opensource/coder
3
0
Fork 0
mirror of https://github.com/coder/coder.git synced 2026-09-22 05:05:20 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: refine chat retry status UX (#23651)

Browse Source 
Follow-up to #23282. The retry and terminal error callouts had a few UX
oddities:

- Auto-retrying states reused backend error text that said "Please try
again" even while the UI was already retrying on behalf of the user.
- Terminal error states also said "Please try again" with no action the
user could take.
- `startup_timeout` had no specific title or retry copy — it fell
through to the generic "Retrying request" heading.
- The kind pill showed raw enum values like `startup_timeout` and
`rate_limit`.
- Terminal error metadata showed a "Retryable" / "Not retryable" label
that does not help users.
- A separate "Provider anthropic" metadata row duplicated information
already present in the message body.
- The `usage-limit` error kind used a hyphen while every backend kind
uses underscores.

Changes:

**Backend (`chaterror/message.go`)**

- Split message generation into `terminalMessage()` and
`retryMessage()`, replacing the old `userFacingMessage()`.
- Terminal messages include HTTP status codes and actionable guidance
(e.g. "Check the API key, permissions, and billing settings.").
- Retry messages are clean factual statements without status codes or
remediation, suitable for the retry countdown UI (e.g. "Anthropic is
temporarily overloaded.").
- Removed "Please try again" / "Please try again later" from all paths.
- `StreamRetryPayload` calls `retryMessage()` instead of forwarding
`classified.Message`.

**Frontend**

- Removed the parallel frontend message-generation system:
`getRetryMessage()`, `getProviderDisplayName()`,
`getRetryProviderSubject()`, and the `PROVIDER_DISPLAY_NAMES` map are
all deleted from `chatStatusHelpers.ts`.
- `liveStatusModel.ts` passes `retryState.error` through directly — the
backend owns the copy.
- Added specific title and retry copy for `startup_timeout`, and
extended the title mapping to cover `auth` and `config`.
- Kind pills now show humanized labels ("Startup timeout", "Rate limit",
etc.) instead of raw enum strings.
- Removed the redundant "Provider anthropic" metadata row.
- Removed the terminal "Retryable" / "Not retryable" badge.
- Normalized `"usage-limit"` → `"usage_limit"` and added it to
`ChatProviderFailureKind` so all error kinds follow the same underscore
convention and live in one enum.

Refs #23282.
This commit is contained in:
Ethan
2026-03-26 17:37:27 +11:00
committed by GitHub
parent 411714cd73
commit 21c2acbad5
20 changed files with 335 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files
coderd/x/chatd/chatd_internal_test.go
+2 -2
View File
@@ -732,7 +732,7 @@ func TestSubscribeDeliversRetryEventViaPubsubOnce(t *testing.T) {
expected := &codersdk.ChatStreamRetry{
Attempt: 1,
DelayMs: (1500 * time.Millisecond).Milliseconds(),
Error: "OpenAI is rate limiting requests (HTTP 429). Please try again later.",
Error: "OpenAI is rate limiting requests (HTTP 429).",
Kind: chaterror.KindRateLimit,
Provider: "openai",
StatusCode: 429,
@@ -773,7 +773,7 @@ func TestSubscribePrefersStructuredErrorPayloadViaPubsub(t *testing.T) {
defer cancel()
classified := chaterror.ClassifiedError{
Message: "OpenAI is rate limiting requests (HTTP 429). Please try again later.",
Message: "OpenAI is rate limiting requests (HTTP 429).",
Kind: chaterror.KindRateLimit,
Provider: "openai",
Retryable: true,
coderd/x/chatd/chaterror/classify.go
+1 -1
View File
@@ -178,7 +178,7 @@ func normalizeClassification(classified ClassifiedError) ClassifiedError {
classified.Kind = KindGeneric
}
if classified.Message == "" {
classified.Message = userFacingMessage(classified)
classified.Message = terminalMessage(classified)
}
return classified
}
coderd/x/chatd/chaterror/classify_test.go
+9 -9
View File
@@ -22,7 +22,7 @@ func TestClassify(t *testing.T) {
name: "AmbiguousOverloadKeepsProviderUnknown",
err: xerrors.New("status 529 from upstream"),
want: chaterror.ClassifiedError{
Message: "The AI provider is temporarily overloaded (HTTP 529). Please try again later.",
Message: "The AI provider is temporarily overloaded (HTTP 529).",
Kind: chaterror.KindOverloaded,
Provider: "",
Retryable: true,
@@ -33,7 +33,7 @@ func TestClassify(t *testing.T) {
name: "ExplicitAnthropicOverload",
err: xerrors.New("anthropic overloaded_error"),
want: chaterror.ClassifiedError{
Message: "Anthropic is temporarily overloaded. Please try again later.",
Message: "Anthropic is temporarily overloaded.",
Kind: chaterror.KindOverloaded,
Provider: "anthropic",
Retryable: true,
@@ -110,7 +110,7 @@ func TestClassify(t *testing.T) {
name: "ExplicitStatus429ClassifiesAsRateLimit",
err: xerrors.New("status 429 from upstream"),
want: chaterror.ClassifiedError{
Message: "The AI provider is rate limiting requests (HTTP 429). Please try again later.",
Message: "The AI provider is rate limiting requests (HTTP 429).",
Kind: chaterror.KindRateLimit,
Provider: "",
Retryable: true,
@@ -132,7 +132,7 @@ func TestClassify(t *testing.T) {
name: "ServiceUnavailableClassifiesAsRetryableTimeout",
err: xerrors.New("service unavailable"),
want: chaterror.ClassifiedError{
Message: "The AI provider is temporarily unavailable. Please try again later.",
Message: "The AI provider is temporarily unavailable.",
Kind: chaterror.KindTimeout,
Provider: "",
Retryable: true,
@@ -176,7 +176,7 @@ func TestClassify(t *testing.T) {
name: "DeadlineExceededStaysNonRetryableTimeout",
err: context.DeadlineExceeded,
want: chaterror.ClassifiedError{
Message: "The request timed out before it completed. Please try again.",
Message: "The request timed out before it completed.",
Kind: chaterror.KindTimeout,
Provider: "",
Retryable: false,
@@ -279,7 +279,7 @@ func TestClassify_TransportFailuresUseBroaderRetryMessage(t *testing.T) {
require.True(t, classified.Retryable)
require.Equal(
t,
"The AI provider is temporarily unavailable. Please try again later.",
"The AI provider is temporarily unavailable.",
classified.Message,
)
})
@@ -299,7 +299,7 @@ func TestClassify_StartupTimeoutWrappedClassificationWins(t *testing.T) {
)
require.Equal(t, chaterror.ClassifiedError{
Message: "OpenAI did not start responding in time. Please try again.",
Message: "OpenAI did not start responding in time.",
Kind: chaterror.KindStartupTimeout,
Provider: "openai",
Retryable: true,
@@ -315,7 +315,7 @@ func TestWithProviderUsesExplicitHint(t *testing.T) {
enriched := classified.WithProvider("azure openai")
require.Equal(t, chaterror.ClassifiedError{
Message: "Azure OpenAI is rate limiting requests (HTTP 429). Please try again later.",
Message: "Azure OpenAI is rate limiting requests (HTTP 429).",
Kind: chaterror.KindRateLimit,
Provider: "azure",
Retryable: true,
@@ -331,7 +331,7 @@ func TestWithProviderAddsProviderWhenUnknown(t *testing.T) {
enriched := classified.WithProvider("openai")
require.Equal(t, chaterror.ClassifiedError{
Message: "OpenAI is rate limiting requests (HTTP 429). Please try again later.",
Message: "OpenAI is rate limiting requests (HTTP 429).",
Kind: chaterror.KindRateLimit,
Provider: "openai",
Retryable: true,
coderd/x/chatd/chaterror/message.go
+80 -45
View File
@@ -5,78 +5,113 @@ import (
"strings"
)
func userFacingMessage(classified ClassifiedError) string {
// terminalMessage produces the user-facing error description shown
// when retries are exhausted. It includes HTTP status codes and
// actionable remediation guidance.
func terminalMessage(classified ClassifiedError) string {
subject := providerSubject(classified.Provider)
switch classified.Kind {
case KindOverloaded:
return optionalStatusMessage(
subject,
classified.StatusCode,
"%s is temporarily overloaded (HTTP %d). Please try again later.",
"%s is temporarily overloaded. Please try again later.",
)
if classified.StatusCode > 0 {
return fmt.Sprintf(
"%s is temporarily overloaded (HTTP %d).",
subject, classified.StatusCode,
)
}
return fmt.Sprintf("%s is temporarily overloaded.", subject)
case KindRateLimit:
return optionalStatusMessage(
subject,
classified.StatusCode,
"%s is rate limiting requests (HTTP %d). Please try again later.",
"%s is rate limiting requests. Please try again later.",
)
if classified.StatusCode > 0 {
return fmt.Sprintf(
"%s is rate limiting requests (HTTP %d).",
subject, classified.StatusCode,
)
}
return fmt.Sprintf("%s is rate limiting requests.", subject)
case KindTimeout:
if classified.StatusCode > 0 {
return fmt.Sprintf(
"%s is temporarily unavailable (HTTP %d). Please try again later.",
subject,
classified.StatusCode,
"%s is temporarily unavailable (HTTP %d).",
subject, classified.StatusCode,
)
}
if classified.Retryable {
return fmt.Sprintf("%s is temporarily unavailable. Please try again later.", subject)
if !classified.Retryable {
return "The request timed out before it completed."
}
return "The request timed out before it completed. Please try again."
return fmt.Sprintf("%s is temporarily unavailable.", subject)
case KindStartupTimeout:
return fmt.Sprintf("%s did not start responding in time. Please try again.", subject)
return fmt.Sprintf(
"%s did not start responding in time.", subject,
)
case KindAuth:
if displayName := providerDisplayName(classified.Provider); displayName != "" {
return fmt.Sprintf(
"Authentication with %s failed. Check the API key, permissions, and billing settings.",
displayName,
)
displayName := providerDisplayName(classified.Provider)
if displayName == "" {
displayName = "the AI provider"
}
return "Authentication with the AI provider failed. Check the API key, permissions, and billing settings."
return fmt.Sprintf(
"Authentication with %s failed."+
" Check the API key, permissions, and billing settings.",
displayName,
)
case KindConfig:
return fmt.Sprintf(
"%s rejected the model configuration. Check the selected model and provider settings.",
"%s rejected the model configuration."+
" Check the selected model and provider settings.",
subject,
)
default:
if classified.StatusCode > 0 {
suffix := " Please try again."
if classified.Retryable {
suffix = " Please try again later."
}
return fmt.Sprintf(
"%s returned an unexpected error (HTTP %d).%s",
subject,
classified.StatusCode,
suffix,
"%s returned an unexpected error (HTTP %d).",
subject, classified.StatusCode,
)
}
if classified.Retryable {
return fmt.Sprintf(
"%s returned an unexpected error. Please try again later.",
subject,
)
if !classified.Retryable {
return "The chat request failed unexpectedly."
}
return "The chat request failed unexpectedly. Please try again."
return fmt.Sprintf("%s returned an unexpected error.", subject)
}
}
func optionalStatusMessage(subject string, statusCode int, withStatus string, withoutStatus string) string {
if statusCode > 0 {
return fmt.Sprintf(withStatus, subject, statusCode)
// retryMessage produces a clean factual description suitable for
// display alongside the retry countdown UI. It omits HTTP status
// codes (surfaced separately in the payload) and remediation
// guidance (not actionable while auto-retrying).
func retryMessage(classified ClassifiedError) string {
subject := providerSubject(classified.Provider)
switch classified.Kind {
case KindOverloaded:
return fmt.Sprintf("%s is temporarily overloaded.", subject)
case KindRateLimit:
return fmt.Sprintf("%s is rate limiting requests.", subject)
case KindTimeout:
return fmt.Sprintf("%s is temporarily unavailable.", subject)
case KindStartupTimeout:
return fmt.Sprintf(
"%s did not start responding in time.", subject,
)
case KindAuth:
displayName := providerDisplayName(classified.Provider)
if displayName == "" {
displayName = "the AI provider"
}
return fmt.Sprintf(
"Authentication with %s failed.", displayName,
)
case KindConfig:
return fmt.Sprintf(
"%s rejected the model configuration.", subject,
)
default:
return fmt.Sprintf(
"%s returned an unexpected error.", subject,
)
}
return fmt.Sprintf(withoutStatus, subject)
}
func providerSubject(provider string) string {
coderd/x/chatd/chaterror/payload.go
+1 -1
View File
@@ -30,7 +30,7 @@ func StreamRetryPayload(
return &codersdk.ChatStreamRetry{
Attempt: attempt,
DelayMs: delay.Milliseconds(),
Error: classified.Message,
Error: retryMessage(classified),
Kind: classified.Kind,
Provider: classified.Provider,
StatusCode: classified.StatusCode,
coderd/x/chatd/chaterror/payload_test.go
+5 -3
View File
@@ -20,7 +20,7 @@ func TestStreamErrorPayloadUsesNormalizedClassification(t *testing.T) {
payload := chaterror.StreamErrorPayload(classified)
require.Equal(t, &codersdk.ChatStreamError{
Message: "Azure OpenAI is rate limiting requests (HTTP 429). Please try again later.",
Message: "Azure OpenAI is rate limiting requests (HTTP 429).",
Kind: chaterror.KindRateLimit,
Provider: "azure",
Retryable: true,
@@ -40,7 +40,7 @@ func TestStreamRetryPayloadUsesNormalizedClassification(t *testing.T) {
delay := 3 * time.Second
startedAt := time.Now()
payload := chaterror.StreamRetryPayload(2, delay, chaterror.ClassifiedError{
Message: "retry me",
Message: "OpenAI returned an unexpected error (HTTP 503).",
Kind: chaterror.KindGeneric,
Provider: "openai",
Retryable: true,
@@ -50,7 +50,9 @@ func TestStreamRetryPayloadUsesNormalizedClassification(t *testing.T) {
require.NotNil(t, payload)
require.Equal(t, 2, payload.Attempt)
require.Equal(t, delay.Milliseconds(), payload.DelayMs)
require.Equal(t, "retry me", payload.Error)
// Retry messages omit the HTTP status code; the status code is
// surfaced separately in the payload's StatusCode field.
require.Equal(t, "OpenAI returned an unexpected error.", payload.Error)
require.Equal(t, chaterror.KindGeneric, payload.Kind)
require.Equal(t, "openai", payload.Provider)
require.Equal(t, 503, payload.StatusCode)
coderd/x/chatd/chatloop/chatloop_test.go
+4 -4
View File
@@ -144,7 +144,7 @@ func TestRun_OnRetryEnrichesProvider(t *testing.T) {
require.Equal(t, 429, records[0].classified.StatusCode)
require.Equal(
t,
"OpenAI is rate limiting requests (HTTP 429). Please try again later.",
"OpenAI is rate limiting requests (HTTP 429).",
records[0].classified.Message,
)
}
@@ -254,7 +254,7 @@ func TestRun_RetriesStartupTimeoutWhileOpeningStream(t *testing.T) {
require.Equal(t, "openai", retries[0].Provider)
require.Equal(
t,
"OpenAI did not start responding in time. Please try again.",
"OpenAI did not start responding in time.",
retries[0].Message,
)
require.ErrorIs(t, <-attemptCause, errStartupTimeout)
@@ -313,7 +313,7 @@ func TestRun_RetriesStartupTimeoutBeforeFirstPart(t *testing.T) {
require.Equal(t, "openai", retries[0].Provider)
require.Equal(
t,
"OpenAI did not start responding in time. Please try again.",
"OpenAI did not start responding in time.",
retries[0].Message,
)
require.ErrorIs(t, <-attemptCause, errStartupTimeout)
@@ -475,7 +475,7 @@ func TestRun_RetriesStartupTimeoutWhenStreamClosesSilently(t *testing.T) {
require.Equal(t, "openai", retries[0].Provider)
require.Equal(
t,
"OpenAI did not start responding in time. Please try again.",
"OpenAI did not start responding in time.",
retries[0].Message,
)
require.ErrorIs(t, <-attemptCause, errStartupTimeout)
site/src/pages/AgentsPage/AgentDetail.tsx
+2 -2
View File
@@ -247,7 +247,7 @@ const getPersistedDetailError = ({
chatRecord: TypesGen.Chat | undefined;
cachedError: ChatDetailError | undefined;
}): ChatDetailError | undefined => {
if (cachedError?.kind === "usage-limit") {
if (cachedError?.kind === "usage_limit") {
return cachedError;
}
if (chatStatus === "error") {
@@ -595,7 +595,7 @@ const AgentDetail: FC = () => {
isUsageLimitData(error.response.data)
) {
const reason: ChatDetailError = {
kind: "usage-limit",
kind: "usage_limit",
message: formatUsageLimitMessage(error.response.data),
};
store.setStreamError(reason);
site/src/pages/AgentsPage/components/AgentDetail/ChatStatusCallout.tsx
+4 -9
View File
@@ -4,7 +4,7 @@ import { Button } from "components/Button/Button";
import { Pill } from "components/Pill/Pill";
import { ExternalLinkIcon } from "lucide-react";
import { type FC, useEffect, useState } from "react";
import { getProviderStatusURL } from "./chatStatusHelpers";
import { getKindLabel, getProviderStatusURL } from "./chatStatusHelpers";
import type { LiveStatusModel } from "./liveStatusModel";
const RESPONSE_STARTUP_GRACE_MS = 15_000;
@@ -130,9 +130,7 @@ const StatusAlert: FC<{ status: RetryOrFailedStatus }> = ({ status }) => {
: "warning";
const hasMetadata =
status.phase === "retrying" ||
status.provider !== undefined ||
(status.phase === "failed" && status.statusCode !== undefined) ||
(status.phase === "failed" && status.retryable !== undefined);
(status.phase === "failed" && status.statusCode !== undefined);
return (
<Alert
@@ -156,7 +154,7 @@ const StatusAlert: FC<{ status: RetryOrFailedStatus }> = ({ status }) => {
className="h-5 px-2.5 text-[10px] font-semibold"
type={pillType}
>
{status.kind}
{getKindLabel(status.kind)}
</Pill>
</div>
<AlertDescription>{status.message}</AlertDescription>
@@ -171,13 +169,10 @@ const StatusAlert: FC<{ status: RetryOrFailedStatus }> = ({ status }) => {
{status.phase === "retrying" && (
<span>Attempt {status.attempt}</span>
)}
{status.provider && <span>Provider {status.provider}</span>}
{status.phase === "failed" && status.statusCode !== undefined && (
<span>HTTP {status.statusCode}</span>
)}
{status.phase === "failed" && status.retryable !== undefined && (
<span>{status.retryable ? "Retryable" : "Not retryable"}</span>
)}
</div>
)}
</div>
site/src/pages/AgentsPage/components/AgentDetail/LiveStreamTail.stories.tsx
+40 -3
View File
@@ -58,7 +58,7 @@ export const UsageLimitExceeded: Story = {
...defaultArgs,
liveStatus: buildLiveStatus({
persistedError: {
kind: "usage-limit",
kind: "usage_limit",
message:
"You've used $50.00 of your $50.00 spend limit. Your limit resets on July 1, 2025.",
},
@@ -80,7 +80,7 @@ export const TerminalOverloadedError: Story = {
liveStatus: buildLiveStatus({
persistedError: {
kind: "overloaded",
message: "Anthropic is currently overloaded. Please try again shortly.",
message: "Anthropic is currently overloaded.",
provider: "anthropic",
retryable: true,
statusCode: 529,
@@ -92,9 +92,46 @@ export const TerminalOverloadedError: Story = {
expect(
canvas.getByRole("heading", { name: /service overloaded/i }),
).toBeVisible();
expect(canvas.getByText("overloaded")).toBeVisible();
expect(canvas.getByText("Overloaded")).toBeVisible();
expect(
canvas.getByText(/anthropic is currently overloaded./i),
).toBeVisible();
expect(canvas.queryByText(/please try again/i)).not.toBeInTheDocument();
expect(canvas.queryByText(/^retryable$/i)).not.toBeInTheDocument();
expect(canvas.getByText(/http 529/i)).toBeVisible();
expect(canvas.getByRole("link", { name: /status/i })).toBeVisible();
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
/** Terminal startup timeouts get a specific heading without provider metadata. */
export const TerminalStartupTimeoutError: Story = {
args: {
...defaultArgs,
liveStatus: buildLiveStatus({
persistedError: {
kind: "startup_timeout",
message: "Anthropic did not start responding in time.",
provider: "anthropic",
retryable: true,
},
}),
},
play: async ({ canvasElement }) => {
const canvas = within(canvasElement);
expect(
canvas.getByRole("heading", { name: /startup timed out/i }),
).toBeVisible();
expect(canvas.getByText("Startup timeout")).toBeVisible();
expect(
canvas.getByText(/anthropic did not start responding in time./i),
).toBeVisible();
expect(canvas.queryByText(/please try again/i)).not.toBeInTheDocument();
expect(canvas.queryByText(/^retryable$/i)).not.toBeInTheDocument();
expect(
canvas.queryByRole("link", { name: /status/i }),
).not.toBeInTheDocument();
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
site/src/pages/AgentsPage/components/AgentDetail/LiveStreamTail.tsx
+1 -1
View File
@@ -55,7 +55,7 @@ export const LiveStreamTailContent = ({
const shouldRenderStreamSection = shouldRenderStreamingSection(liveStatus);
const terminalStatus = liveStatus.phase === "failed" ? liveStatus : null;
const usageLimitStatus =
terminalStatus?.kind === "usage-limit" ? terminalStatus : null;
terminalStatus?.kind === "usage_limit" ? terminalStatus : null;
const shouldRenderEmptyState =
isTranscriptEmpty && liveStatus.phase === "idle";
site/src/pages/AgentsPage/components/AgentDetail/StreamingOutput.stories.tsx
+64 -14
View File
@@ -71,13 +71,13 @@ export const ReconnectingAfterDisconnect: Story = {
await waitFor(() => {
expect(canvasElement.textContent).toMatch(/reconnecting in \d+s/i);
});
expect(canvas.queryByText("generic")).not.toBeInTheDocument();
expect(canvas.queryByText("Unexpected error")).not.toBeInTheDocument();
const thinkingMatches = canvas.getAllByText(/thinking\.\.\./i);
expect(thinkingMatches.length).toBeGreaterThanOrEqual(1);
},
};
/** Generic retry reasons show the mux-style retry callout. */
/** Generic retry reasons show automatic retry copy without a manual CTA. */
export const RetryWithVisibleReason: Story = {
args: {
streamState: null,
@@ -92,9 +92,13 @@ export const RetryWithVisibleReason: Story = {
expect(
canvas.getByRole("heading", { name: /retrying request/i }),
).toBeVisible();
expect(canvas.getByText(/transient upstream failure/i)).toBeVisible();
expect(canvas.getByText("generic")).toBeVisible();
expect(
canvas.getByText(/anthropic returned an unexpected error/i),
).toBeVisible();
expect(canvas.getByText("Unexpected error")).toBeVisible();
expect(canvas.getByText(/attempt 1/i)).toBeVisible();
expect(canvas.queryByText(/please try again/i)).not.toBeInTheDocument();
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
@@ -106,7 +110,7 @@ export const RetryRateLimited: Story = {
liveStatus: buildLiveStatus({
retryState: buildRetryState({
attempt: 3,
error: "Anthropic asked us to back off briefly before retrying.",
error: "Anthropic is rate limiting requests.",
kind: "rate_limit",
delayMs: 3000,
}),
@@ -118,13 +122,17 @@ export const RetryRateLimited: Story = {
expect(
canvas.getByRole("heading", { name: /rate limited/i }),
).toBeVisible();
expect(canvas.getByText("rate_limit")).toBeVisible();
expect(
canvas.getByText(/anthropic is rate limiting requests/i),
).toBeVisible();
expect(canvas.getByText("Rate limit")).toBeVisible();
await waitFor(() => {
expect(canvasElement.textContent).toMatch(/retrying in \d+s/i);
});
expect(
canvas.queryByRole("link", { name: /status/i }),
).not.toBeInTheDocument();
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
@@ -136,7 +144,7 @@ export const RetryInvalidTimestamp: Story = {
liveStatus: buildLiveStatus({
retryState: buildRetryState({
attempt: 3,
error: "Anthropic asked us to back off briefly before retrying.",
error: "Anthropic is rate limiting requests.",
kind: "rate_limit",
delayMs: 3000,
retryingAt: "not-a-date",
@@ -149,12 +157,16 @@ export const RetryInvalidTimestamp: Story = {
expect(
canvas.getByRole("heading", { name: /rate limited/i }),
).toBeVisible();
expect(canvas.getByText("rate_limit")).toBeVisible();
expect(
canvas.getByText(/anthropic is rate limiting requests/i),
).toBeVisible();
expect(canvas.getByText("Rate limit")).toBeVisible();
expect(canvas.getByText(/attempt 3/i)).toBeVisible();
await waitFor(() => {
expect(canvas.queryByText(/retrying in nan/i)).not.toBeInTheDocument();
expect(canvas.queryByText(/retrying in \d+s/i)).not.toBeInTheDocument();
});
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
@@ -167,7 +179,7 @@ export const RetryOverloaded: Story = {
retryState: buildRetryState({
kind: "overloaded",
provider: "anthropic",
error: "Anthropic is currently overloaded. Retrying your request.",
error: "Anthropic is temporarily overloaded.",
}),
isAwaitingFirstStreamChunk: true,
}),
@@ -177,14 +189,18 @@ export const RetryOverloaded: Story = {
expect(
canvas.getByRole("heading", { name: /service overloaded/i }),
).toBeVisible();
expect(canvas.getByText("overloaded")).toBeVisible();
expect(
canvas.getByText(/anthropic is temporarily overloaded/i),
).toBeVisible();
expect(canvas.getByText("Overloaded")).toBeVisible();
const statusLink = screen.getByRole("link", { name: /status/i });
expect(statusLink).toBeVisible();
expect(statusLink).toHaveAttribute("href", "https://status.anthropic.com");
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
/** Timeout retries render the timeout-specific heading without a status CTA. */
/** Timeout retries render timeout-specific copy without a status CTA. */
export const RetryTimeout: Story = {
args: {
streamState: null,
@@ -192,7 +208,7 @@ export const RetryTimeout: Story = {
liveStatus: buildLiveStatus({
retryState: buildRetryState({
kind: "timeout",
error: "The provider took too long to respond. Retrying now.",
error: "Anthropic is temporarily unavailable.",
}),
isAwaitingFirstStreamChunk: true,
}),
@@ -200,9 +216,43 @@ export const RetryTimeout: Story = {
play: async ({ canvasElement }) => {
const canvas = within(canvasElement);
expect(
canvas.getByRole("heading", { name: /request time(?:out|d out)/i }),
canvas.getByRole("heading", { name: /request timed out/i }),
).toBeVisible();
expect(canvas.getByText("timeout")).toBeVisible();
expect(
canvas.getByText(/anthropic is temporarily unavailable/i),
).toBeVisible();
expect(canvas.getByText("Timeout")).toBeVisible();
expect(
canvas.queryByRole("link", { name: /status/i }),
).not.toBeInTheDocument();
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
},
};
/** Startup timeouts explain the first-token delay before retrying. */
export const RetryStartupTimeout: Story = {
args: {
streamState: null,
streamTools: [],
liveStatus: buildLiveStatus({
retryState: buildRetryState({
kind: "startup_timeout",
error: "Anthropic did not start responding in time.",
}),
isAwaitingFirstStreamChunk: true,
}),
},
play: async ({ canvasElement }) => {
const canvas = within(canvasElement);
expect(
canvas.getByRole("heading", { name: /startup timed out/i }),
).toBeVisible();
expect(
canvas.getByText(/anthropic did not start responding in time/i),
).toBeVisible();
expect(canvas.getByText("Startup timeout")).toBeVisible();
expect(canvas.queryByText(/please try again/i)).not.toBeInTheDocument();
expect(canvas.queryByText(/provider anthropic/i)).not.toBeInTheDocument();
expect(
canvas.queryByRole("link", { name: /status/i }),
).not.toBeInTheDocument();
site/src/pages/AgentsPage/components/AgentDetail/chatStatusHelpers.ts
+69 -5
View File
@@ -1,9 +1,43 @@
import type { ChatProviderFailureKind } from "../../utils/usageLimitMessage";
const PROVIDER_STATUS_URLS: Record<string, string> = {
anthropic: "https://status.anthropic.com",
};
const normalizeProvider = (provider?: string): string | undefined => {
const normalized = provider?.trim().toLowerCase();
if (!normalized) {
return undefined;
}
switch (normalized) {
case "azure openai":
case "azure-openai":
return "azure";
case "openai compat":
case "openai compatible":
case "openai_compat":
return "openai-compat";
default:
return normalized;
}
};
const humanizeKind = (kind: string): string => {
const words = kind
.trim()
.split(/[_\-\s]+/)
.filter(Boolean);
if (words.length === 0) {
return "Unexpected error";
}
return words
.map((word) => word.charAt(0).toUpperCase() + word.slice(1))
.join(" ");
};
export const getErrorTitle = (
kind: string,
kind: ChatProviderFailureKind | (string & {}),
mode: "retry" | "error",
): string => {
switch (kind) {
@@ -12,18 +46,48 @@ export const getErrorTitle = (
case "rate_limit":
return "Rate limited";
case "timeout":
return "Request timeout";
return "Request timed out";
case "startup_timeout":
return "Startup timed out";
case "auth":
return "Authentication failed";
case "config":
return "Configuration error";
default:
return mode === "retry" ? "Retrying request" : "Request failed";
}
};
export const getKindLabel = (
kind: ChatProviderFailureKind | (string & {}),
): string => {
switch (kind) {
case "generic":
return "Unexpected error";
case "overloaded":
return "Overloaded";
case "rate_limit":
return "Rate limit";
case "timeout":
return "Timeout";
case "startup_timeout":
return "Startup timeout";
case "auth":
return "Authentication";
case "config":
return "Configuration";
default:
return humanizeKind(kind);
}
};
export const getProviderStatusURL = (
kind: string,
kind: ChatProviderFailureKind | (string & {}),
provider?: string,
): string | undefined => {
if (!provider || kind !== "overloaded") {
if (kind !== "overloaded") {
return undefined;
}
return PROVIDER_STATUS_URLS[provider.toLowerCase()];
const normalized = normalizeProvider(provider);
return normalized ? PROVIDER_STATUS_URLS[normalized] : undefined;
};
site/src/pages/AgentsPage/components/AgentDetail/liveStatusModel.test.ts
+2 -3
View File
@@ -15,7 +15,7 @@ const makeStreamState = (
const makeRetryState = (overrides: Partial<RetryState> = {}): RetryState => ({
attempt: 2,
error: "Retrying request shortly.",
error: "Anthropic returned an unexpected error.",
kind: "generic",
provider: "anthropic",
delayMs: 2000,
@@ -62,7 +62,7 @@ describe("deriveLiveStatus", () => {
hasAccumulatedOutput: false,
title: "Retrying request",
kind: "generic",
message: "Retrying request shortly.",
message: "Anthropic returned an unexpected error.",
attempt: 2,
provider: "anthropic",
delayMs: 2000,
@@ -84,7 +84,6 @@ describe("deriveLiveStatus", () => {
kind: "generic",
message: "Chat processing failed.",
provider: "anthropic",
retryable: false,
statusCode: 500,
};
site/src/pages/AgentsPage/components/AgentDetail/liveStatusModel.ts
+16 -13
View File
@@ -37,7 +37,6 @@ export type LiveStatusModel =
kind: string;
message: string;
provider?: string;
retryable?: boolean;
statusCode?: number;
} & LiveStatusBase);
@@ -64,6 +63,21 @@ const toReconnectingLiveStatus = (
...reconnectState,
});
const toRetryingLiveStatus = (
retryState: RetryState,
options: { hasAccumulatedOutput?: boolean } = {},
): Extract<LiveStatusModel, { phase: "retrying" }> => ({
phase: "retrying",
hasAccumulatedOutput: options.hasAccumulatedOutput ?? false,
title: getErrorTitle(retryState.kind, "retry"),
kind: retryState.kind,
message: retryState.error,
attempt: retryState.attempt,
provider: retryState.provider,
delayMs: retryState.delayMs,
retryingAt: retryState.retryingAt,
});
const toFailedLiveStatus = (
error: ChatDetailError,
options: { hasAccumulatedOutput?: boolean } = {},
@@ -74,7 +88,6 @@ const toFailedLiveStatus = (
kind: error.kind,
message: error.message,
provider: error.provider,
retryable: error.retryable,
statusCode: error.statusCode,
});
@@ -89,17 +102,7 @@ export const deriveLiveStatus = ({
const hasAccumulatedOutput = getHasAccumulatedOutput(streamState);
if (retryState) {
return {
phase: "retrying",
hasAccumulatedOutput,
title: getErrorTitle(retryState.kind, "retry"),
kind: retryState.kind,
message: retryState.error,
attempt: retryState.attempt,
provider: retryState.provider,
delayMs: retryState.delayMs,
retryingAt: retryState.retryingAt,
};
return toRetryingLiveStatus(retryState, { hasAccumulatedOutput });
}
if (streamError) {
site/src/pages/AgentsPage/components/AgentDetail/storyFixtures.ts
+1 -2
View File
@@ -73,8 +73,7 @@ export const buildRetryState = (
overrides: Partial<RetryState> = {},
): RetryState => ({
attempt: 1,
error:
"Anthropic is retrying your request after a transient upstream failure.",
error: "Anthropic returned an unexpected error.",
kind: "generic",
provider: "anthropic",
delayMs: 2000,
site/src/pages/AgentsPage/components/AgentDetail/types.ts
+2 -1
View File
@@ -1,5 +1,6 @@
import type { ReconnectSchedule } from "utils/reconnectingWebSocket";
import type * as TypesGen from "#/api/typesGenerated";
import type { ChatProviderFailureKind } from "../../utils/usageLimitMessage";
export type ParsedToolCall = {
id: string;
@@ -66,7 +67,7 @@ export type ReconnectState = ReconnectSchedule;
export type RetryState = {
attempt: number;
error: string;
kind: string;
kind: ChatProviderFailureKind | (string & {});
provider?: string;
delayMs?: number;
retryingAt?: string;
site/src/pages/AgentsPage/components/AgentDetailView.stories.tsx
+13 -1
View File
@@ -206,13 +206,25 @@ export const WithError: Story = {
<StoryAgentDetailView
persistedError={{
kind: "overloaded",
message: "Anthropic is currently overloaded. Please try again shortly.",
message: "Anthropic is currently overloaded.",
provider: "anthropic",
retryable: true,
statusCode: 529,
}}
/>
),
play: async ({ canvasElement }) => {
const canvas = within(canvasElement);
expect(
canvas.getByRole("heading", { name: /service overloaded/i }),
).toBeVisible();
expect(
canvas.getByText(/anthropic is currently overloaded\./i),
).toBeVisible();
expect(canvas.queryByText(/please try again/i)).not.toBeInTheDocument();
expect(canvas.queryByText(/^retryable$/i)).not.toBeInTheDocument();
expect(canvas.getByText(/http 529/i)).toBeVisible();
},
};
/** Input area appears disabled when `isInputDisabled` is true. */
site/src/pages/AgentsPage/utils/usageLimitMessage.test.ts
+2 -2
View File
@@ -104,10 +104,10 @@ describe("isUsageLimitData", () => {
it("accepts a fully populated valid payload", () => {
const error: ChatDetailError = {
message: "Your usage limit has been reached.",
kind: "usage-limit",
kind: "usage_limit",
};
expect(error.kind).toBe("usage-limit");
expect(error.kind).toBe("usage_limit");
expect(
isUsageLimitData({
spent_micros: 900_000,
site/src/pages/AgentsPage/utils/usageLimitMessage.ts
+17 -9
View File
@@ -10,21 +10,29 @@ interface UsageLimitData {
resets_at?: string; // RFC3339
}
/**
* Known provider failure kinds surfaced in chat retry/error events.
*/
export type ChatProviderFailureKind =
| "generic"
| "overloaded"
| "rate_limit"
| "timeout"
| "startup_timeout"
| "auth"
| "config"
| "usage_limit";
/**
* Typed classification for errors surfaced in the agent detail view.
* - "usage-limit": the user hit a spending cap (409 + valid usage data).
* - "usage_limit": the user hit a spending cap (409 + valid usage data).
* - other kinds come from normalized stream/provider failures such as
* "generic", "overloaded", "rate_limit", or "timeout".
* "generic", "overloaded", "rate_limit", "timeout",
* "startup_timeout", "auth", and "config".
*/
export type ChatDetailError = {
message: string;
kind:
| "usage-limit"
| "generic"
| "overloaded"
| "rate_limit"
| "timeout"
| (string & {});
kind: ChatProviderFailureKind | (string & {});
provider?: string;
retryable?: boolean;
statusCode?: number;