Open Trogluddite opened 5 months ago
some brainstorming from Discord -- we'll eventually need something like this in the API: crawler interface
read seeds
remove seed
add seed
start crawl
check crawl status
query interfaces
set query (set filters, query strings)
do query
build new MatrixMarkov instance (either add a new instance, or replace current)
list matrix markov instances (if we handle multiple concurrent networks)
delete matrix markov instances (if we handle multiples)
stub out new discord commands. The discord bot should allow these commands to run, but should just echo back "ok" or something; they don't need to do real work yet.
The idea here is to develop a user-flow (UX) for the crawler, querier, and markov selection process.
Crawling the web and building bayesian networks will be batch tasks. We want the discord bot to issue commands to trigger those processes, and wait for 'done' responses, which will be returned to Discord.
'searching' the Bayesian network is really going to be generating candidate texts, comparing them to a target, and choosing the best-match from the candidates as a result.
I'm envisioning a workflow like this:
/search
command, with some 'target_text' and 'show_sources' bool. This, as far as the UX is concerned, operate the same as it currently does./query
withfilters
andquery_params
values./reload_from_query
or the like