Repository navigation
Commit 044e763
authored
fix(webui): first-run call to action, onboarding path for a fresh setup, and "what needs me" on the landing page (#1049)
* fix(webui): give first run a way forward (F3, F19) and keep "what needs me" on the landing page (F4)
Web UI UX audit (docs/qa/ux-audit-webui-2026-08.md), first-run and onboarding.
F3 — Servers page, zero servers. The page a new user lands on after closing
the wizard said "No servers found / No servers available" and offered nothing
to click, above four zero-valued stat tiles, four zero-count filter pills and
a search box with nothing to search. It now renders the AgentTokens-style
empty state: a headline, one line of why upstream servers matter, Add Server
(primary), Browse Registry, and a quiet "Import from your AI client configs"
link into the wizard's Servers step. The stat/filter/search chrome is hidden
until a server exists, and "you have nothing yet" is now a separate state from
"nothing matched your filter" — the two need different answers.
F4 residual (after #1044) — #1044 routed /usage, made the analytics panel the
default landing page and demoted the hub diagram, which is most of the fix.
What it did not carry across is the other half the finding asked for: the
landing page still answered "what happened" and not "what needs me". The
servers-needing-attention and tools-pending-approval banners were inside the
Overview panel, so a fresh landing could render charts while a server was down
and an unreviewed tool was waiting, and say neither. They move above the panel
switcher: visible on whichever panel is active, silent when there is nothing
to act on.
F19 — onboarding wizard, Servers step. It was entirely "pick which servers
from your existing AI clients to import", so a user with no prior MCP config
hit a dead end. Adds a nothing-to-import branch (browse the registry / add a
server manually). Import & quarantine becomes the single primary and the
unreviewed path becomes a link that says what it skips, "Import without
review" — previously two similar primaries with the safer one rendered weaker.
The Docker-isolation/quarantine panel is expanded by default and moved out of
the sticky footer into the step body, where it can be open without squeezing
the list; the import list is capped and scrolls in place instead of clipping
its last row; and every step gets a Back.
* fix(webui): initialise the wizard when it is mounted already-open, and keep every step exitable
Cross-model review (opencode/gpt-sol) on the first-run changes.
P1 — the Servers page's import link sets the store flag and then routes to the
Dashboard, which is the component that mounts the wizard. The wizard therefore
came into existence with `show` already true and never saw a transition to
watch: it rendered, but the open sequence never ran. The onMounted fallback
that existed for this case kicked off the fetches but never chose a tab and
never consumed the request — so the wizard opened on Clients regardless of what
the opener asked for, and left the request behind to hijack the next plain
open. One `immediate` watcher now owns opening; the fallback is gone. The
requested tab is read before the first await, so it cannot outlive its own open
or be consumed twice, and it applies immediately so the wizard never paints the
wrong step while the fetches are in flight.
P2 — `goToRegistry` did not await `dismiss()`. `dismiss()` awaits its
mark-skipped calls before emitting `close`, so routing first tore down the
Dashboard that owns the open flag and the emit could land with nothing left to
clear it, springing the wizard back open on return.
P2 — the Servers page fetches on neither mount nor route: App.vue and the
Dashboard do. That leaves a window where `loading` is false and the list is
still unknown, in which the filter/search empty state claimed "no servers match
your search criteria" about a list nobody had fetched. Adds an explicit pending
state, gated so servers already in hand are never held behind it.
Also restores Close to the import action footer, which the review asked about
and the live run then proved: a step whose only exits are "import" strands the
user, and the committed e2e sweep dismisses an auto-opened wizard through
exactly that control — on a fresh instance the wizard can land on this step.
The security disclosure gains `group` so its arrow reflects the now-open state.
* fix(webui): abandon superseded wizard opens, and stop a failed refresh from hiding the server list
Round 2 of the cross-model review (opencode/gpt-sol).
P1 — nothing cancels the wizard's open sequence. A close, or a close followed
by a reopen, while its five fetches were still in flight let the abandoned run
finish and stamp its tab over the newer state — and call startPolling() on a
wizard that was no longer open, leaving a 5s poll running until unmount. Each
run now takes a generation token and checks it still owns the wizard before it
writes; closing and unmounting both retire whatever is in flight.
P2 — `loading.error` was write-only. Silent background refreshes set it and no
success ever cleared it, so a single transient blip replaced the Servers page
with a full-page "Failed to load servers" for the rest of the session, hiding a
list we were still holding. A list that arrives now clears the error, and the
blocking error state is reserved for the case with no fallback: when servers
are already in hand, a failed refresh degrades to a non-blocking notice above
the grid that says the data may be stale and offers Retry.
* fix(webui): sequence the wizard's config read and the server-list failure path
Round 3 of the cross-model review (opencode/gpt-sol).
The open sequence's config read is the one fetch whose result the user can also
edit: it drives the quarantine and Docker-isolation checkboxes, and this branch
expands that panel by default, so toggling one the moment the step appears is
now the expected path rather than a corner. A read left over from a superseded
open landing afterwards would silently flip the checkbox back to the old server
value. It now carries the open's generation token and does not commit if it has
been superseded. The other fetches in the sequence write read-only display
state that either run populates identically, so they are left alone.
Server-list failures now take the same ticket as successes. Only the list
itself was sequenced, so an older request failing after a newer one had already
delivered a list would raise a stale error over fresh data — the mirror image
of the stale-list problem the sequencing exists to prevent.
* fix(webui): let a failed server fetch advance the sequencing mark
Round 4 of the cross-model review (opencode/gpt-sol).
Only a delivered list advanced the mark, so a newer request that failed left it
behind — and an older list arriving afterwards still passed the check, applying
stale servers over the newer outcome and clearing its error on the way. Both
outcomes are news about the same list, so both now claim the ticket.1 parent 7e33f68 commit 044e763
9 files changed
Lines changed: 1469 additions & 240 deletions
File tree
- frontend
- src
- components
- stores
- views
- tests/unit
Large diffs are not rendered by default.
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
17 | 17 | | |
18 | 18 | | |
19 | 19 | | |
| 20 | + | |
| 21 | + | |
| 22 | + | |
20 | 23 | | |
21 | 24 | | |
22 | 25 | | |
| |||
29 | 32 | | |
30 | 33 | | |
31 | 34 | | |
| 35 | + | |
| 36 | + | |
| 37 | + | |
| 38 | + | |
| 39 | + | |
| 40 | + | |
| 41 | + | |
| 42 | + | |
32 | 43 | | |
33 | 44 | | |
34 | 45 | | |
| |||
146 | 157 | | |
147 | 158 | | |
148 | 159 | | |
149 | | - | |
| 160 | + | |
| 161 | + | |
150 | 162 | | |
151 | 163 | | |
152 | 164 | | |
153 | 165 | | |
154 | 166 | | |
| 167 | + | |
| 168 | + | |
| 169 | + | |
| 170 | + | |
| 171 | + | |
| 172 | + | |
| 173 | + | |
155 | 174 | | |
156 | 175 | | |
157 | 176 | | |
| |||
161 | 180 | | |
162 | 181 | | |
163 | 182 | | |
| 183 | + | |
164 | 184 | | |
165 | 185 | | |
166 | 186 | | |
| |||
179 | 199 | | |
180 | 200 | | |
181 | 201 | | |
| 202 | + | |
182 | 203 | | |
183 | 204 | | |
184 | 205 | | |
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
105 | 105 | | |
106 | 106 | | |
107 | 107 | | |
108 | | - | |
109 | | - | |
| 108 | + | |
| 109 | + | |
| 110 | + | |
| 111 | + | |
| 112 | + | |
| 113 | + | |
| 114 | + | |
110 | 115 | | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
| 119 | + | |
| 120 | + | |
111 | 121 | | |
112 | 122 | | |
113 | 123 | | |
| 124 | + | |
| 125 | + | |
| 126 | + | |
| 127 | + | |
| 128 | + | |
114 | 129 | | |
115 | 130 | | |
116 | 131 | | |
| |||
120 | 135 | | |
121 | 136 | | |
122 | 137 | | |
| 138 | + | |
| 139 | + | |
| 140 | + | |
| 141 | + | |
| 142 | + | |
| 143 | + | |
| 144 | + | |
| 145 | + | |
| 146 | + | |
123 | 147 | | |
124 | 148 | | |
125 | 149 | | |
126 | 150 | | |
127 | 151 | | |
128 | | - | |
| 152 | + | |
129 | 153 | | |
130 | 154 | | |
131 | | - | |
| 155 | + | |
132 | 156 | | |
133 | 157 | | |
134 | 158 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
6 | 6 | | |
7 | 7 | | |
8 | 8 | | |
9 | | - | |
10 | | - | |
11 | | - | |
12 | | - | |
13 | | - | |
14 | | - | |
15 | | - | |
16 | | - | |
17 | | - | |
18 | | - | |
19 | | - | |
20 | | - | |
21 | | - | |
22 | | - | |
23 | | - | |
24 | | - | |
25 | | - | |
26 | | - | |
27 | | - | |
28 | | - | |
29 | | - | |
30 | | - | |
31 | | - | |
32 | | - | |
33 | | - | |
34 | | - | |
35 | | - | |
36 | | - | |
37 | | - | |
38 | | - | |
39 | | - | |
40 | | - | |
41 | | - | |
42 | | - | |
43 | | - | |
44 | | - | |
45 | | - | |
46 | | - | |
47 | | - | |
48 | | - | |
49 | | - | |
50 | | - | |
51 | | - | |
52 | | - | |
53 | | - | |
54 | | - | |
55 | | - | |
56 | | - | |
57 | | - | |
58 | | - | |
59 | | - | |
60 | | - | |
61 | | - | |
62 | | - | |
63 | | - | |
64 | | - | |
65 | | - | |
66 | | - | |
67 | | - | |
68 | | - | |
69 | | - | |
70 | | - | |
71 | | - | |
72 | | - | |
73 | | - | |
74 | | - | |
75 | | - | |
76 | | - | |
77 | | - | |
78 | | - | |
79 | | - | |
80 | | - | |
81 | | - | |
82 | | - | |
83 | | - | |
84 | | - | |
85 | | - | |
86 | | - | |
87 | | - | |
88 | | - | |
89 | | - | |
90 | | - | |
91 | | - | |
92 | | - | |
93 | | - | |
94 | | - | |
| 9 | + | |
| 10 | + | |
| 11 | + | |
| 12 | + | |
| 13 | + | |
| 14 | + | |
95 | 15 | | |
96 | 16 | | |
97 | 17 | | |
| |||
179 | 99 | | |
180 | 100 | | |
181 | 101 | | |
| 102 | + | |
| 103 | + | |
| 104 | + | |
| 105 | + | |
| 106 | + | |
| 107 | + | |
| 108 | + | |
| 109 | + | |
| 110 | + | |
| 111 | + | |
| 112 | + | |
| 113 | + | |
| 114 | + | |
| 115 | + | |
| 116 | + | |
| 117 | + | |
| 118 | + | |
| 119 | + | |
| 120 | + | |
| 121 | + | |
| 122 | + | |
| 123 | + | |
| 124 | + | |
| 125 | + | |
| 126 | + | |
| 127 | + | |
| 128 | + | |
| 129 | + | |
| 130 | + | |
| 131 | + | |
| 132 | + | |
| 133 | + | |
| 134 | + | |
| 135 | + | |
| 136 | + | |
| 137 | + | |
| 138 | + | |
| 139 | + | |
| 140 | + | |
| 141 | + | |
| 142 | + | |
| 143 | + | |
| 144 | + | |
| 145 | + | |
| 146 | + | |
| 147 | + | |
| 148 | + | |
| 149 | + | |
| 150 | + | |
| 151 | + | |
| 152 | + | |
| 153 | + | |
| 154 | + | |
| 155 | + | |
| 156 | + | |
| 157 | + | |
| 158 | + | |
| 159 | + | |
| 160 | + | |
| 161 | + | |
| 162 | + | |
| 163 | + | |
| 164 | + | |
| 165 | + | |
| 166 | + | |
| 167 | + | |
| 168 | + | |
| 169 | + | |
| 170 | + | |
| 171 | + | |
| 172 | + | |
| 173 | + | |
| 174 | + | |
| 175 | + | |
| 176 | + | |
| 177 | + | |
| 178 | + | |
| 179 | + | |
| 180 | + | |
| 181 | + | |
| 182 | + | |
| 183 | + | |
| 184 | + | |
| 185 | + | |
| 186 | + | |
| 187 | + | |
182 | 188 | | |
183 | 189 | | |
184 | 190 | | |
| |||
0 commit comments