You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I know I said that generics lead to complections, and I still stand by that sentiment. However, folks should be free to use alternative search engines.
Thankfully, LLMs already have a pretty generic interface: text. As long as connectors are able to return text, then we should be able to use that as an interfacing layer.
Another consideration is the frontend. Slack is both a frontend and a backend. Perhaps those ideas could be split into distinct connectors.
Finally, is the matter of business logic layer, which does not currently exist.
Splitting CogniQ into three layers may provide the modularity to permit users to write their own business logic.
In fact, the design goal should be that connector packages should be able to be used as a package rather than as part of a whole.
The text was updated successfully, but these errors were encountered:
Naturally, these interfaces will change as the project evolves.
However, I hope we can keep in mind the goal of not overly genericizing the interfaces.
Fixes#7
* refactor: define an interface for search
* chore: remove unneessary slack_app_token
This is used for socket mode, which I am not using
* chore: permit writing linting changes
* refactor: black 🤖 [skip ci]
* chore: only trigger autolinter on pull request
* refactor: do an interface pass, and rename some things.
Naturally, these interfaces will change as the project evolves.
However, I hope we can keep in mind the goal of not overly genericizing the interfaces.
Fixes#7
* refactor: black 🤖 [skip ci]
* fix: await correctly
* refactor: black 🤖 [skip ci]
---------
Co-authored-by: jim80net <jim80net@users.noreply.github.com>
I waffled on this for a while, before deciding to do a refactor pass and close this issue.
On second thought, I don't want to be in the business of building a framework. The complexity costs seem to be problematic.
The project will continue to evolve as I add connectors. We'll let the problem of modularity remain a high level concern, and close this issue. It'll come up again as the situation evolves.
I know I said that generics lead to complections, and I still stand by that sentiment. However, folks should be free to use alternative search engines.
Thankfully, LLMs already have a pretty generic interface: text. As long as connectors are able to return text, then we should be able to use that as an interfacing layer.
Another consideration is the frontend. Slack is both a frontend and a backend. Perhaps those ideas could be split into distinct connectors.
Finally, is the matter of business logic layer, which does not currently exist.
Splitting CogniQ into three layers may provide the modularity to permit users to write their own business logic.
In fact, the design goal should be that connector packages should be able to be used as a package rather than as part of a whole.
The text was updated successfully, but these errors were encountered: