Archives

MANNA ships archive notes for these archives (src/manna/archives/), listed here in priority order (the order list_archives and the workflow tools’ archive selection use):

short_name	Archive	Waveband
datalab	NOIRLab Astro Data Lab	optical
alma	ALMA Science Archive	millimeter
nrao	NRAO Science Data Archive — paused	radio
eso	ESO Science Archive	optical
cadc	Canadian Astronomy Data Centre	multi
gaia	ESA Gaia Archive	optical
gaia_ari	Gaia ARI Heidelberg	optical
sdss	Sloan Digital Sky Survey	optical

list_archives returns the active ones with their endpoints, usage notes, and notable tables; describe_table returns per-table facts. Each archive is one file, so adding one is dropping a file in and deleting one is removing it: Adding an archive is the step-by-step procedure and Modular archives — archive notes, one file per archive is the design behind it.

The active set

Which archives make curated claims is decided at startup:

• Unset MANNA_ARCHIVES (the default): every archive file present, except paused ones.

• MANNA_ARCHIVES=datalab,alma: only those two. An unknown name is logged as a warning and ignored rather than crashing the server, so a typo silently drops that archive instead of stopping startup — check the server log if an archive you named is missing.

• Paused archives (Archive.paused = "<reason>") ship in the package but stay out of the default set. Naming one in MANNA_ARCHIVES activates it: MANNA_ARCHIVES=datalab,alma,nrao.

nrao has been paused since 2026-09-11 at NRAO’s request while its TAP service is rebuilt. When active, its obscore data reads need mode="auto" or mode="async"; the archive notes say so.

priority (ascending) orders archives in list_archives and in the workflow tools’ archive selection.

Absence is not inaccessibility

Dropping or deselecting an archive removes only MANNA’s claims about it: its usage notes, schema facts, endpoint examples, and the archive label on envelopes. The archive is still reachable: search_ivoa_registry finds it, describe_ivoa_service introspects it, and run_adql_query queries it. The active set narrows what the model is told, not where it may go. (Where it may go is the job of MANNA_ALLOWED_HOSTS; see Security.)

Forking for a deployment

• Physical — delete unwanted src/manna/archives/<short_name>.py files from your fork. Discovery picks up whatever remains.

• Runtime — keep the shared image and set MANNA_ARCHIVES per deployment.