AI model library
Models by content, not by trust.
Model files are imported, verified against their SHA-256, pinned by CID in a private Kubo swarm with IPFS Cluster, and served only through a read-only gateway that refuses unapproved CIDs.
AI model library: overview
| File | CID | SHA-256 | Size | Gateway serving |
|---|---|---|---|---|
- Gateway
- CONNECTING serves GET/HEAD /ipfs/<approved CID> only
- Replication status
- UNAVAILABLE — cluster status is not exposed publicly
- Consensus
- IPFS is outside consensus; models never enter block validation
- Storage reward
- IPFS storage reward: Live on testnet IPFS reward v2: Deployed on testnet 5% of issuance pays IPFS Kubo and Cluster operators who prove the manifest content they pin. From block 68,961 the bonded, slashable v2 pool
0x16362198b0685B4a1d4C1f531B215C21568002b7takes over. Operator guide
| Component | Status | Registry ID |
|---|---|---|
| IPFS storage sidecar and replication experiments | PARTIAL | CMP-IPFS |
| Release Fabric (package manifest, TUF catalog, IPFS, CAR, update engine) | LOCAL VERTICAL SLICE (fixture + isolated component packages only) | CMP-RELEASE-FABRIC |
| AI-sharing node, gateway and provider deployment scripts | IMPLEMENTED; operator-run only | CMP-AI-SHARING-DEPLOY |