I'd like to know more about prospects for LLM-based conversational interfaces to Neo4j. Unfortunately, I missed out on watching this talk and can't find a copy online. (Help?)
Use Case
I'm interested in using Neo4j as the data store for software development projects. In particular, I'd like to model and track the entities and relationships involved in a running Elixir / Erlang program. This information could, I believe, allow a programmer to answer a variety of questions about a running system and the underlying code base.
There are a lot of entities involved in any substantial software system. In the case of Elixir, these would include applications, attributes, functions, logs, messages, macros, modules, NIFs, nodes, packages, ports, processes, tasks, tests, VM status, etc. Clearly, there would be even more relationships, e.g.:
- a module defines an attribute, function, or macro
- a function invokes a function or macro
- a function is used within a process
- a process sends messages to others
- a process supervises other processes
- ...
By way of clarification, I don't want to record every dynamic event (e.g., function call, message), just the fact that the event has been observed and (optionally) the number of times it has occurred.
Discussion
Although Cypher works well for "serious" users, its concepts and syntax have a substantial learning curve. So, more casual users might do better with something like chatGPT. They could then ask questions in (say) English and get results as if they had generated the necessary Cypher query.
Is anything like this in prospect? Inquiring gnomes need to mine...
-r