AIREloom: Asynchronous Python client for the OpenAIRE API

Samuel Mok · s.mok@utwente.nl · 2025–2026

AIREloom is an async Python client for the OpenAIRE Graph API and Scholexplorer API, built on bibliofabric.

Docs: utsmok.github.io/AIREloom · PyPI: aireloom · License: MIT

Features

• Full API coverage — Research Products (v2), Projects, Organizations, Data Sources, Persons, Research Product Links (v1), Scholexplorer (v3)
• Async by design — built on httpx + asyncio with proper connection pooling
• Typed throughout — Pydantic models for all inputs/outputs, PEP 561 py.typed marker
• Ergonomics layer — computed properties, SafeStr/SafeList defaults, convenience queries, iterator helpers
• Flexible auth — auto-detection from env vars, static tokens, OAuth2 client credentials, or no auth
• Resilient — retries with backoff, rate-limit handling (Retry-After), optional client-side caching, hook system

Installation

uv add aireloom

Or with pip: pip install aireloom. Requires Python ≥3.12.

Quick Start

import asyncio
from aireloom import AireloomSession
from aireloom.endpoints import ResearchProductsFilters

async def main():
    async with AireloomSession() as session:
        # Search publications
        filters = ResearchProductsFilters(
            type="article", mainTitle="climate change",
            fromPublicationDate="2024-01-01",
        )
        response = await session.research_products.search(
            filters=filters, page_size=5, sortBy="publicationDate desc",
        )
        for product in response.results:
            print(f"{product.title} — DOI: {product.doi or 'N/A'}")

        # Iterate all results (cursor-based auto-pagination)
        filters2 = ResearchProductsFilters(type="dataset", countryCode="NL")
        async for product in session.research_products.iterate(
            filters=filters2, page_size=50,
        ):
            print(product.title)
            break  # stop when you want

        # Get a single entity
        product = await session.research_products.get("doi:10.1038/s41586-021-03964-9")
        print(product.title, product.doi)

        # Convenience queries
        citations = await session.queries.citing_works("doi:10.1038/s41586-021-03964-9")
        print(f"{len(citations)} citations")

asyncio.run(main())

No authentication required — the OpenAIRE API works without it. For higher rate limits, see Authentication.

Core API

Retrieval methods

Every resource client (session.research_products, session.organizations, etc.) provides:

Resource clients

Ergonomics

Computed properties on models — derived fields that don't exist in the raw API response:

product.doi              # → "10.1234/example" (extracted from pids)
product.publication_year  # → 2024 (from publicationDate)
product.is_open_access   # → True (from bestAccessRight)
org.ror_id               # → "https://ror.org/..." (from rorId)
project.funder_name      # → "EC" (from funding array)

SafeStr/SafeList — None is never returned for string/list fields; you get "" or [] instead:

product.subjects  # [] instead of None
product.title     # "" instead of None

Convenience queries via session.queries:

session.queries.publications_by_doi("10.1234/example")
session.queries.citing_works("doi:10.1234/example")
session.queries.datasets_by_organization("openaire____::orgID:grid.5522.e")
session.queries.projects_by_funder("EC")

See Ergonomics docs for the full list.

Filters & Sorting

Filters are Pydantic models — import from aireloom.endpoints:

from aireloom.endpoints import ResearchProductsFilters, ProjectsFilters

filters = ResearchProductsFilters(
    type="article",
    mainTitle="machine learning",
    fromPublicationDate="2024-01-01",
    countryCode="NL",
)

Sort with sortBy="field asc" or sortBy="field desc". Valid fields depend on the endpoint (see ENDPOINT_DEFINITIONS).

Authentication

AIREloom auto-detects auth from environment variables or .env files (prefixed with AIRELOOM_). No auth is the default if nothing is configured.

# Option 1: Static token
AIRELOOM_OPENAIRE_API_TOKEN=your_token

# Option 2: OAuth2 client credentials
AIRELOOM_OPENAIRE_CLIENT_ID=your_id
AIRELOOM_OPENAIRE_CLIENT_SECRET=your_secret

Or pass explicitly:

from bibliofabric.auth import NoAuth, StaticTokenAuth, ClientCredentialsAuth

session = AireloomSession(auth_strategy=StaticTokenAuth(token="..."))

See Authentication docs for details.

Error Handling

from bibliofabric.exceptions import (
    BibliofabricError, APIError, NotFoundError, RateLimitError,
    TimeoutError, NetworkError, AuthError, ValidationError,
)

See Error Handling docs for the full hierarchy.

Examples

All examples in examples/ are dual-purpose — run as scripts or as interactive marimo notebooks:

# As a script
uv run examples/simple_example.py

# As an interactive notebook
uv run marimo edit examples/simple_example.py

See examples/README.md for marimo embedding and WASM details.

Known OpenAIRE API Issues

Full bug report with reproduction steps: OPENAIRE_BUG_REPORT.md.

• Persons givenName/lastName filters — cause HTTP 500 despite being in the spec
• pageSize=100 on Links/Scholix — silently falls back to 10. Use pageSize ≤ 99
• Undocumented endpoints — Persons has no filtering docs; Links (/v1/researchProducts/links) is entirely undocumented
• Sort field issues — Persons only accepts relevance; Data Sources error messages say "organizations"

Development

uv sync                  # install dev deps
uv run pytest -x -q      # run tests
uvx ruff check src/      # lint
uvx ty check src/        # type check

Pre-commit hooks for ruff format + lint are configured in .pre-commit-config.yaml.

Contributions welcome — see Contributing.