feat(platform): wire feature flags resolver and add governance tab#1375
Merged
Conversation
📝 WalkthroughWalkthroughThis PR introduces per-user, per-team, and per-role feature flag governance enforcement for an organization. It adds a React component ( Estimated code review effort🎯 3 (Moderate) | ⏱️ ~35 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
There was a problem hiding this comment.
Your free trial has ended. If you'd like to continue receiving code reviews, you can add a payment method here.
Wire the existing resolveFeatureFlags() into the chat pipeline so governance feature flag rules are enforced before agent generation. Add validation for feature_flags config in upsertPolicy, pass maxContextTokens through to context builder, and build a FeatureFlagsEditor admin UI on the governance settings page. - Call resolveFeatureFlags() in startAgentChat after budget check - Enforce webSearch disable (remove web tool, set mode off) - Block file uploads with assistant message when fileUpload is false - Forward maxContextTokens through scheduler to generate_response - Override maxHistoryTokens in all buildStructuredContext calls - Add featureFlagsConfigSchema validation in upsertPolicy mutation - Build FeatureFlagsEditor with scope/target rules table and dialog - Add Feature Controls tab to governance page with i18n tab labels - Add translation keys to en.json and de.json - Add unit tests for resolveFeatureFlags priority chain - Add component tests with accessibility checks - Add Storybook story for FeatureFlagsEditor Closes #1360
Add tests verifying governance feature flag enforcement: - maxContextTokens forwarded to scheduled generation args - web search disabled removes web tool and sets mode off - file upload disabled blocks request with assistant message - file upload disabled allows request without attachments
Replace all hardcoded English strings in feature-flags-editor with i18n translation keys (scope/role labels, search placeholders, aria-labels, submit text, actions header). Use formatNumber() instead of toLocaleString(). Show loading skeleton with aria-busy instead of returning null. Normalize maxContextTokens once before reuse in generate_response. Remove redundant object spread in start_agent_chat.
larryro
force-pushed
the
feat/issue-1360-feature-flags
branch
from
eccf802 to
917e25f
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 9
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In
`@services/platform/app/features/settings/governance/components/feature-flags-editor.test.tsx`:
- Around line 63-66: Replace the unsafe "as never" casts on
mockedUseGovernancePolicy fixtures with a typed mock that preserves the hook's
return shape: create a small typed factory or use TypeScript's "satisfies" on
the mock objects so they conform to the actual useGovernancePolicy return type
(e.g., the shape used by mockedUseGovernancePolicy in
feature-flags-editor.test.tsx) and update all instances (lines matching the
other fixtures at 78-93, 116-119, 135-150) to use that factory or "satisfies"
expression so the tests will error if the hook contract drifts.
In
`@services/platform/app/features/settings/governance/components/feature-flags-editor.tsx`:
- Around line 255-275: Replace the hardcoded placeholder string in the Input
component for maxContextTokens with a translated string: add a new translation
key (e.g., "featureFlags.maxContextTokensPlaceholder" or
"maxContextTokensPlaceholder") to the locale JSON and use the i18n function
t(...) in the Input's placeholder prop in the FeatureFlagsEditor component (the
Input where value={draft.maxContextTokens ?? ''} and onChange updates
draft.maxContextTokens). Ensure the key matches your translations naming
convention and update any tests or snapshots if needed.
- Around line 505-547: The table rows currently use the array index as the React
key in rules.map which can break reconciliation; generate a stable client-side
ID for each rule when rules are created/loaded (e.g., assign rule._clientId =
crypto.randomUUID() in the functions that initialize or add rules) and use
rule._clientId as the key in rules.map; ensure functions referenced here
(rules.map rendering, the rule-creation handler where new rules are pushed, and
any loader that populates initial rules) preserve that _clientId and strip it
only when sending payloads to the backend (or omit it in save logic), keeping
openEditDialog and removeRule working with index or find-by-_clientId as needed.
In `@services/platform/app/routes/dashboard/`$id/settings/governance.tsx:
- Around line 61-69: Replace the hardcoded label 'Retention' with the
translation hook (e.g. t('governance.tabs.retention')) so the tab uses localized
text; update the object with value: 'retention' (the same one that renders
<RetentionEditor organizationId={organizationId} />) to set label:
t('governance.tabs.retention') and ensure the translation key exists in the
locale files.
In `@services/platform/convex/governance/__tests__/feature_enforcement.test.ts`:
- Around line 17-25: The test fixture createMockCtx currently uses the unsafe
assertion "as never"; replace this with TypeScript's "satisfies" to preserve
type safety by ensuring the mock shape matches the real context type. Update
createMockCtx to return the object with "satisfies <ContextType>" where
<ContextType> is the actual context/interface used by the code under test (or
define an expectedCtxShape type reflecting db.query and auth.getUserIdentity),
removing "as never" so the compiler validates the mock structure while keeping
proper inferred types.
In `@services/platform/convex/lib/agent_chat/start_agent_chat.ts`:
- Around line 269-285: The file-upload feature-flag check
(featureFlags.fileUpload) runs after buildMessageWithAttachments() and
saveMessage(), causing forbidden attachments to be persisted; move the guard
that inspects attachments and featureFlags.fileUpload to run before calling
buildMessageWithAttachments() and before any saveMessage() call so rejected
uploads are never written. Specifically, relocate the block that references
featureFlags.fileUpload, attachments, saveMessage, threadMeta, threadId,
messageAlreadyExists, and streamId to an earlier point in start_agent_chat where
attachments are first available (before buildMessageWithAttachments and before
the initial saveMessage), and keep the same return behavior and threadMeta
generationStatus update when rejecting.
- Around line 250-289: Call resolveFeatureFlags with the user's role (pass the
`role` variable) so role-scoped rules are evaluated, and then enforce the
returned codeExecution flag into `enforcedConfig` (alongside the existing
webSearch handling): when `featureFlags.codeExecution` is false, set the
appropriate code-execution toggle in `enforcedConfig` (e.g., `codeExecutionMode:
'off'` or the boolean field your agent uses) and remove any code-execution tools
from `enforcedConfig.convexToolNames` (similar to how 'web' is filtered out);
update the block around `resolveFeatureFlags`, `enforcedConfig`, and
`agentConfig` to apply these changes so governance actually disables code
execution for this chat path.
In `@services/platform/messages/de.json`:
- Around line 3588-3612: The JSON adds governance.tabs and
governance.featureFlags blocks that are being overwritten by later duplicate
keys; merge the new keys into the existing governance.tabs and
governance.featureFlags objects instead of adding duplicate top-level objects so
translations aren't lost — update the German locale to merge keys like
tabs.featureControls into governance.tabs and featureFlags.fileUpload,
featureFlags.enabled, featureFlags.scope, featureFlags.target,
featureFlags.noRules, featureFlags.maxContextTokensHint, featureFlags.saved, and
featureFlags.saveFailed into governance.featureFlags (ensure you preserve any
existing keys in those objects and only add/replace the missing/new entries).
In `@services/platform/messages/en.json`:
- Around line 3641-3646: The JSON contains two governance.tabs objects causing
the later one to override the first and drop governance.tabs.featureControls;
fix by merging the unique key "pii" from the second governance.tabs into the
first governance.tabs object (which already includes systemPrompt, budgets,
featureControls, usage) and then remove the duplicate governance.tabs entry so
only the merged governance.tabs remains.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: fa26d370-d083-452a-bcd9-04523607a725
📒 Files selected for processing (14)
services/platform/app/features/settings/governance/components/feature-flags-editor.stories.tsxservices/platform/app/features/settings/governance/components/feature-flags-editor.test.tsxservices/platform/app/features/settings/governance/components/feature-flags-editor.tsxservices/platform/app/routes/dashboard/$id/settings/governance.tsxservices/platform/convex/governance/__tests__/feature_enforcement.test.tsservices/platform/convex/governance/mutations.tsservices/platform/convex/lib/agent_chat/__tests__/start_agent_chat.test.tsservices/platform/convex/lib/agent_chat/internal_actions.tsservices/platform/convex/lib/agent_chat/start_agent_chat.tsservices/platform/convex/lib/agent_chat/types.tsservices/platform/convex/lib/agent_response/generate_response.tsservices/platform/convex/lib/agent_response/types.tsservices/platform/messages/de.jsonservices/platform/messages/en.json
Comment on lines
+63
to
+66
| mockedUseGovernancePolicy.mockReturnValue({ | ||
| data: null, | ||
| isLoading: false, | ||
| } as never); |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Prefer typed mock fixtures over as never.
These casts erase the useGovernancePolicy contract, so the tests will still compile if the hook return shape drifts. A small typed factory or satisfies keeps the mocks honest without losing inference. Based on learnings, "In TypeScript test fixtures ... prefer the 'satisfies' operator over 'as' type assertions for config objects and test data."
Also applies to: 78-93, 116-119, 135-150
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In
`@services/platform/app/features/settings/governance/components/feature-flags-editor.test.tsx`
around lines 63 - 66, Replace the unsafe "as never" casts on
mockedUseGovernancePolicy fixtures with a typed mock that preserves the hook's
return shape: create a small typed factory or use TypeScript's "satisfies" on
the mock objects so they conform to the actual useGovernancePolicy return type
(e.g., the shape used by mockedUseGovernancePolicy in
feature-flags-editor.test.tsx) and update all instances (lines matching the
other fixtures at 78-93, 116-119, 135-150) to use that factory or "satisfies"
expression so the tests will error if the hook contract drifts.
Comment on lines
+255
to
+275
| <div> | ||
| <Input | ||
| label={t('featureFlags.maxContextTokens')} | ||
| type="number" | ||
| value={draft.maxContextTokens ?? ''} | ||
| onChange={(e) => | ||
| updateDraft({ | ||
| maxContextTokens: e.target.value | ||
| ? Number(e.target.value) | ||
| : undefined, | ||
| }) | ||
| } | ||
| disabled={cannotManage} | ||
| size="sm" | ||
| placeholder="e.g. 50000" | ||
| min={0} | ||
| /> | ||
| <Text className="text-muted-foreground mt-1 text-xs"> | ||
| {t('featureFlags.maxContextTokensHint')} | ||
| </Text> | ||
| </div> |
There was a problem hiding this comment.
Hardcoded placeholder text should use translation key.
Line 269 has a hardcoded placeholder "e.g. 50000" which should be internationalized per coding guidelines.
🌐 Proposed fix
Add a translation key in en.json:
"maxContextTokensPlaceholder": "e.g. 50000"Then use it in the component:
placeholder="e.g. 50000"
+ placeholder={t('featureFlags.maxContextTokensPlaceholder')}🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In
`@services/platform/app/features/settings/governance/components/feature-flags-editor.tsx`
around lines 255 - 275, Replace the hardcoded placeholder string in the Input
component for maxContextTokens with a translated string: add a new translation
key (e.g., "featureFlags.maxContextTokensPlaceholder" or
"maxContextTokensPlaceholder") to the locale JSON and use the i18n function
t(...) in the Input's placeholder prop in the FeatureFlagsEditor component (the
Input where value={draft.maxContextTokens ?? ''} and onChange updates
draft.maxContextTokens). Ensure the key matches your translations naming
convention and update any tests or snapshots if needed.
Comment on lines
+505
to
+547
| <tbody> | ||
| {rules.map((rule, index) => ( | ||
| <tr key={index} className="border-border border-b"> | ||
| <td className="px-3 py-2 capitalize">{rule.scope}</td> | ||
| <td className="px-3 py-2">{resolveTarget(rule)}</td> | ||
| <td className="px-3 py-2 text-center"> | ||
| {rule.webSearch === false ? '\u2718' : '\u2714'} | ||
| </td> | ||
| <td className="px-3 py-2 text-center"> | ||
| {rule.codeExecution === false ? '\u2718' : '\u2714'} | ||
| </td> | ||
| <td className="px-3 py-2 text-center"> | ||
| {rule.fileUpload === false ? '\u2718' : '\u2714'} | ||
| </td> | ||
| <td className="px-3 py-2 text-right"> | ||
| {rule.maxContextTokens != null | ||
| ? formatNumber(rule.maxContextTokens) | ||
| : '\u2014'} | ||
| </td> | ||
| <td className="px-3 py-2 text-right"> | ||
| <HStack gap={1} justify="end"> | ||
| <Button | ||
| variant="ghost" | ||
| size="icon" | ||
| onClick={() => openEditDialog(index)} | ||
| disabled={cannotManage} | ||
| aria-label={`${t('featureFlags.editRule')} ${index + 1}`} | ||
| > | ||
| <Pencil className="size-4" /> | ||
| </Button> | ||
| <Button | ||
| variant="ghost" | ||
| size="icon" | ||
| onClick={() => removeRule(index)} | ||
| disabled={cannotManage} | ||
| aria-label={`${t('featureFlags.deleteRule')} ${index + 1}`} | ||
| > | ||
| <Trash2 className="size-4" /> | ||
| </Button> | ||
| </HStack> | ||
| </td> | ||
| </tr> | ||
| ))} |
There was a problem hiding this comment.
🧹 Nitpick | 🔵 Trivial
Consider generating unique IDs for rules to improve React reconciliation.
Using array index as the React key (Line 507) works for the current implementation but can cause subtle bugs if rules are ever reordered or if list operations become more complex. Since rules lack a natural unique identifier, consider generating a client-side ID (e.g., using crypto.randomUUID()) when creating rules.
♻️ Example approach
function emptyRule(): FeatureFlagRule {
return {
+ _clientId: crypto.randomUUID(),
scope: 'default',
webSearch: true,
codeExecution: true,
fileUpload: true,
};
}Then use rule._clientId as the key. The _clientId would be stripped before saving to the backend.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In
`@services/platform/app/features/settings/governance/components/feature-flags-editor.tsx`
around lines 505 - 547, The table rows currently use the array index as the
React key in rules.map which can break reconciliation; generate a stable
client-side ID for each rule when rules are created/loaded (e.g., assign
rule._clientId = crypto.randomUUID() in the functions that initialize or add
rules) and use rule._clientId as the key in rules.map; ensure functions
referenced here (rules.map rendering, the rule-creation handler where new rules
are pushed, and any loader that populates initial rules) preserve that _clientId
and strip it only when sending payloads to the backend (or omit it in save
logic), keeping openEditDialog and removeRule working with index or
find-by-_clientId as needed.
Comment on lines
61
to
+69
| { | ||
| value: 'retention', | ||
| label: 'Retention', | ||
| content: <RetentionEditor organizationId={organizationId} />, | ||
| }, | ||
| { | ||
| value: 'feature-controls', | ||
| label: t('tabs.featureControls'), | ||
| content: <FeatureFlagsEditor organizationId={organizationId} />, |
There was a problem hiding this comment.
Localize the retention tab label.
'Retention' is still hardcoded while the surrounding tabs were moved to governance.tabs.*, so this tab will remain English in localized UIs. As per coding guidelines, "Do NOT hardcode text, use the translation hooks/functions instead for user-facing UI".
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@services/platform/app/routes/dashboard/`$id/settings/governance.tsx around
lines 61 - 69, Replace the hardcoded label 'Retention' with the translation hook
(e.g. t('governance.tabs.retention')) so the tab uses localized text; update the
object with value: 'retention' (the same one that renders <RetentionEditor
organizationId={organizationId} />) to set label: t('governance.tabs.retention')
and ensure the translation key exists in the locale files.
Comment on lines
+17
to
+25
| function createMockCtx() { | ||
| return { | ||
| db: { | ||
| query: vi.fn(), | ||
| }, | ||
| auth: { | ||
| getUserIdentity: vi.fn(), | ||
| }, | ||
| } as never; |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Verify whether unsafe cast escape hatches remain in this test.
# Expected after fix: no output.
rg -nP --type=ts '\bas\s+never\b' services/platform/convex/governance/__tests__/feature_enforcement.test.tsRepository: tale-project/tale
Length of output: 77
Replace as never with satisfies in test fixture (Line 25).
Use satisfies instead of as never to maintain type safety in mock objects. For example:
Suggested change
function createMockCtx() {
return {
db: {
query: vi.fn(),
},
auth: {
getUserIdentity: vi.fn(),
},
} satisfies typeof expectedCtxShape; // Define expectedCtxShape based on actual context interfaceType assertions like as bypass compile-time validation; satisfies verifies the object structure matches the type while preserving proper typing.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@services/platform/convex/governance/__tests__/feature_enforcement.test.ts`
around lines 17 - 25, The test fixture createMockCtx currently uses the unsafe
assertion "as never"; replace this with TypeScript's "satisfies" to preserve
type safety by ensuring the mock shape matches the real context type. Update
createMockCtx to return the object with "satisfies <ContextType>" where
<ContextType> is the actual context/interface used by the code under test (or
define an expectedCtxShape type reflecting db.query and auth.getUserIdentity),
removing "as never" so the compiler validates the mock structure while keeping
proper inferred types.
Comment on lines
+250
to
+289
| if (userId) { | ||
| const userTeamIds = await getUserTeamIds(ctx, userId); | ||
| const featureFlags = await resolveFeatureFlags( | ||
| ctx, | ||
| organizationId, | ||
| userId, | ||
| userTeamIds, | ||
| ); | ||
|
|
||
| if (!featureFlags.webSearch) { | ||
| enforcedConfig = { | ||
| ...agentConfig, | ||
| webSearchMode: 'off', | ||
| convexToolNames: (agentConfig.convexToolNames ?? []).filter( | ||
| (t) => t !== 'web', | ||
| ), | ||
| }; | ||
| } | ||
|
|
||
| if (!featureFlags.fileUpload && attachments && attachments.length > 0) { | ||
| await saveMessage(ctx, components.agent, { | ||
| threadId, | ||
| message: { | ||
| role: 'assistant', | ||
| content: | ||
| 'File uploads are disabled for your account by organization policy. Please contact your administrator.', | ||
| }, | ||
| }); | ||
| if (threadMeta) { | ||
| await ctx.db.patch(threadMeta._id, { | ||
| generationStatus: 'idle' as const, | ||
| updatedAt: Date.now(), | ||
| }); | ||
| } | ||
| return { messageAlreadyExists, streamId }; | ||
| } | ||
|
|
||
| if (featureFlags.maxContextTokens != null) { | ||
| governanceMaxContextTokens = featureFlags.maxContextTokens; | ||
| } |
There was a problem hiding this comment.
Role-based and code-execution rules are still no-ops here.
resolveFeatureFlags() supports a role parameter, but this call omits it, so role-scoped rules can never match. The returned codeExecution flag is also never applied to enforcedConfig, leaving that governance control ineffective on this chat path.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@services/platform/convex/lib/agent_chat/start_agent_chat.ts` around lines 250
- 289, Call resolveFeatureFlags with the user's role (pass the `role` variable)
so role-scoped rules are evaluated, and then enforce the returned codeExecution
flag into `enforcedConfig` (alongside the existing webSearch handling): when
`featureFlags.codeExecution` is false, set the appropriate code-execution toggle
in `enforcedConfig` (e.g., `codeExecutionMode: 'off'` or the boolean field your
agent uses) and remove any code-execution tools from
`enforcedConfig.convexToolNames` (similar to how 'web' is filtered out); update
the block around `resolveFeatureFlags`, `enforcedConfig`, and `agentConfig` to
apply these changes so governance actually disables code execution for this chat
path.
Comment on lines
+269
to
+285
| if (!featureFlags.fileUpload && attachments && attachments.length > 0) { | ||
| await saveMessage(ctx, components.agent, { | ||
| threadId, | ||
| message: { | ||
| role: 'assistant', | ||
| content: | ||
| 'File uploads are disabled for your account by organization policy. Please contact your administrator.', | ||
| }, | ||
| }); | ||
| if (threadMeta) { | ||
| await ctx.db.patch(threadMeta._id, { | ||
| generationStatus: 'idle' as const, | ||
| updatedAt: Date.now(), | ||
| }); | ||
| } | ||
| return { messageAlreadyExists, streamId }; | ||
| } |
There was a problem hiding this comment.
Reject blocked uploads before persisting the user message.
This guard runs after buildMessageWithAttachments() and saveMessage() at Lines 179-198, so a forbidden upload is still written into thread history with attachment URLs/file IDs before the assistant warning is sent. Move the fileUpload check ahead of attachment markdown generation and message persistence.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@services/platform/convex/lib/agent_chat/start_agent_chat.ts` around lines 269
- 285, The file-upload feature-flag check (featureFlags.fileUpload) runs after
buildMessageWithAttachments() and saveMessage(), causing forbidden attachments
to be persisted; move the guard that inspects attachments and
featureFlags.fileUpload to run before calling buildMessageWithAttachments() and
before any saveMessage() call so rejected uploads are never written.
Specifically, relocate the block that references featureFlags.fileUpload,
attachments, saveMessage, threadMeta, threadId, messageAlreadyExists, and
streamId to an earlier point in start_agent_chat where attachments are first
available (before buildMessageWithAttachments and before the initial
saveMessage), and keep the same return behavior and threadMeta generationStatus
update when rejecting.
Comment on lines
+3588
to
+3612
| "tabs": { | ||
| "systemPrompt": "System-Prompt", | ||
| "budgets": "Budgets", | ||
| "featureControls": "Feature-Steuerung", | ||
| "usage": "Nutzung" | ||
| }, | ||
| "featureFlags": { | ||
| "title": "Feature-Steuerung", | ||
| "description": "Funktionen pro Benutzer, Team oder Rolle aktivieren oder deaktivieren.", | ||
| "webSearch": "Websuche", | ||
| "codeExecution": "Code-Ausführung", | ||
| "fileUpload": "Datei-Upload", | ||
| "maxContextTokens": "Max. Kontext-Tokens", | ||
| "maxContextTokensHint": "Maximale Kontext-Tokens für KI-Antworten. Leer lassen für unbegrenzt.", | ||
| "enabled": "Feature-Steuerung aktivieren", | ||
| "scope": "Geltungsbereich", | ||
| "target": "Ziel", | ||
| "allUsers": "Alle Benutzer", | ||
| "addRule": "Regel hinzufügen", | ||
| "editRule": "Regel bearbeiten", | ||
| "deleteRule": "Regel löschen", | ||
| "noRules": "Keine Feature-Regeln konfiguriert. Füge eine Regel hinzu, um Funktionen pro Benutzer, Team oder Rolle zu steuern.", | ||
| "saved": "Feature-Steuerung gespeichert", | ||
| "saveFailed": "Feature-Steuerung konnte nicht gespeichert werden" | ||
| }, |
There was a problem hiding this comment.
Merge these into the existing governance.tabs and governance.featureFlags objects.
JSON keeps the last duplicate key, so the later tabs block at Line 3717 and featureFlags block at Line 3667 overwrite this entire addition. In de, that drops tabs.featureControls plus most of the new feature-flag strings (fileUpload, enabled, scope, target, noRules, etc.), so the new governance UI will render missing translations.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@services/platform/messages/de.json` around lines 3588 - 3612, The JSON adds
governance.tabs and governance.featureFlags blocks that are being overwritten by
later duplicate keys; merge the new keys into the existing governance.tabs and
governance.featureFlags objects instead of adding duplicate top-level objects so
translations aren't lost — update the German locale to merge keys like
tabs.featureControls into governance.tabs and featureFlags.fileUpload,
featureFlags.enabled, featureFlags.scope, featureFlags.target,
featureFlags.noRules, featureFlags.maxContextTokensHint, featureFlags.saved, and
featureFlags.saveFailed into governance.featureFlags (ensure you preserve any
existing keys in those objects and only add/replace the missing/new entries).
Comment on lines
+3641
to
+3646
| "tabs": { | ||
| "systemPrompt": "System prompt", | ||
| "budgets": "Budgets", | ||
| "featureControls": "Feature controls", | ||
| "usage": "Usage" | ||
| }, |
There was a problem hiding this comment.
Duplicate governance.tabs object will cause missing translation key.
There are two governance.tabs objects in this JSON file:
- Lines 3641-3646: Contains
systemPrompt,budgets,featureControls,usage - Lines 3783-3788: Contains
systemPrompt,budgets,usage,pii
JSON parsers will use the last occurrence, so governance.tabs.featureControls will be undefined at runtime, breaking the new Feature Controls tab UI.
🐛 Proposed fix: merge the two objects
Remove the duplicate at lines 3783-3788 and merge its unique key (pii) into the first definition:
"tabs": {
"systemPrompt": "System prompt",
"budgets": "Budgets",
"featureControls": "Feature controls",
- "usage": "Usage"
+ "usage": "Usage",
+ "pii": "PII protection"
},Then delete lines 3783-3788 entirely.
Also applies to: 3783-3788
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@services/platform/messages/en.json` around lines 3641 - 3646, The JSON
contains two governance.tabs objects causing the later one to override the first
and drop governance.tabs.featureControls; fix by merging the unique key "pii"
from the second governance.tabs into the first governance.tabs object (which
already includes systemPrompt, budgets, featureControls, usage) and then remove
the duplicate governance.tabs entry so only the merged governance.tabs remains.
Remove duplicate governance.tabs and governance.featureFlags entries, merging feature controls tab label into the canonical tabs section.
…load when disabled Add getMyFeatureFlags query that resolves the current user's feature flags server-side (reactive via Convex subscriptions). Use it in ChatInterface to hide the file upload button, drop zone, and paste handler when fileUpload is disabled by governance policy. Also disable the Confirm button in the rule edit dialog until the user makes a change.
This was referenced Apr 11, 2026
This was referenced Apr 19, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
resolveFeatureFlags()into the chat pipeline (startAgentChat) so governance feature flag rules are enforced server-side before agent generation startsfeatureFlagsConfigSchemavalidation to theupsertPolicymutation (rejects invalid payloads)webfromconvexToolNamesand settingwebSearchModeto'off'fileUploadisfalsefor the usermaxContextTokensthrough the scheduler args togenerateAgentResponse, overridingmaxHistoryTokensin all threebuildStructuredContextcallsFeatureFlagsEditoradmin UI component (modeled afterBudgetEditor) with scope/target rules table, add/edit dialog with toggle switches, and number input for max context tokensgovernance.tabs.*en.jsonandde.jsonTest plan
resolveFeatureFlagspriority chain (user > team > role > default), partial rules, disabled policy, empty rulesstartAgentChattests updated with mocks for new dependenciesFeatureFlagsEditorcomponent tests: empty state, rules table, add button, enabled toggle, loading statecheckAccessibility()) for empty state and rules tableFeatureFlagsEditorbun run --filter @tale/platform test --project server)bun run --filter @tale/platform test --project client)bun run --filter @tale/platform lint)bun run --filter @tale/platform typecheck)Closes #1360
Summary by CodeRabbit
Release Notes
New Features
Tests
Documentation