Image sources

Where quarry acquires public-domain and CC0 images: Wikimedia Commons, plus the PD12M and Megalith metadata on Hugging Face for the /gallery Sources view.

Used by

Package	What for	Reads
argus-quarry (opens in new tab)	`--source commons` (the default). Wikimedia asks for a contact User-Agent.	`--source commons`, `COMMONS_USER_AGENT`, `QUARRY_USER_AGENT`
argus-quarry (opens in new tab)	PD12M and Megalith metadata, plus optional CLIP text search.	`QUARRY_PD12M_DIR`, `QUARRY_MEGALITH_DIR`, `QUARRY_ALLOW_FETCH`, `HF_TOKEN`, `[search]`, `QUARRY_CLIP_MODEL`

Official guides

Minimal setup

Wikimedia Commons needs no account. Set COMMONS_USER_AGENT (or QUARRY_USER_AGENT) to something that identifies your deployment and gives a contact, as Wikimedia's API policy asks.
PD12M and Megalith are metadata-only datasets on Hugging Face. Fetch them with quarry's scripts/fetch_public_domain_metadata.py (--hf-token or HF_TOKEN) into QUARRY_PD12M_DIR and QUARRY_MEGALITH_DIR. The server's fetch button stays off unless QUARRY_ALLOW_FETCH=1.
Text search over PD12M needs the [search] extra and a CLIP model (QUARRY_CLIP_MODEL, an OpenAI CLIP ViT-L/14 from the HF cache by default).

Every acquired image keeps its source, author and licence. See public-domain archives (opens in new tab).