I run the node c-otto.de and left a few notes on https://c-otto.de. Aside from that, I think that having lots of capital is crucial. Your node needs to be well connected, but it also needs to have enough inbound and outbound liquidity for a lucrative large forward to happen (I love those 10M+ forwards, even though they are rather rare). Most implementations take the channel capacity into account when finding routes, which also means that you should try to have large channels.
I think that LnRouter (https://lnrouter.app/) is really helpful. Before you open a (large) channel to a node, make sure that it can actually put those sats to good use. If the peer lacks outbound liquidity, it doesn't help to open yet another channel TO them.
Regarding rebalances: Don't rebalance if it's not worth it. If you use rebalance-lnd, only routes are used that (if routing happens) put you in a better position.
Another advice: weed out bad peers, keep the good peers, use available sats to improve your position and/or try out new peers. With lnd-manageJ (https://github.com/C-Otto/lnd-manageJ) you can get a rating for each peer based on routing activity, and use that number to guide your decisions (which is what I do).
I know this doesn't directly address your bounty. But I thought maybe this would be a better alternative.
One of the challenges of working with OpenAI API is the rate limiting (3/minute in my case). Instead of consuming a call on every link post, this bot will just summarize where a user has asked for a tl;dr summary.
Still a WIP. Trying different ways to get the text from a website. Settled on python-readability and bs4.
I run the node c-otto.de and left a few notes on https://c-otto.de. Aside from that, I think that having lots of capital is crucial. Your node needs to be well connected, but it also needs to have enough inbound and outbound liquidity for a lucrative large forward to happen (I love those 10M+ forwards, even though they are rather rare). Most implementations take the channel capacity into account when finding routes, which also means that you should try to have large channels.
I think that LnRouter (https://lnrouter.app/) is really helpful. Before you open a (large) channel to a node, make sure that it can actually put those sats to good use. If the peer lacks outbound liquidity, it doesn't help to open yet another channel TO them.
Some helpful stuff from Alex Bosworth:
How I set my fees (and why):
https://yalls.org/articles/97d67df1-d721-417d-a6c0-11d793739be9:0965AC5E-56CD-4870-9041-E69616660E6F/a81e96f9-b1a7-4763-9f7a-adf226ebde59
Regarding rebalances:
Don't rebalance if it's not worth it. If you use rebalance-lnd, only routes are used that (if routing happens) put you in a better position.
Another advice: weed out bad peers, keep the good peers, use available sats to improve your position and/or try out new peers. With lnd-manageJ (https://github.com/C-Otto/lnd-manageJ) you can get a rating for each peer based on routing activity, and use that number to guide your decisions (which is what I do).
I know this doesn't directly address your bounty. But I thought maybe this would be a better alternative.
One of the challenges of working with OpenAI API is the rate limiting (3/minute in my case). Instead of consuming a call on every link post, this bot will just summarize where a user has asked for a tl;dr summary.
Still a WIP. Trying different ways to get the text from a website. Settled on python-readability and bs4.