One nice feature of this solution is that there is no required structure for your JSON documents in Cosmos DB. They can be structured as if it is a non-graph application, per general NoSQL/Mongo design best practices.
This is because the graph, in the CosmosAIGraph solution, only exists in-memory within the Python rdflib process.
However, one suggestion is to use an edges attribute in your documents, and populate it with the list of outgoing relationships from the given entity/vertex document.
In the case of the CosmosAIGraph reference dataset of python libraries, however, an edges attribute is not used. Instead, the developers and dependency_ids are used at runtime to construct the in-memory RDF graph.
...
"developers": [
"contact@palletsprojects.com"
],
...
"dependency_ids": [
"pypi_asgiref",
"pypi_blinker",
"pypi_click",
"pypi_importlib_metadata",
"pypi_itsdangerous",
"pypi_jinja2",
"pypi_python_dotenv",
"pypi_werkzeug"
],
...
The impl\data directory in this repo contains a curated set of PyPi (Python) library JSON documents.
This domain of software libraries was chosen because it should be relatable to most customers, and it also suitable for Bill-of-Materials graphs.
The PyPi JSON files were obtained with HTTP requests to public URLs such as https://pypi.org/pypi/{libname}/json, and their HTML contents were transformed into JSON.
Subsequent data wrangling fetched referenced HTML documentation, produced text summarization with Azure OpenAI and semantic-kernel and produced a vectorized embedding value from several concatenated text attributes within each library JSON document. A full description of this data wrangling process is beyond the scope of this documentation, but the process itself is in file 'impl/app/wrangle.py' in the repo.
Depending on the Cosmos DB API you chose: