Storage Operations | IPFS

Why composites exist: "upload a file and make sure it's durably pinned" or "confirm this content is still intact" are common workflow needs that would otherwise take two or three chained nodes (content/add + pin/add or cluster/add, or content/cat + a manual hash check). The storage resource packages each pattern as a single node call.

uploadAndPin

Binary input. Adds content, then pins it — via the Cluster (replicated) when a Cluster connection is configured, otherwise via Kubo's own local pinset.

Field	Type	Required	Default	Description
`inputDataKey`	text	No	`data`	Key in the upstream node's InputData holding the bytes to store.
`dataMode`	select	No	`base64`	`base64` \| `text`.
`inlineContent`	text	No	—	Inline content fallback.
`fileName`	text	No	`file`	File name recorded on the multipart part.
`baseUrl`	text	Conditional	—	Required only when not using a Cluster connection (the Kubo-only path).
`clusterBaseUrl`	text	No	—	When set, content is added and pinned through the Cluster instead of Kubo directly.
`clusterAuthMode` / `destinationCredentialID`	—	Conditional	—	Required together when pinning via the Cluster with Basic auth.

Response: { cid, name, size, usedCluster, pinStatus }.

Step-by-step:

Resolve the upstream binary content (same inputDataKey/inlineContent pattern as content/add).
If clusterBaseUrl is set: call the Cluster's add-and-replicate-pin endpoint (equivalent to cluster/add).
Otherwise: call Kubo's content/add with pin: true directly.
Return a unified record regardless of which path was taken.

verifyIntegrity

Kubo-only (no Cluster connection needed). Re-fetches content at a CID and asks Kubo to recompute its CID, then compares the two — a tamper/corruption check without reimplementing hashing client-side.

Field	Type	Required	Default	Description
`cid`	text	✓ Yes	—	The expected CID — also the path fetched to verify.
`chunker`	text	No	—	Chunker to use when recomputing — should match the original add for a meaningful comparison.
`cidVersion`	int	No	inferred	CID version to recompute with. When unset, inferred from the expected CID's own encoding (`0` for a `Qm…`-prefixed v0 CID, otherwise `1`) so the comparison is apples-to-apples.

Missing cid → CFG_MISSING_CID. Response: { matches, expectedCid, computedCid, size }.

Step-by-step:

Fetch the content at cid via Kubo's cat endpoint.
Re-add the fetched bytes with Kubo's only-hash flag (recompute the CID, no datastore write).
Compare the recomputed CID against the expected cid string-for-string.
matches: false is a legitimate result, not an execution error — it still routes to the success output port.

The CIDv1 inference fix: an earlier version of this operation always recomputed using CIDv0, which produced a false mismatch for any content that was originally added as CIDv1. The fix infers the version from the expected CID's own prefix by default — pass cidVersion explicitly only if you need to override that inference.