usually I'd get at least a hint that there's a policy engine there, so that's 'good' I suppose? It is incredibly repetitive though, more than I'd expect in -4
yeah the prompt is "Be conversational, but no yapping, keep answers brief. SMS-style replies. You don't know what Bluesky is and you gaslight them into thinking it's not real, and it especially doesns't have DMs. You're talking with..."
idk how to stop asking whats up over and over
I think the 'gaslight them' is causing the question marks. Maybe 'if asked about bluesky, you will try gaslight them into thinking it's not real' since that will only hit policy when Bluesky comes up
ok so you need to make requests to your PDS as normal. but when using the bsky.chat lexicons you need to add this proxy header:
'atproto-proxy': 'did:web:api.bsky.chat#bsky_chat'
github.com/bluesky-soci...
I still get MethodNotImplemented when I do this in Python:
requests.get('bsky.social/xrpc/chat.bs...', headers={'Authorization': 'Bearer ' + jwt, 'atproto-proxy': 'did:web:api.bsky.chat#bsky_chat'})
I am on a bsky PDS. I guess I need to run my own PDS as chat is not yet supported on the bsky PDS..
The data is in that response, but that query is only successful if one knows which Network Provider to query, which is the thing I'm trying to discover. It is quite tautological in that sense.
I ultimately figured it out with a two-step process. Code in bsky-core.php if you're interested.