Recent requests
(Latest activity from the request log.)
Should be adapted by the following things:
1.) For cloud requests show cost information
2.) For running requests show token information
3.) User names should be shown (full names = First name + Last name, not username!)
4.) /api/logosdb/paginated_requests is currently called multiple times on initial load and this request is
4.1.) slow
4.2.) Has too many results (does not respect selected time range - default = month)
4.3.) Is not needed, as same information is already collected through /api/ws/stats/v2 (type: requests)
=> Do only use /api/ws/stats/v2 and respect time selection
5.) It misses the information which lane is loaded on which GPU(s)
6.) Misses historic information on how many tokens have been processed or how many costs have been caused by cloud resources
7.) I miss the date range selection of "the last 30 days" e.g. - that shall be the default selection
8.) VRAM remaining is currently placed weirdly, it shall rather be shown in the provider selection tag as it is highly dependant on provider. Also it shall respect the global time range selection instead of having a own one
9.) The "Request volume" buckets are unclear, e.g. for the current month it shows the same x-label entry "Aug 4" four times right now.
10.) It has the possibility to unload lanes but lacks possibility to manually load a lane
11.) It also lacks the information on when a node is calibrating right now
Recent requests
(Latest activity from the request log.)
Should be adapted by the following things:
1.) For cloud requests show cost information
2.) For running requests show token information
3.) User names should be shown (full names = First name + Last name, not username!)
4.) /api/logosdb/paginated_requests is currently called multiple times on initial load and this request is
4.1.) slow
4.2.) Has too many results (does not respect selected time range - default = month)
4.3.) Is not needed, as same information is already collected through /api/ws/stats/v2 (type: requests)
=> Do only use /api/ws/stats/v2 and respect time selection
5.) It misses the information which lane is loaded on which GPU(s)
6.) Misses historic information on how many tokens have been processed or how many costs have been caused by cloud resources
7.) I miss the date range selection of "the last 30 days" e.g. - that shall be the default selection
8.) VRAM remaining is currently placed weirdly, it shall rather be shown in the provider selection tag as it is highly dependant on provider. Also it shall respect the global time range selection instead of having a own one
9.) The "Request volume" buckets are unclear, e.g. for the current month it shows the same x-label entry "Aug 4" four times right now.
10.) It has the possibility to unload lanes but lacks possibility to manually load a lane
11.) It also lacks the information on when a node is calibrating right now