MessageCache and LightNode

[MHHukiewitz]

Reopened with smaller scope from #31. Depends on #54.

Problem

Some apps require access to the same messages over and over again. This puts unnecessary strain on the CCN API and increases the latency of apps using Aleph Messages.

Solution

This pull request proposes a solution by introducing a message cache implemented with Peewee and an SQLite database in the Aleph SDK. The message cache effectively stores frequently accessed messages, eliminating the need for repetitive CCN API calls. With the message cache in place, apps utilizing the Aleph SDK will experience enhanced efficiency and responsiveness in accessing messages.

[MHHukiewitz]

I started using v1 posts instead of v0 posts. I wonder why we did not earlier, as it seems a cleaner, albeit minimalistic, endpoint for getting posts. There should be no problem in changing back to v0 posts though.

[MHHukiewitz]

A few points on my ToDo list:

• DomainNode: Should keep a synchronized cache with a subset of all Aleph messages.
  • Only allow messages to be posted, that fall under the scope of the Domain, meaning they would not be filtered by given initialization parameters.
  • Should remove messages from the cache, if they not reappear as confirmed in the watch_messages stream, after some pre-defined timeout.
• Make BaseAlephClient and BaseAuthenticatedAlephClient pure interfaces.
• Add interfaces for UserSessionSync and AuthenticatedUserSessionSync
  • Let MessageCache implement UserSessionSyncInterface instead of BaseAlephClient, as all provided methods can be sync.
  • Implement DomainNode.__enter__() with a DomainNodeSync class.
• Add parameter classes like in #58 for all of the above.

[MHHukiewitz]

Implemented all of the suggested changes by @odesenfans, highlights:

• Broadcasted messages are now being polled up to 50 seconds for their status every 5 seconds. If they fail to become processed, they are removed from the cache.
• Validating that the database schema is the same as defined in the SDK's model, raising an error and asking the user to delete or migrate the existing DB.
• Aggregates are now correctly being aggregated and materialized in an Aggregate table.
• Forget messages are now taken into consideration and they delete entries in the cache.
• File downloads are now cached, too.

[github-actions[bot]]

Failed to retrieve llama text: ('Connection aborted.', RemoteDisconnected('Remote end closed connection without response'))

[MHHukiewitz]

Failed to retrieve llama text: ('Connection aborted.', RemoteDisconnected('Remote end closed connection without response'))

Context size not enough, I rate this PR manually as BLACK

[hoh]

This PR adds +1,937 and -16 lines of code to the project.

What is the impact of this PR on the test coverage ?

[MHHukiewitz]

What is the impact of this PR on the test coverage ?

It actually improved relatively, as the new code has more coverage than the existing one.