We do not extract. We synchronize.

Federation is a peer relationship, not a harvest.

Five institutional archives already hold the Indic manuscript record. The Bodleian. The Cambridge Digital Library. The University of Pennsylvania’s OPenn. The Internet Archive. Wikidata. Vamshanidhi does not duplicate their work; it synchronizes with them, preserves their provenance, and emits its own cite-able URNs that they can synchronize against in return.

This is the difference between a scraper and a peer. A scraper takes; a peer trades. Outbound IIIF v3 manifests are first-class. Outbound OAI-PMH responses are first-class. Wikidata identifiers are cross-walked both ways. A scholar at Oxford who cites a Vamshanidhi record is citing a URL that resolves stably for a century, and a peer institution that mirrors that record is mirroring a citation, not a copy.

Every manuscript we hold in trust carries its provenance back to its source. Every record we publish emits a citation other scholars can trust.

How federation is engineered.

Three protocols carry the weight. OAI-PMH for bibliographic metadata harvesting and emission. IIIF v3 for deep-zoom image manifests and viewer interoperability. Wikidata SPARQL for identifier cross-walks. Every record ingested through any of the three writes a provenance row first: source URL, capture timestamp, SHA-256 hash, integrity status. The claim it implies does not surface until that row is intact.

The identifier scheme, codified in ADR-0026, gives every primary record a URN of the form urn:vamshanidhi:<type>:<ULID>. A peer reading the string knows what kind of record it points to without a side channel. A peer mirroring a record can dereference the URN to a stable manifest. A scholar citing it gets a URL that survives platform redesigns.

Bulk content mirrors follow the preservation-mirror policy in ADR-0041. Mirror tier, retention, and refetch cadence are documented per peer; takedown requests from a source institution propagate within the same cycle.

The peers, named.

The live registry is temporarily unavailable; the curated list below reflects the soft-launch composition.

University of Pennsylvania · OPenn
Open-data, BagIt-served Indic manuscripts. Ninety-one records mirrored.
Protocol: BagIt + custom JSON
Cambridge Digital Library
Sanskrit collection. Five hundred IIIF v3 manifests synchronized.
Protocol: IIIF v3 + OAI-PMH
Bodleian Libraries · University of Oxford
IIIF-served Indic holdings. Twenty-five records in initial mirror.
Protocol: IIIF v3
Internet Archive
Public-domain Indic texts. Custom-JSON adapter, refined ingest.
Protocol: Custom JSON adapter
Wikidata
Cross-walked identifier graph. Bidirectional sync.
Protocol: Wikidata SPARQL

Why a mirror is not a verdict.

Federation produces provenance, not verification. A record ingested from a peer arrives with its source intact but no scholar has yet weighed its claims in this platform’s register. That work belongs to the Scholar Council. Until a named scholar concurs, a federated record sits at provisional, exactly as honestly as it would on the source institution’s own surface.

Some manuscripts held in trust carry a scholar’s concurrence; the rest are accessible and citable but do not yet wear the verified pill. Until the Scholar Council is seated, even a scholar-reviewed reading is shown provisionally. The platform refuses to round up.

Inbound and outbound, both first-class.

Outbound endpoints are not an afterthought. A peer institution can harvest from Vamshanidhi’s OAI-PMH endpoint, dereference any URN as a IIIF v3 manifest, and cross-walk identifiers via Wikidata. The contracts are versioned. Breaking changes go through the same scholar countersignature protocol that binds the rest of the platform.

The developer surface lives at /developer. Institutional peers interested in onboarding should write to the custodial entity through /contribute. Synchronization terms are negotiated per peer and published with the agreement.

We do not extract. We synchronize.

How federation is engineered.

The peers, named.

Why a mirror is not a verdict.

Inbound and outbound, both first-class.

Further reading.