cmoriarty/braid
1
0
Fork 0
Code Issues 15 Pull requests Projects Releases Packages 1 Wiki Activity Actions

UI improvements #49

New issue
Closed
opened 2026-09-26 13:56:25 -04:00 by cmoriarty · 1 comment
cmoriarty commented 2026-09-26 13:56:25 -04:00
Owner
Copy link

INBOX:

  1. Remove all text descriptions. ("Nothing is waiting on you", "Every open gate across every run. Answering here is the same card the transcript shows.", "When an agent asks something, or a graph gate opens, it lands here.")
  2. When there are no current reviews waiting, create a simple "You are caught up on reviews!" message, with some appropriate image conveying something being done (maybe a flag?)
  3. Move the inbox link to the top bar, it is currently on the left pane at the bottom. It should have a little mail icon rather than a diamond. It should go just to the left of the current settings gear icon, and just to the right of the running label.

INFERENCE:

  1. Move the graphic that shows how many slots are active to the top bar, to the right of the running label. Remove the kv percentage. Clicking on either the running label or this active slots labal should bring to the inference page.
  2. Remove the label showing "http://192.168.1.105:8090/metrics"
  3. Remove "No one is waiting for a slot."
  4. Change name of "MEMORY IN USE" to be "KV CACHE USAGE"
  5. When queued sequential tasks are active, the INFERENCE page incorrectly says "Waiting for a slot", which is should say "Queued requests" and indicate which task they are waiting on. "Waiting for a slot" should only be used when the system is truly backed up, and vLLM has more than 6 requests.
  6. I'd like the boxes used as "SLOTS" to take up a full row, and display more info about what it is doing. Same with the boxes used for "Waiting for a slot" or soon to be "Queued requests"
  7. Remove the "OVER TIME" and info icon for it, the controls are intuitive enough without them.

BRAID TITLE (top left):

  1. Design a logo/wordmark for braid that tries to incorporate the "braid" nature of the system.
  2. Currently, clicking the BRAID logo is the only way to get to the main dashboard, perhaps it should have a little home icon to the left to make it more clear, and then clicking either home or braid takes you to the dashboard.

SETTINGS:

  1. Make the gear a bigger, its currently very small and can be overlooked easily.
  2. Remove "How Braid runs"
  3. Tuck all of the helper text into info icons, use the same as what is used on the inference page for the various time series charts (ex "Any server that speaks the OpenAI API at /v1. vLLM’s own metrics feed the Inference charts; other kinds leave them empty.", "Tokens, from the model card or the server’s max model length.", etc)

LEFT EXPANDED PANE:

  1. When the left pane is expanded, I would like each run listed to have a box 2-3x the size it is currently, so that text can wrap, and it be more clear what the run is doing. If it could show some kind of progress bar for whichever pipeline it is using, that would be cool. I suppose we should also keep the boxes the same size when the left pane collapses, and try and also use the extra space for something useful.

DASHBOARD:

  1. REmove "x on the ledger" label

GRAPH PANE:

  1. By default, an active step should be selected, so if the one selected ends, it should move to an active one. This should be a setting in the settings form, by default it is on, but if the user doesn't want the graph always be showing an active agent, they could disable it.
  2. The grey text under the top progress bar includes the "osf.run_2lk3j45hl2k3jh45lk23jh45k2j3h45" branch name which is just noise, and can be removed.
  3. The repo label (ex soundcheck) under the progress bar should be a link to the repo URL.
  4. Finished tasks shouldn't get hidden.
  5. Finished tasks should keep how long they took when they are in a finished state.
  6. There should be a total running total time kept near the top progress bar.
  7. The step should indicate if has active subagents running.
  8. The "silent for ..." label should not be shown if subagents ARE running actively. Only if absolutely nothing is happening in that session.
  9. When auto-approve is on, the "human." steps should say "agent." instead.
  10. The chosen pipeline name should be at the top of the pane.

AGENT TOP PANE:

  1. The title of the run is currently in grey letters, and the step is in white, lets make the title more prominent, perhaps use the orange shade I have seen for the "Waiting for a slot" message. If the title came directly from a forgejo issue, it should become a link to that issue. No need to show the "- Step 15 openspec.apply[1]" after the title, that is already clear from the graph pane.

STEP DETAILS:

  1. "past the 112k alarm — a structured handoff and a respawn come next · 1 respawn · respawned" this text is too free form. All of these should be fields showing the process more clearly.
  2. Hide helper text like "Superseded attempts are kept, never deleted. * = derived from the step." in an info icon.
  3. The ATTEMPTS row has a #1 in flight, this text under it: "session gDZ17Ea9 · peak 112,766 in · 1 respawn". Seems like we could display that more visually appealing, and also possibly provide context for why the respawn happened, it sounds bad, but it may just be ok.

OVERALL STYLE:

  1. From a UX standpoint do a review of all forms and make any small tweaks that align with a minimalist, beautiful, easy to understand user interface.
INBOX: 1. Remove all text descriptions. ("Nothing is waiting on you", "Every open gate across every run. Answering here is the same card the transcript shows.", "When an agent asks something, or a graph gate opens, it lands here.") 2. When there are no current reviews waiting, create a simple "You are caught up on reviews!" message, with some appropriate image conveying something being done (maybe a flag?) 3. Move the inbox link to the top bar, it is currently on the left pane at the bottom. It should have a little mail icon rather than a diamond. It should go just to the left of the current settings gear icon, and just to the right of the running label. INFERENCE: 1. Move the graphic that shows how many slots are active to the top bar, to the right of the running label. Remove the kv percentage. Clicking on either the running label or this active slots labal should bring to the inference page. 2. Remove the label showing "http://192.168.1.105:8090/metrics" 3. Remove "No one is waiting for a slot." 4. Change name of "MEMORY IN USE" to be "KV CACHE USAGE" 5. When queued sequential tasks are active, the INFERENCE page incorrectly says "Waiting for a slot", which is should say "Queued requests" and indicate which task they are waiting on. "Waiting for a slot" should only be used when the system is truly backed up, and vLLM has more than 6 requests. 6. I'd like the boxes used as "SLOTS" to take up a full row, and display more info about what it is doing. Same with the boxes used for "Waiting for a slot" or soon to be "Queued requests" 7. Remove the "OVER TIME" and info icon for it, the controls are intuitive enough without them. BRAID TITLE (top left): 1. Design a logo/wordmark for braid that tries to incorporate the "braid" nature of the system. 2. Currently, clicking the BRAID logo is the only way to get to the main dashboard, perhaps it should have a little home icon to the left to make it more clear, and then clicking either home or braid takes you to the dashboard. SETTINGS: 1. Make the gear a bigger, its currently very small and can be overlooked easily. 2. Remove "How Braid runs" 3. Tuck all of the helper text into info icons, use the same as what is used on the inference page for the various time series charts (ex "Any server that speaks the OpenAI API at /v1. vLLM’s own metrics feed the Inference charts; other kinds leave them empty.", "Tokens, from the model card or the server’s max model length.", etc) LEFT EXPANDED PANE: 1. When the left pane is expanded, I would like each run listed to have a box 2-3x the size it is currently, so that text can wrap, and it be more clear what the run is doing. If it could show some kind of progress bar for whichever pipeline it is using, that would be cool. I suppose we should also keep the boxes the same size when the left pane collapses, and try and also use the extra space for something useful. DASHBOARD: 1. REmove "x on the ledger" label GRAPH PANE: 1. By default, an active step should be selected, so if the one selected ends, it should move to an active one. This should be a setting in the settings form, by default it is on, but if the user doesn't want the graph always be showing an active agent, they could disable it. 2. The grey text under the top progress bar includes the "osf.run_2lk3j45hl2k3jh45lk23jh45k2j3h45" branch name which is just noise, and can be removed. 3. The repo label (ex soundcheck) under the progress bar should be a link to the repo URL. 4. Finished tasks shouldn't get hidden. 5. Finished tasks should keep how long they took when they are in a finished state. 6. There should be a total running total time kept near the top progress bar. 7. The step should indicate if has active subagents running. 8. The "silent for ..." label should not be shown if subagents ARE running actively. Only if absolutely nothing is happening in that session. 9. When auto-approve is on, the "human." steps should say "agent." instead. 10. The chosen pipeline name should be at the top of the pane. AGENT TOP PANE: 1. The title of the run is currently in grey letters, and the step is in white, lets make the title more prominent, perhaps use the orange shade I have seen for the "Waiting for a slot" message. If the title came directly from a forgejo issue, it should become a link to that issue. No need to show the "- Step 15 openspec.apply[1]" after the title, that is already clear from the graph pane. STEP DETAILS: 1. "past the 112k alarm — a structured handoff and a respawn come next · 1 respawn · respawned" this text is too free form. All of these should be fields showing the process more clearly. 2. Hide helper text like "Superseded attempts are kept, never deleted. * = derived from the step." in an info icon. 3. The ATTEMPTS row has a #1 in flight, this text under it: "session gDZ17Ea9 · peak 112,766 in · 1 respawn". Seems like we could display that more visually appealing, and also possibly provide context for why the respawn happened, it sounds bad, but it may just be ok. OVERALL STYLE: 1. From a UX standpoint do a review of all forms and make any small tweaks that align with a minimalist, beautiful, easy to understand user interface.
cmoriarty commented 2026-09-26 15:50:37 -04:00
Author
Owner
Copy link

Shipped in #50 (a19ea2e) and archived as openspec/changes/archive/2026-09-26-ui-polish-49 (4f19d6f).

  • Inbox: explanatory text removed; empty state is a flag and "You are caught up on reviews!"; the inbox link is in the top bar with a mail icon, between the slot meter and the gear.
  • Inference: slot meter in the top bar (no kv %), and it and the running label open Inference; metrics URL, "No one is waiting for a slot." and "Over time" removed; "KV cache usage"; slots and queued requests are full-width rows. Queued steps say what they wait on (e.g. after openspec.apply[1]); "Waiting for a slot" only when vLLM is holding requests or every slot is running.
  • Title: home icon + a braid wordmark, both to the dashboard.
  • Settings: bigger gear, "How Braid runs" removed, helper text behind ⓘ, new "Follow the active step" setting.
  • Left pane: taller cards with wrapped titles and a pipeline progress bar; collapsed cards keep the height with a vertical bar and settled/total.
  • Dashboard: "N on the ledger" removed.
  • Graph pane: follows the active step (setting, default on); branch removed; repo links to the forge; finished steps never hidden and keep their duration; run total time by the progress bar; active subagents shown and no "silent for" while they work; human. reads agent. under auto-approve; pipeline name at the top.
  • Agent top pane: title in orange, linked to its issue, no step suffix.
  • Step details: budget as labelled fields; attempts as chips with the respawn explained as the planned hand-off at the alarm; notes behind ⓘ.
  • Overall: New Run hints moved behind ⓘ to match Settings.
Shipped in #50 (a19ea2e) and archived as `openspec/changes/archive/2026-09-26-ui-polish-49` (4f19d6f). - **Inbox:** explanatory text removed; empty state is a flag and "You are caught up on reviews!"; the inbox link is in the top bar with a mail icon, between the slot meter and the gear. - **Inference:** slot meter in the top bar (no kv %), and it and the running label open Inference; metrics URL, "No one is waiting for a slot." and "Over time" removed; "KV cache usage"; slots and queued requests are full-width rows. Queued steps say what they wait on (e.g. `after openspec.apply[1]`); "Waiting for a slot" only when vLLM is holding requests or every slot is running. - **Title:** home icon + a braid wordmark, both to the dashboard. - **Settings:** bigger gear, "How Braid runs" removed, helper text behind ⓘ, new "Follow the active step" setting. - **Left pane:** taller cards with wrapped titles and a pipeline progress bar; collapsed cards keep the height with a vertical bar and settled/total. - **Dashboard:** "N on the ledger" removed. - **Graph pane:** follows the active step (setting, default on); branch removed; repo links to the forge; finished steps never hidden and keep their duration; run total time by the progress bar; active subagents shown and no "silent for" while they work; `human.` reads `agent.` under auto-approve; pipeline name at the top. - **Agent top pane:** title in orange, linked to its issue, no step suffix. - **Step details:** budget as labelled fields; attempts as chips with the respawn explained as the planned hand-off at the alarm; notes behind ⓘ. - **Overall:** New Run hints moved behind ⓘ to match Settings.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
archive/prefect-kestra-experiments
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
cmoriarty/braid#49
No description provided.