Yggdradil Update: HMAS Ontology, refactor, updated dependencies & build tools and various smaller changes

[KaiTries]

Main changes:

• Versioning: updated most dependencies to more recent versions.
  • Gradle: version 8.9 is now used
  • WoT-TD-Java library: v1.2 -> newest MainSnapshot
  • Cartago library: dedicated fork but still local jar dependency -> Updated versions + additional functionality to remove workspaces.
  • currently used versions of different dependencies can be found here.
• Support for multiple Ontologies & HMAS implementation
  • Yggdrasil now runs agnostic of the underlying ontology, which enables easy switching between and adding of different ontologies. To enable a different ontology all you have to do is create a new Implementation of the RepresentationFactory interface and add it to the RepresentationFactoryFactory.
  • To choose a specific ontology a new key value pair can be added to the conf file in the environment-config part. The value corresponds to the chosen identifier in the RepresentationFactoryFactory. (This works even if the environment-config is disabled):

    {
    "environment-config" : 
    {
    "ontology": "td"
    }
    }

  • If you want to add a HypermediaArtifact on a new ontology all you have to do is implement the HypermediaArtifact interface.
• Improved agent handling. <- This should also solve some points mentioned by matteo
  • No longer constrained to specific agent WebID.
  • New header "X-Agent-LocalName" : "wantedName" <- If the name is available it will be used for the bodyArtifact otherwise a random UUID will be used.
  • the /agents/ path is no longer used. Agent body artifact are now also located in the /artifacts/ path.
• New Artifact functionality & API
  • The output of an action must now be specified in the same manner that the input must be specified!
    • The api for registering actions has therefore changed
    • Also due to the change of the HypermediaArtifact registry no longer being a singleton, the getApiKey() method has also been moved to the artifact
  • Multiple feedback parameters are now supported.
  • Signals are now also supported.
  • Currently we are still restricted to only supporting list input & output.
• Refactoring of HypermediaArtifactRegistry
  • Is no longer a Singleton and is now only used in the CartagoVerticle through dependency injection. This enables stronger decoupling and helped clean up the code a little.
• Various changes:
  • Deletion of all entities works correctly now. Workspaces can also be properly deleted in the cartago instance.
  • Namespace handling is improved. Representations will only use those that they actually need & new representations are added at runtime as well.
  • Using @base to give a bit of a cleaner look. The @base is the baseUri for the yggdrasil instance.
  • Making the representation for an entity is now decoupled from the CartagoVerticle. This enables uniformity between the different kinds of artifacts.
  • Probably various other smaller changes

[KaiTries]

TO-DOs before merge:

• Conditional exposure of WebSub Signifiers -> done
• Fixing bug that alters representation of workspace given by file and enabled environment
  • CartagoVerticle and RdfVerticle interpret the config file in a different way need to make uniform -> done
  • Metadata add triple to TD to wrap TD in ResourceProfile -> done

[KaiTries]

TD ontology wraps Thing Descriptions in ResourceProfiles to more closely represent the resources.

• No breaking changes, overall logic stays the same.
• Each workspace, artifact, agent body artifact, are now generated with two additional triples:
  • \ rdf:Type \
  • \ hmas:isProfileOf \<resourceIRI + #(workspace / artifact)>

[danaivach]

@KaiTries, I observe the following issue:

Agent bodies are initialized as CArtAgO artifacts, but focusing on them leads to a NullPointerException and a 500 server error.

Probable Reason In CartagoVerticle.focus(), the body artifact name is passed as artifactName (e.g. artifactName = agent0, given that bodies are named based on the X-Agent-LocalName header value).

However, in Workspace.makeAgentBodyArtifact() (cartago 3.2), body artifacts are initialized with names in the form "body_" + body.getAgentId().getAgentName() (e.g, body_https://wiser-solid-xi.interactions.ics.unisg.ch/agent0/profile/card#me, given that agents are named based on the X-Agent-WebID).

Therefore, the body artifact cannot be found in the workspace due to name mismatch (probably line 349).

To reproduce

• Join a workspace

  curl --location --request POST 'http://172.27.52.55:8080/workspaces/61/join' \
  --header 'X-Agent-WebID: https://wiser-solid-xi.interactions.ics.unisg.ch/agent-a/profile/card#me' \
  --header 'X-Agent-LocalName: agent0'

• Focus on the body of the joined agent

  curl --location 'http://172.27.52.55:8080/workspaces/61/focus' \
  --header 'X-Agent-WebID: https://wiser-solid-xi.interactions.ics.unisg.ch/agent-n/profile/card#me' \
  --header 'Content-Type: application/json' \
  --data '{
  "artifactName" : "agent0",
  "callbackIri" : "http://localhost:8081/notifications/"
  }'

[KaiTries]

Thanks for the detailed report @danaivach. This issue went under the radar because we only tested for focusing on workspaces and artifacts.

• Added additional test to ensure that an agent body artifact can also be focused on.
• To integrate more easily with cartago, the agent name will automatically be prepended with "body_" so that it works internally. This just means that the locations of agent bodies in yggdrasil will uniformely look like this:
  • https://yggdrasil-node.com/workspace/artifacts/body_agentname/
  • When joining through an http request you do not need to add the "body_" part just normally add the webID and prefered agent name.