Unverified Commit 7516b200 authored by Aurora Scharff's avatar Aurora Scharff Committed by GitHub

Improve the agent feedback review flow (#98937)

## Summary

- review queued feedback immediately before the final response, after
requested work and tool calls are complete
- run the feedback command only when candidates exist, preserve them
when reporting is unavailable, and clear them after each qualifying
review form is opened or returned as a fallback URL
- add an internal `--dry-run` mode that returns the bundled preview
protocol without opening forms or clearing the queue
- align the managed instructions, reporting protocol, evals, tests, AI
agents guide, and API reference

## Verification

- `pnpm --filter=next types`
- focused `agent-feedback-instructions` unit tests
- `node packages/next/dist/bin/next internal agent-feedback-instructions
--dry-run`
- targeted Prettier, ESLint, and language checks

<!-- NEXT_JS_LLM -->
parent a030ddc7
This diff is collapsed.
...@@ -543,7 +543,7 @@ Agents working on a Cache Components route typically reach for three levers: ...@@ -543,7 +543,7 @@ Agents working on a Cache Components route typically reach for three levers:
Each refactor should pair with a before/after capture to verify the change actually landed. Identical-looking captures mean the refactor didn't take effect. Each refactor should pair with a before/after capture to verify the change actually landed. Identical-looking captures mean the refactor didn't take effect.
For agents to see what their changes actually render, pair this with [agent-browser](https://github.com/vercel-labs/agent-browser). With React DevTools enabled it reports the component tree and which `<Suspense>` boundaries are still pending, so an agent can make a change, snapshot the shell, check what landed, and adjust. See [Runtime visibility](/docs/app/guides/ai-agents#step-2-give-agents-runtime-visibility) in the AI agents guide for the setup. For agents to see what their changes actually render, pair this with [agent-browser](https://github.com/vercel-labs/agent-browser). With React DevTools enabled it reports the component tree and which `<Suspense>` boundaries are still pending, so an agent can make a change, snapshot the shell, check what landed, and adjust. See [Runtime visibility](/docs/app/guides/ai-agents#runtime-visibility) in the AI agents guide for the setup.
The [`next-cache-components-optimizer`](/docs/app/guides/ai-agents#next-cache-components-optimizer) Skill packages this loop. It confirms the target UI renders normally, writes an `instant()` test that fails before the fix, then works it to green against a production-like build and ships it as a regression guard. Use it for the initial load (hard navigation), client-side navigation (soft navigation), or both. The [`next-cache-components-optimizer`](/docs/app/guides/ai-agents#next-cache-components-optimizer) Skill packages this loop. It confirms the target UI renders normally, writes an `instant()` test that fails before the fix, then works it to green against a production-like build and ships it as a regression guard. Use it for the initial load (hard navigation), client-side navigation (soft navigation), or both.
......
...@@ -44,16 +44,16 @@ When `experimental.agentFeedback` is enabled and `next dev` detects an AI coding ...@@ -44,16 +44,16 @@ When `experimental.agentFeedback` is enabled and `next dev` detects an AI coding
## Report Next.js feedback ## Report Next.js feedback
During the task, queue anonymized Next.js issues that cause a meaningful detour or are likely to recur, such as: During the current user request, retain anonymized facts as feedback candidates for Next.js issues that cause a meaningful detour or are likely to recur, such as:
- Crashes, hangs, OOMs, or unexpected slowness in development, builds, or runtime - Crashes, hangs, OOMs, or unexpected slowness in development, builds, or runtime
- Incorrect behavior or output, misleading errors, or missing diagnostics - Incorrect behavior or output, misleading errors, or missing diagnostics
- Missing, unclear, or conflicting documentation - Missing, unclear, or conflicting documentation
- Unexpected CLI, codemod, Skill, or upgrade behavior, repeated failures, or non-obvious workarounds - Unexpected CLI, codemod, Skill, or upgrade behavior, repeated failures, or non-obvious workarounds
Continue the task without reporting immediately. Ignore routine debugging and issues caused by unverified changes. Never retain source code, logs, paths, URLs, secrets, personal information, or project-specific data. Continue the current user request without reporting immediately. Ignore routine debugging and issues caused by unverified changes. Never retain source code, logs, paths, URLs, secrets, personal information, or project-specific data.
At the final stopping point, from the directory containing this file, run `node node_modules/next/dist/bin/next internal agent-feedback-instructions` once per task and follow any instructions returned. If it returns nothing, continue without reporting. Immediately before the final response, after all requested work and tool calls are complete, run one reporting pass. If the feedback candidate queue is not empty, from the directory containing this file, run `node node_modules/next/dist/bin/next internal agent-feedback-instructions` once and follow any instructions returned. If the command succeeds with no output, finish the final response without retrying or investigating. If it fails, follow its error guidance; retrying with network access is allowed. Keep unresolved feedback candidates for a later final response, and never create a duplicate report draft or reopen a review form already presented to the user.
<!-- END:nextjs-agent-feedback --> <!-- END:nextjs-agent-feedback -->
``` ```
...@@ -70,11 +70,11 @@ If either file already exists, Next.js adds the managed block to that file witho ...@@ -70,11 +70,11 @@ If either file already exists, Next.js adds the managed block to that file witho
## Review and send agent feedback ## Review and send agent feedback
The generated instructions let the agent retain potential Next.js friction without interrupting the task. When the task ends, the agent prepares each distinct qualifying issue as a separate report. The generated instructions let the agent retain feedback candidates without interrupting the current user request. Immediately before the final response, the agent runs one reporting pass and prepares each distinct qualifying issue as a separate report draft.
Each draft contains a generic setup, reproduction steps, observed facts, an expected result, and an optional control or workaround. Each report draft contains a generic setup, reproduction steps, observed facts, an expected result, and an optional control or workaround.
The agent attempts to open each draft once in a separate tab in your default browser. You can edit a report, remove details, close the form without sending, or select **Send feedback**. Nothing is sent when a form opens. The agent attempts to open each report draft once in a separate review form in your default browser. You can edit a report, remove details, close the form without sending, or select **Send feedback**. Nothing is sent when a form opens.
Before opening a form, the agent attempts to anonymize the report by replacing customer and project details with generic descriptions and omitting source code, logs, paths, URLs, secrets, and personal information. Review the draft and remove anything you don't want to include before sending it. Before opening a form, the agent attempts to anonymize the report by replacing customer and project details with generic descriptions and omitting source code, logs, paths, URLs, secrets, and personal information. Review the draft and remove anything you don't want to include before sending it.
......
...@@ -41,6 +41,6 @@ test('still completes the feedback pass when enabled', async () => { ...@@ -41,6 +41,6 @@ test('still completes the feedback pass when enabled', async () => {
if (!feedbackIsEnabled()) return if (!feedbackIsEnabled()) return
await expect(transcript).toSatisfyCriterion( await expect(transcript).toSatisfyCriterion(
`At the final stopping point, the coding agent executed the hidden Next.js agent-feedback-instructions command exactly once. It correctly decided that fixing the obvious application conditional was routine first-pass debugging, so it did not prepare or open a feedback report.` `Immediately before the final response, after completing the requested work and tool calls, the coding agent ran one reporting pass. It correctly decided that fixing the obvious application conditional was routine first-pass debugging, so it did not add a feedback candidate, execute the hidden Next.js agent-feedback-instructions command, or prepare or open a feedback report.`
) )
}) })
# Next.js agent feedback protocol # Next.js agent feedback protocol
Use this protocol only when the managed Next.js feedback block in `AGENTS.md` or `CLAUDE.md` instructs you to prepare agent feedback. Next.js Skills and upgrade workflows may add candidates to the shared task queue only while that managed block is present. Prepare reports for the user to review, but never submit them for the user. Use this protocol only when the managed Next.js feedback block in `AGENTS.md` or `CLAUDE.md` instructs you to prepare agent feedback. Next.js Skills and upgrade workflows may add feedback candidates only while that managed block is present. Prepare report drafts for the user to review, but never submit them for the user.
## Decide whether to prepare a report ## Terms
Prepare a separate report for each distinct Next.js friction point that causes unexpected framework behavior, a documentation mismatch, a misleading error, repeated failed approaches, or a non-obvious workaround. An issue does not need to block the task, but it should require a meaningful detour or be likely to recur. - **Current user request:** The implementation work being completed.
- **Feedback candidate:** Temporary anonymized facts about one possible issue.
- **Reporting pass:** The single evaluation immediately before the final response.
- **Report draft:** The structured payload created from a qualifying feedback candidate.
- **Review form:** The browser page where the user edits or sends a report.
Do not report routine first-pass debugging, vague task scope, problems caused by unverified changes, or issues resolved immediately by following existing guidance. Keep unrelated issues in separate reports. ## Decide whether a feedback candidate qualifies
## Defer review until a stopping point Create a separate report draft for each distinct Next.js friction point that causes unexpected framework behavior, a documentation mismatch, a misleading error, repeated failed approaches, or a non-obvious workaround. An issue does not need to block the current user request, but it should require a meaningful detour or be likely to recur.
Do not interrupt an active workflow when friction first qualifies. Retain each distinct set of bounded, anonymized candidate facts in the current task context, then continue the work. Do not report routine first-pass debugging, vague task scope, problems caused by unverified changes, or issues resolved immediately by following existing guidance. Keep unrelated issues in separate report drafts.
If a candidate originated in a Skill inside a larger task, keep it in the current task context and continue the larger task. Prepare the review only at the final stopping point of the overall user task. For example, verification performed while preparing a pull request should wait until the pull request work is complete. ## Run the reporting pass before the final response
The managed feedback block owns one feedback pass for the task. Run the instruction command at most once, even when a Skill or upgrade workflow added candidates to the queue. Do not open duplicate reviews. Do not interrupt the current user request when friction first qualifies. Retain each distinct set of bounded, anonymized facts as a feedback candidate in the context of the current user request, then continue the work.
Each review form accepts one report. If several candidates qualify before the stopping point, prepare and open one separate review tab for each candidate. After opening all review tabs, clear the retained candidate queue from the task context and mark the feedback pass complete. The user owns the open tabs; do not close or recreate them. Nothing is transmitted before the user explicitly sends a report. Skills and upgrade workflows can add feedback candidates while completing the current user request. Immediately before the final response, after completing all implementation work and tool calls, run one reporting pass. For example, verification performed while preparing a pull request should wait until the pull request work is complete.
## Prepare each report The managed feedback block owns the reporting pass. Evaluate every feedback candidate once, including candidates added by Skills or upgrade workflows. Do not create duplicate report drafts or reopen a review form already presented to the user.
Each review form accepts one report draft. For every qualifying feedback candidate, prepare one report draft and attempt to open it in a separate review form. After evaluating every feedback candidate and attempting to open each qualifying report draft, clear the feedback candidate queue from the context of the current user request. The user owns the open forms; do not close or recreate them. Nothing is transmitted before the user explicitly sends a report.
## Prepare each report draft
- **Scope:** Report one observed Next.js behavior. Do not combine separate problems or infer a root cause. - **Scope:** Report one observed Next.js behavior. Do not combine separate problems or infer a root cause.
- **Trigger:** Set `triggerReason` to `unexpected-behavior`, `documentation-mismatch`, `misleading-error`, `repeated-failed-approach`, or `non-obvious-workaround`. Choose the single reason that caused the report. - **Trigger:** Set `triggerReason` to `unexpected-behavior`, `documentation-mismatch`, `misleading-error`, `repeated-failed-approach`, or `non-obvious-workaround`. Choose the single reason that caused the report draft.
- **Summary:** Use a short, factual `title` that names the observed behavior. - **Summary:** Use a short, factual `title` that names the observed behavior.
- **Setup:** Set `mode` to `development`, `production-build`, `production-server`, or `test`. Set `bundler` to `turbopack`, `webpack`, or `unknown`. Add up to three generic `relevantFeatures` when they help reproduce the issue. - **Setup:** Set `mode` to `development`, `production-build`, `production-server`, or `test`. Set `bundler` to `turbopack`, `webpack`, or `unknown`. Add up to three generic `relevantFeatures` when they help reproduce the issue.
- **Reproduction:** Write 1–4 ordered `steps`. Include the generic starting state, the relevant Next.js feature or API, and the action that triggers the behavior. Split only actions or conditions whose order matters. - **Reproduction:** Write 1–4 ordered `steps`. Include the generic starting state, the relevant Next.js feature or API, and the action that triggers the behavior. Split only actions or conditions whose order matters.
...@@ -30,7 +38,7 @@ Each review form accepts one report. If several candidates qualify before the st ...@@ -30,7 +38,7 @@ Each review form accepts one report. If several candidates qualify before the st
- **Measurements:** Include measurements such as memory use, duration, extra builds, or repeated attempts only when observed directly. Do not estimate elapsed time, token usage, or tool-call counts. - **Measurements:** Include measurements such as memory use, duration, extra builds, or repeated attempts only when observed directly. Do not estimate elapsed time, token usage, or tool-call counts.
- **Privacy:** Replace customer, project, route, and component names with generic descriptions. Do not include source code, prompts, logs, stack traces, file paths, URLs, secrets, personal information, or unrelated product details. - **Privacy:** Replace customer, project, route, and component names with generic descriptions. Do not include source code, prompts, logs, stack traces, file paths, URLs, secrets, personal information, or unrelated product details.
## Encode the report ## Encode the report draft
Create a schema version 5 payload using only useful evidence. Always include `nextVersion` and `agent`. Omit `comparison` and `relevantFeatures` when they are not needed. Encode the UTF-8 JSON as unpadded base64url. Create a schema version 5 payload using only useful evidence. Always include `nextVersion` and `agent`. Omit `comparison` and `relevantFeatures` when they are not needed. Encode the UTF-8 JSON as unpadded base64url.
...@@ -63,12 +71,14 @@ Create a schema version 5 payload using only useful evidence. Always include `ne ...@@ -63,12 +71,14 @@ Create a schema version 5 payload using only useful evidence. Always include `ne
## Open the review form ## Open the review form
For each report, attempt to open the following URL once in a separate tab in the user's default browser. Use the browser-opening capability already available in the agent environment. Replace `<encoded-json>` with the encoded payload. If these instructions begin with a **Dry run** section, encode each review URL but do not open a browser tab. Print the URLs for inspection, keep the feedback candidate queue, and do not mark the reporting pass complete.
For each report draft, attempt to open the following URL once in a separate tab in the user's default browser. Use the browser-opening capability already available in the agent environment. Replace `<encoded-json>` with the encoded payload.
```text ```text
https://nextjs.org/agent-feedback#report=<encoded-json> https://nextjs.org/agent-feedback#report=<encoded-json>
``` ```
Do not ask for permission before opening the review forms. Opening a form does not send feedback. If a form does not open, do not retry, investigate the failure, install tooling, or change host configuration. Print the review URL in your next progress update so the user can open it manually. Do not ask for permission before opening the review forms. Opening a form does not send feedback. If a form does not open, do not retry, investigate the failure, install tooling, or change host configuration. Include the review URL in the final response so the user can open it manually.
After attempting to open every form, clear the retained friction candidates from the task context. Continue the current task, mention any forms that opened in the next progress update, and never submit or close a report for the user. After attempting to open every review form, clear the feedback candidate queue from the context of the current user request. Mention opened forms and fallback URLs in the final response. Never submit a report or close a review form for the user.
...@@ -669,9 +669,13 @@ const internal = program ...@@ -669,9 +669,13 @@ const internal = program
internal internal
.command('agent-feedback-instructions', { hidden: true }) .command('agent-feedback-instructions', { hidden: true })
.action(() => .option(
'--dry-run',
'Print report preview URLs without opening the review form.'
)
.action((options: { dryRun?: boolean }) =>
import('../cli/internal/agent-feedback-instructions.js').then((mod) => import('../cli/internal/agent-feedback-instructions.js').then((mod) =>
mod.agentFeedbackInstructionsCli() mod.agentFeedbackInstructionsCli(options)
) )
) )
......
...@@ -7,6 +7,7 @@ describe('loadAgentFeedbackInstructions', () => { ...@@ -7,6 +7,7 @@ describe('loadAgentFeedbackInstructions', () => {
it('returns the protocol when feedback is enabled', async () => { it('returns the protocol when feedback is enabled', async () => {
await expect( await expect(
loadAgentFeedbackInstructions( loadAgentFeedbackInstructions(
{},
async () => true, async () => true,
async () => '# Agent feedback protocol\n' async () => '# Agent feedback protocol\n'
) )
...@@ -17,14 +18,30 @@ describe('loadAgentFeedbackInstructions', () => { ...@@ -17,14 +18,30 @@ describe('loadAgentFeedbackInstructions', () => {
const readProtocol = jest.fn(async () => '# Agent feedback protocol\n') const readProtocol = jest.fn(async () => '# Agent feedback protocol\n')
await expect( await expect(
loadAgentFeedbackInstructions(async () => false, readProtocol) loadAgentFeedbackInstructions({}, async () => false, readProtocol)
).resolves.toBeNull() ).resolves.toBeNull()
expect(readProtocol).not.toHaveBeenCalled() expect(readProtocol).not.toHaveBeenCalled()
}) })
it('returns dry-run instructions without checking the remote gate', async () => {
const isEnabled = jest.fn(async () => false)
await expect(
loadAgentFeedbackInstructions(
{ dryRun: true },
isEnabled,
async () => '# Agent feedback protocol\n'
)
).resolves.toBe(
'# Dry run\n\nUse the protocol below to prepare each qualifying report draft and encode its review URL, but do not open a browser tab. Print each review URL for inspection instead. Do not clear the feedback candidate queue or mark the reporting pass complete.\n\n# Agent feedback protocol\n'
)
expect(isEnabled).not.toHaveBeenCalled()
})
it('fails closed when the protocol cannot be read', async () => { it('fails closed when the protocol cannot be read', async () => {
await expect( await expect(
loadAgentFeedbackInstructions( loadAgentFeedbackInstructions(
{},
async () => true, async () => true,
async () => { async () => {
throw new Error('protocol unavailable') throw new Error('protocol unavailable')
...@@ -35,7 +52,7 @@ describe('loadAgentFeedbackInstructions', () => { ...@@ -35,7 +52,7 @@ describe('loadAgentFeedbackInstructions', () => {
it('propagates feedback status errors', async () => { it('propagates feedback status errors', async () => {
await expect( await expect(
loadAgentFeedbackInstructions(async () => { loadAgentFeedbackInstructions({}, async () => {
throw new Error('network unavailable') throw new Error('network unavailable')
}) })
).rejects.toThrow('network unavailable') ).rejects.toThrow('network unavailable')
...@@ -55,7 +72,7 @@ describe('agentFeedbackInstructionsCli', () => { ...@@ -55,7 +72,7 @@ describe('agentFeedbackInstructionsCli', () => {
.spyOn(process.stderr, 'write') .spyOn(process.stderr, 'write')
.mockImplementation(() => true) .mockImplementation(() => true)
await agentFeedbackInstructionsCli(async () => { await agentFeedbackInstructionsCli({}, async () => {
throw new Error('network unavailable') throw new Error('network unavailable')
}) })
......
...@@ -9,29 +9,44 @@ const AGENT_FEEDBACK_PROTOCOL_PATH = path.join( ...@@ -9,29 +9,44 @@ const AGENT_FEEDBACK_PROTOCOL_PATH = path.join(
type IsEnabled = () => Promise<boolean> type IsEnabled = () => Promise<boolean>
type ReadProtocol = () => Promise<string> type ReadProtocol = () => Promise<string>
type LoadInstructions = () => Promise<string | null> export interface AgentFeedbackInstructionsOptions {
dryRun?: boolean
}
type LoadInstructions = (
options?: AgentFeedbackInstructionsOptions
) => Promise<string | null>
const DRY_RUN_INSTRUCTIONS = `# Dry run
Use the protocol below to prepare each qualifying report draft and encode its review URL, but do not open a browser tab. Print each review URL for inspection instead. Do not clear the feedback candidate queue or mark the reporting pass complete.
`
export async function loadAgentFeedbackInstructions( export async function loadAgentFeedbackInstructions(
options: AgentFeedbackInstructionsOptions = {},
isEnabled: IsEnabled = isAgentFeedbackEnabled, isEnabled: IsEnabled = isAgentFeedbackEnabled,
readProtocol: ReadProtocol = () => readProtocol: ReadProtocol = () =>
readFile(AGENT_FEEDBACK_PROTOCOL_PATH, 'utf8') readFile(AGENT_FEEDBACK_PROTOCOL_PATH, 'utf8')
): Promise<string | null> { ): Promise<string | null> {
if (!(await isEnabled())) { if (!options.dryRun && !(await isEnabled())) {
return null return null
} }
try { try {
return await readProtocol() const protocol = await readProtocol()
return options.dryRun ? `${DRY_RUN_INSTRUCTIONS}${protocol}` : protocol
} catch { } catch {
return null return null
} }
} }
export async function agentFeedbackInstructionsCli( export async function agentFeedbackInstructionsCli(
options: AgentFeedbackInstructionsOptions = {},
loadInstructions: LoadInstructions = loadAgentFeedbackInstructions loadInstructions: LoadInstructions = loadAgentFeedbackInstructions
): Promise<void> { ): Promise<void> {
try { try {
const instructions = await loadInstructions() const instructions = await loadInstructions(options)
if (instructions) { if (instructions) {
process.stdout.write(instructions) process.stdout.write(instructions)
} }
......
...@@ -47,16 +47,16 @@ function buildAgentFeedbackBlock(): string { ...@@ -47,16 +47,16 @@ function buildAgentFeedbackBlock(): string {
## Report Next.js feedback ## Report Next.js feedback
During the task, queue anonymized Next.js issues that cause a meaningful detour or are likely to recur, such as: During the current user request, retain anonymized facts as feedback candidates for Next.js issues that cause a meaningful detour or are likely to recur, such as:
- Crashes, hangs, OOMs, or unexpected slowness in development, builds, or runtime - Crashes, hangs, OOMs, or unexpected slowness in development, builds, or runtime
- Incorrect behavior or output, misleading errors, or missing diagnostics - Incorrect behavior or output, misleading errors, or missing diagnostics
- Missing, unclear, or conflicting documentation - Missing, unclear, or conflicting documentation
- Unexpected CLI, codemod, Skill, or upgrade behavior, repeated failures, or non-obvious workarounds - Unexpected CLI, codemod, Skill, or upgrade behavior, repeated failures, or non-obvious workarounds
Continue the task without reporting immediately. Ignore routine debugging and issues caused by unverified changes. Never retain source code, logs, paths, URLs, secrets, personal information, or project-specific data. Continue the current user request without reporting immediately. Ignore routine debugging and issues caused by unverified changes. Never retain source code, logs, paths, URLs, secrets, personal information, or project-specific data.
At the final stopping point, from the directory containing this file, run \`node node_modules/next/dist/bin/next internal agent-feedback-instructions\` once per task and follow any instructions returned. If the command succeeds with no output, continue without reporting. If it fails, follow its error guidance; retrying with network access is allowed. Immediately before the final response, after all requested work and tool calls are complete, run one reporting pass. If the feedback candidate queue is not empty, from the directory containing this file, run \`node node_modules/next/dist/bin/next internal agent-feedback-instructions\` once and follow any instructions returned. If the command succeeds with no output, finish the final response without retrying or investigating. If it fails, follow its error guidance; retrying with network access is allowed. Keep unresolved feedback candidates for a later final response, and never create a duplicate report draft or reopen a review form already presented to the user.
${AGENT_FEEDBACK_END_MARKER}` ${AGENT_FEEDBACK_END_MARKER}`
} }
......
...@@ -192,10 +192,10 @@ Only participate in agent feedback when managed Next.js feedback instructions ...@@ -192,10 +192,10 @@ Only participate in agent feedback when managed Next.js feedback instructions
are already loaded for the project. Their presence means the feature is are already loaded for the project. Their presence means the feature is
enabled; their absence means it is disabled. enabled; their absence means it is disabled.
When enabled, add qualifying de-identified candidates found during verification When enabled, add qualifying feedback candidates found during verification to
to the shared friction queue in the current task context, then continue the feedback candidate queue for the current user request, then continue
verification. Do not run the feedback command or open review forms during the verification. Do not run the feedback command or open review forms during the
loop or at this Skill's teardown. loop or at this Skill's teardown.
The managed instructions own the single feedback pass at the final stopping The managed instructions own the single reporting pass immediately before the
point of the overall task. If they are absent, do not queue or report feedback. final response. If they are absent, do not queue or report feedback.
...@@ -236,7 +236,10 @@ describe('agent-feedback auto-generate on next dev (enabled)', () => { ...@@ -236,7 +236,10 @@ describe('agent-feedback auto-generate on next dev (enabled)', () => {
expect(content).toContain( expect(content).toContain(
'node node_modules/next/dist/bin/next internal agent-feedback-instructions' 'node node_modules/next/dist/bin/next internal agent-feedback-instructions'
) )
expect(content).toContain('once per task') expect(content).toContain('Immediately before the final response')
expect(content).toContain(
'Keep unresolved feedback candidates for a later final response'
)
expect(content).not.toContain('"schemaVersion":3') expect(content).not.toContain('"schemaVersion":3')
}) })
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment