that asks for your own key is optional unless the input schema marks it required.

What does the research paper RAG charge for?

The primary event in the live price list is research-paper (Research paper). Actor Start is separate. An event the source does not call is not billed by the source.

Can this Actor run on a schedule?

Yes. Save the input and schedule it. The next run compares against the baseline when the source stores one.

Where do the results go?

Results land in the run dataset. Download JSON, CSV or Excel, or read them with the API.

⚖️ Is it legal to use the research paper RAG?

You are responsible for the source terms and for privacy law that applies to you, including the GDPR if you handle personal data. This Actor stores the public fields described above. The source checks robots.txt before a fetch when that check is in the code. This Actor is not affiliated with arXiv, Europe PMC and OpenAlex.

Not affiliated with arXiv, Europe PMC and OpenAlex. Source license: Apache-2.0. Where the source checks robots.txt, a disallowed URL is not fetched as a normal result.

🔗 Which related actors sit beside this one?

Other public Actors from mmdincer in this group:

• EPO Patent Monitor - New publications for queries
• ERIC Monitor - Education records for queries
• INSPIRE Monitor - HEP papers for literature queries
• Company Registry Lookup - Filings for seven registers
• SAM.gov Monitor - Contract notices for the US

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.

Run the research paper RAG again when you want a fresh arXiv check.