Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Showed failing tests properly in the UI #45

Merged
merged 4 commits into from
Dec 10, 2024
Merged

Showed failing tests properly in the UI #45

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
43c3e50
WIP
mattpocock Dec 10, 2024
cf4108d
Made tests pass
mattpocock Dec 10, 2024
5606e75
Added visual statuses to UI
mattpocock Dec 10, 2024
ba8eaa4
Reduced example size
mattpocock Dec 10, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
80 changes: 47 additions & 33 deletions apps/evalite-ui/app/components/score.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@ import {
ChevronRightCircleIcon,
ChevronUpCircleIcon,
LoaderCircleIcon,
XCircleIcon,
} from "lucide-react";
import { cn } from "~/lib/utils";

Expand All @@ -12,44 +13,57 @@ export const Score = (props: {
score: number;
state: ScoreState;
isRunning: boolean;
evalStatus: "success" | "fail";
iconClassName?: string;
}) => {
return (
<span className="flex items-center space-x-2">
<span>{Math.round(props.score * 100)}%</span>
{props.isRunning ? (
<span className="text-gray-500">
<LoaderCircleIcon
className={cn(
"size-3 text-blue-500 animate-spin",
props.iconClassName
)}
/>
</span>
) : (
<>
{props.state === "up" && (
<ChevronUpCircleIcon
className={cn("size-3 text-green-600", props.iconClassName)}
/>
)}
{props.state === "down" && (
<ChevronDownCircleIcon
className={cn("size-3 text-red-600", props.iconClassName)}
/>
)}
{props.state === "same" && (
<ChevronRightCircleIcon
className={cn("size-3 text-blue-500", props.iconClassName)}
/>
)}
{props.state === "first" && (
<ChevronRightCircleIcon
className={cn("size-3 text-blue-500", props.iconClassName)}
/>
)}
</>
)}
{(() => {
switch (true) {
case props.isRunning:
return (
<LoaderCircleIcon
className={cn(
"size-3 text-blue-500 animate-spin",
props.iconClassName
)}
/>
);
case props.evalStatus === "fail":
return (
<XCircleIcon
className={cn("size-3 text-red-500", props.iconClassName)}
/>
);
case props.state === "up":
return (
<ChevronUpCircleIcon
className={cn("size-3 text-green-600", props.iconClassName)}
/>
);
case props.state === "down":
return (
<ChevronDownCircleIcon
className={cn("size-3 text-red-600", props.iconClassName)}
/>
);
case props.state === "same":
return (
<ChevronRightCircleIcon
className={cn("size-3 text-blue-500", props.iconClassName)}
/>
);
case props.state === "first":
return (
<ChevronRightCircleIcon
className={cn("size-3 text-blue-500", props.iconClassName)}
/>
);
default:
return null;
}
})()}
</span>
);
};
Expand Down
14 changes: 12 additions & 2 deletions apps/evalite-ui/app/root.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,10 +62,11 @@ export function Layout({ children }: { children: React.ReactNode }) {
}

export const clientLoader = async () => {
const { archivedEvals, currentEvals, prevScore, score } =
const { archivedEvals, currentEvals, prevScore, score, evalStatus } =
await getMenuItems();

return {
evalStatus,
prevScore,
score,
archivedEvals: archivedEvals.map((e) => {
Expand Down Expand Up @@ -118,6 +119,7 @@ export default function App() {
score={data.score}
state={getScoreState(data.score, data.prevScore)}
iconClassName="size-4"
evalStatus={data.evalStatus}
/>
</div>
</div>
Expand All @@ -133,6 +135,7 @@ export default function App() {
name={e.name}
score={e.score}
state={e.state}
evalStatus={e.evalStatus}
/>
);
})}
Expand All @@ -150,6 +153,7 @@ export default function App() {
name={e.name}
score={e.score}
state={e.state}
evalStatus={e.evalStatus}
/>
);
})}
Expand All @@ -169,6 +173,7 @@ const SidebarItem = (props: {
name: string;
state: ScoreState;
score: number;
evalStatus: "success" | "fail";
}) => {
let isRunning = false;

Expand All @@ -191,7 +196,12 @@ const SidebarItem = (props: {
>
<span>{props.name}</span>

<Score score={props.score} state={props.state} isRunning={isRunning} />
<Score
score={props.score}
state={props.state}
isRunning={isRunning}
evalStatus={props.evalStatus}
/>
</NavLink>
</SidebarMenuItem>
);
Expand Down
4 changes: 3 additions & 1 deletion apps/evalite-ui/app/routes/eval.$name.trace.$index.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -55,7 +55,7 @@ export const clientLoader = async (args: ClientLoaderFunctionArgs) => {

export default function Page() {
const {
data: { result, prevResult, filepath },
data: { result, prevResult, filepath, evalStatus },
name,
resultIndex,
} = useLoaderData<typeof clientLoader>();
Expand Down Expand Up @@ -108,6 +108,7 @@ export default function Page() {
isRunning={isRunning}
score={result.score}
state={getScoreState(result.score, prevResult?.score)}
evalStatus={evalStatus}
/>
</BreadcrumbItem>
<Separator orientation="vertical" className="mx-1 h-4" />
Expand Down Expand Up @@ -223,6 +224,7 @@ export default function Page() {
(prevScore) => prevScore.name === score.name
)?.score
)}
evalStatus={evalStatus}
/>
</MainBodySection>
{score.metadata && (
Expand Down
184 changes: 101 additions & 83 deletions apps/evalite-ui/app/routes/eval.$name.tsx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ import {
useLoaderData,
type ClientLoaderFunctionArgs,
} from "@remix-run/react";
import { XCircleIcon } from "lucide-react";
import React, { useContext } from "react";
import { DisplayInput } from "~/components/display-input";
import { InnerPageLayout } from "~/components/page-layout";
Expand Down Expand Up @@ -58,96 +59,113 @@ export default function Page() {
filepath={evaluation.filepath.split(/(\/|\\)/).slice(-1)[0]!}
>
{/* {history.length > 1 && <MyLineChart data={history} />} */}
<Table>
<TableHeader>
<TableRow>
<TableHead>Input</TableHead>
<TableHead>Output</TableHead>
{showExpectedColumn && <TableHead>Expected</TableHead>}
{firstResult?.scores.map((scorer, index) => (
<TableHead
key={scorer.name}
className={cn(index === 0 && "border-l")}
>
{scorer.name}
</TableHead>
))}
</TableRow>
</TableHeader>
<TableBody>
{evaluation.results.map((result, index) => {
const Wrapper = (props: { children: React.ReactNode }) => (
<NavLink
prefetch="intent"
to={`trace/${index}`}
preventScrollReset
className={({ isActive }) => {
return cn("block h-full p-4", isActive && "active");
}}
>
{props.children}
</NavLink>
);
return (
<TableRow
key={JSON.stringify(result.input)}
className="has-[.active]:bg-gray-100"
>
<td className="align-top">
<DisplayInput
input={result.input}
shouldTruncateText
Wrapper={Wrapper}
/>
</td>
<td className="align-top">
<DisplayInput
input={result.output}
shouldTruncateText
Wrapper={Wrapper}
/>
</td>
{showExpectedColumn && (
{evaluation.status === "fail" && (
<div className="flex gap-4 px-4">
<div className="flex-shrink-0">
<XCircleIcon className="text-red-500 size-7" />
</div>
<div className="text-sm text-gray-600 gap-1 flex flex-col">
<h3 className="font-semibold text-gray-700 mb-1 text-lg">
Evaluation Failed
</h3>
<p>This is likely because of an exception from your code.</p>
<p>Check the terminal for more information.</p>
</div>
</div>
)}
{evaluation.status === "success" && (
<Table>
<TableHeader>
<TableRow>
<TableHead>Input</TableHead>
<TableHead>Output</TableHead>
{showExpectedColumn && <TableHead>Expected</TableHead>}
{firstResult?.scores.map((scorer, index) => (
<TableHead
key={scorer.name}
className={cn(index === 0 && "border-l")}
>
{scorer.name}
</TableHead>
))}
</TableRow>
</TableHeader>
<TableBody>
{evaluation.results.map((result, index) => {
const Wrapper = (props: { children: React.ReactNode }) => (
<NavLink
prefetch="intent"
to={`trace/${index}`}
preventScrollReset
className={({ isActive }) => {
return cn("block h-full p-4", isActive && "active");
}}
>
{props.children}
</NavLink>
);
return (
<TableRow
key={JSON.stringify(result.input)}
className="has-[.active]:bg-gray-100"
>
<td className="align-top">
<DisplayInput
input={result.input}
shouldTruncateText
Wrapper={Wrapper}
/>
</td>
<td className="align-top">
<DisplayInput
input={result.expected}
input={result.output}
shouldTruncateText
Wrapper={Wrapper}
/>
</td>
)}
{result.scores.map((scorer, index) => {
const scoreInPreviousEvaluation = prevEvaluation?.results
.find((r) => r.input === result.input)
?.scores.find((s) => s.name === scorer.name);
return (
<td
key={scorer.name}
className={cn(index === 0 && "border-l", "align-top")}
>
<Wrapper>
<Score
score={scorer.score}
isRunning={
serverState.state.type === "running" &&
serverState.state.filepaths.has(
evaluation.filepath
)
}
state={getScoreState(
scorer.score,
scoreInPreviousEvaluation?.score
)}
/>
</Wrapper>
{showExpectedColumn && (
<td className="align-top">
<DisplayInput
input={result.expected}
shouldTruncateText
Wrapper={Wrapper}
/>
</td>
);
})}
</TableRow>
);
})}
</TableBody>
</Table>
)}
{result.scores.map((scorer, index) => {
const scoreInPreviousEvaluation = prevEvaluation?.results
.find((r) => r.input === result.input)
?.scores.find((s) => s.name === scorer.name);
return (
<td
key={scorer.name}
className={cn(index === 0 && "border-l", "align-top")}
>
<Wrapper>
<Score
score={scorer.score}
isRunning={
serverState.state.type === "running" &&
serverState.state.filepaths.has(
evaluation.filepath
)
}
evalStatus={evaluation.status}
state={getScoreState(
scorer.score,
scoreInPreviousEvaluation?.score
)}
/>
</Wrapper>
</td>
);
})}
</TableRow>
);
})}
</TableBody>
</Table>
)}
</InnerPageLayout>
<div
className={cn(
Expand Down
Loading
Loading