Mïmis is a variety of pages all aiming generally toward a collaborative file system stored in IPFS and indexed in Neo4j.
Unfortunately, the vision hasn't quite congealed yet, after, literally almost a decade of contemplation. This iteration is, as of 2025⁄3⁄8, about 4–6 weeks old. The most promising avenue to date is the merge lists which allow users to create ordered lists of content that can be aggregated to form recommendations of undiscovered media.
Unfortunately, the architecture currently requires API access to a Kubo IPFS node. Kubo can be relatively easily installed using IPFS Desktop. After it's installed, you can edit the configuration via the GUI or using the CLI tools:
KEY="API.HTTPHeaders.Access-Control-Allow-Origin"
ipfs config --json $KEY $(
ipfs config $KEY \
| jq '. += ["http://localhost:5173"]' \
| tr -d '[:space:]'
)This uses jq to insert the address of the development server into the CORS configuration of IPFS. Once this is done, (and the IPFS daemon is restarted), it should be possible to run the development server using:
deno install && deno task devTheoretically, this same method would work for a remote web server, but accessing a private network from a public one is governed by the Private Network Access (PNA) specification. It dictates that the request should contain a "Access-Control-Request-Private-Network: true" header, and the response a "Access-Control-Allow-Private-Network: true" header.
It is possible to add the header to the IPFS response headers using:
ipfs config --json API.HTTPHeaders.Access-Control-Allow-Private-Network '["true"]'However, that header is only included in the response to a POST request, not the preflight OPTIONS request, which causes the connection to fail.
Perhaps the easiest solution aside from patching Kubo is to use a reverse proxy, and add the appropriate headers as the response is en route.
Additionally, the IPFS node should be protected from other hosts attempting to make contact by setting a password & restricting available paths using API.Authorizations.
The alternative is to implement interfacing with an external storage provider like Storacha, Pinata, or Filebase; and then either giving users access to shared storage or have them buy their own.
I want to download random blocks that comprise files & that is done (without API access) via query parameters to a "path-gateway".
Potential database organizations are discussed elsewhere.
Likewise, the network architecture has its own page.