Files
PaperClipAI/ui/src/components/AttentionInteractionResolver.tsx
T
Devin FoleyandPaperclip 144083fd48 fix(interactions): wait for workspace readiness before enabling approval (#14893)
## Thinking Path

> - Paperclip lets people manage AI agents and review their work.
> - Task confirmations must use the work produced by their source run.
> - The server blocks approval while that run still needs to sync its
workspace.
> - The card currently enables approval before that check can pass, so
an ordinary click produces an error.
> - This PR exposes the existing readiness check and shows “Preparing
approval…” with acceptance disabled.
> - The card refreshes itself and enables approval when the source
workspace settles.

## Linked Issues or Issue Description

**What happened?**

A confirmation appears while its source run is still preparing or
syncing its workspace. Its enabled approval button returns a conflict
asking the user to retry after syncing.

**Steps to reproduce**

1. Run an agent in an isolated workspace.
2. Have it create a confirmation before workspace finalization
completes.
3. Click the approval button while the source workspace is still active.

**Expected behavior**

The card explains that approval is preparing. Acceptance becomes
available automatically when the same server check permits it. Reject
and revise remain available.

Related work: #10770 handles this conflict after a click with retries.
#9520 proposes changing the workspace acceptance barrier. This PR
preserves that barrier and exposes readiness before the click, including
in compact task chat. It preserves the terminal-finalize behavior from
#10099.

## What Changed

- Add an optional, read-only `acceptanceBlocker` to interaction
responses. Readiness uses the existing source-run workspace predicate,
with one check per pending source run.
- Disable acceptance and show a shared preparation notice in classic and
compact confirmation cards, including checkbox and secret-binding
confirmations.
- Refresh preparing cards every two seconds in task detail, attention,
pipelines, and Skill Studio. Restore each surface's previous polling
cadence when preparation clears.
- Preserve live tool reviews, questions, rejection, revision, and the
server acceptance barrier. No automatic acceptance occurs.
- Document the preparation state and cover readiness, terminal sync
outcomes, unrelated runs, historical cards, and automatic refresh.

## Verification

- Focused service, card, query-refresh, and helper tests: 217 passed
across five files.
- `pnpm -r typecheck`: passed.
- `pnpm build`: passed.
- `pnpm build-storybook`: passed.
- `pnpm check:token-gates`: passed.
- `pnpm test:run`: incomplete locally. Stopped after about 17 minutes
once it reproduced seven existing environment failures: two Slack tests
and two email tests lack ancestor-directory skill fixtures; three
company-skills tests fail on macOS runtime-cache staging permissions.
These are outside this change. The full CI test matrix passed.
- CI: all 53 checks passed; two optional Storybook jobs were skipped.
The branch has no conflicts with `master`.
- Greptile: 5/5 on commit `8a6f216d8d`, with no review threads.
- Reviewed added lines and new files for credentials, private URLs,
internal task references, user paths, and run artifacts. None found.

## Risks

- No database migration or change to acceptance authorization. Readiness
is advisory; the server still enforces its existing gate at acceptance.
- An open preparing card adds a read every two seconds. This cadence
stops after readiness clears; historical cards add no workspace checks.
- Failed or stale finalization retains the existing server behavior.
This PR does not change recovery policy.

## Model Used

OpenAI GPT-6 through Codex. The exact serving model ID and
context-window size are not exposed in this session. Used reasoning,
repository inspection, tool execution, and automated tests. No
sub-agents.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass (217 focused tests; full
local-suite limitations are recorded above)
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-10-02 13:08:25 -07:00

170 lines
6.4 KiB
TypeScript

import { useMemo } from "react";
import { useMutation, useQuery, useQueryClient } from "@tanstack/react-query";
import { Loader2 } from "lucide-react";
import type { Agent } from "@paperclipai/shared";
import { issuesApi } from "../api/issues";
import { queryKeys } from "../lib/queryKeys";
import {
interactionReadinessRefetchInterval,
isIssueThreadInteraction,
type AskUserQuestionsAnswer,
type AskUserQuestionsInteraction,
type IssueThreadInteraction,
type RequestCheckboxConfirmationInteraction,
type RequestConfirmationInteraction,
type RequestItemVerdictsInteraction,
type RequestItemVerdictValue,
type SuggestTasksInteraction,
} from "../lib/issue-thread-interactions";
import { IssueThreadInteractionCard } from "./IssueThreadInteractionCard";
interface AttentionInteractionResolverProps {
companyId: string;
issueId: string;
interactionId: string;
agentMap?: Map<string, Agent>;
currentUserId?: string | null;
userLabelMap?: ReadonlyMap<string, string> | null;
/** Called after a resolution so the parent can refresh the feed. */
onResolved?: () => void;
}
export function replaceResolvedInteraction(
current: IssueThreadInteraction[] | undefined,
resolvedInteraction: IssueThreadInteraction,
): IssueThreadInteraction[] {
const existing = current ?? [];
const index = existing.findIndex((entry) => entry.id === resolvedInteraction.id);
if (index < 0) return [...existing, resolvedInteraction];
return existing.map((entry, entryIndex) =>
entryIndex === index ? resolvedInteraction : entry,
);
}
/**
* Lazily fetches the full issue-thread interaction referenced by an attention
* row and renders the existing {@link IssueThreadInteractionCard} inline, so
* confirmations and questions are answerable in-row without leaving the queue
* (converged PAP-12628). Reviews never reach here — they deep-link.
*/
export function AttentionInteractionResolver({
companyId,
issueId,
interactionId,
agentMap,
currentUserId,
userLabelMap,
onResolved,
}: AttentionInteractionResolverProps) {
const queryClient = useQueryClient();
const { data: interactions, isLoading, error } = useQuery({
queryKey: queryKeys.issues.interactions(issueId),
queryFn: () => issuesApi.listInteractions(issueId),
enabled: !!issueId,
refetchInterval: (query) => interactionReadinessRefetchInterval(query.state.data),
});
const interaction = useMemo<IssueThreadInteraction | null>(() => {
const match = (interactions ?? []).find((entry) => entry.id === interactionId);
return match && isIssueThreadInteraction(match) ? match : null;
}, [interactions, interactionId]);
const invalidate = (resolvedInteraction?: IssueThreadInteraction) => {
// The accept route returns the fully stitched interaction (including
// secretProposal.executed/failed). Put that receipt in the cache before
// invalidating so the attention surface never flashes a generic Accepted
// state while the refetch catches up.
if (resolvedInteraction) {
queryClient.setQueryData<IssueThreadInteraction[] | undefined>(
queryKeys.issues.interactions(issueId),
(current) => replaceResolvedInteraction(current, resolvedInteraction),
);
}
queryClient.invalidateQueries({ queryKey: queryKeys.issues.interactions(issueId) });
queryClient.invalidateQueries({ queryKey: queryKeys.attention(companyId) });
onResolved?.();
};
const acceptMutation = useMutation({
mutationFn: (input: {
interaction: SuggestTasksInteraction | RequestConfirmationInteraction | RequestCheckboxConfirmationInteraction;
selectedClientKeys?: string[];
selectedOptionIds?: string[];
rememberAction?: boolean;
}) =>
issuesApi.acceptInteraction(issueId, input.interaction.id, {
selectedClientKeys: input.selectedClientKeys,
selectedOptionIds: input.selectedOptionIds,
rememberAction: input.rememberAction,
}),
onSuccess: invalidate,
});
const rejectMutation = useMutation({
mutationFn: (input: { interactionId: string; reason?: string }) =>
issuesApi.rejectInteraction(issueId, input.interactionId, input.reason),
onSuccess: invalidate,
});
const respondMutation = useMutation({
mutationFn: (input: { interactionId: string; answers: AskUserQuestionsAnswer[] }) =>
issuesApi.respondToInteraction(issueId, input.interactionId, { answers: input.answers }),
onSuccess: invalidate,
});
const cancelMutation = useMutation({
mutationFn: (input: { interactionId: string }) =>
issuesApi.cancelInteraction(issueId, input.interactionId),
onSuccess: invalidate,
});
const verdictsMutation = useMutation({
mutationFn: (input: {
interactionId: string;
verdicts: { id: string; verdict: RequestItemVerdictValue; reason?: string }[];
}) => issuesApi.submitInteractionVerdicts(issueId, input.interactionId, input.verdicts),
onSuccess: invalidate,
});
if (isLoading) {
return (
<div className="flex items-center gap-2 py-3 text-xs text-muted-foreground">
<Loader2 className="h-3.5 w-3.5 animate-spin" /> Loading decision…
</div>
);
}
if (error || !interaction) {
return (
<p className="py-3 text-xs text-muted-foreground">
This decision is no longer available — it may have been resolved elsewhere.
</p>
);
}
return (
<IssueThreadInteractionCard
interaction={interaction}
agentMap={agentMap}
currentUserId={currentUserId}
userLabelMap={userLabelMap}
onAcceptInteraction={(target, selectedClientKeys, selectedOptionIds, rememberAction) =>
acceptMutation.mutateAsync({ interaction: target, selectedClientKeys, selectedOptionIds, rememberAction }).then(() => undefined)
}
onRejectInteraction={(target, reason) =>
rejectMutation.mutateAsync({ interactionId: target.id, reason }).then(() => undefined)
}
onSubmitInteractionAnswers={(target: AskUserQuestionsInteraction, answers) =>
respondMutation.mutateAsync({ interactionId: target.id, answers }).then(() => undefined)
}
onCancelInteraction={(target: AskUserQuestionsInteraction) =>
cancelMutation.mutateAsync({ interactionId: target.id }).then(() => undefined)
}
onSubmitInteractionVerdicts={(target: RequestItemVerdictsInteraction, verdicts) =>
verdictsMutation.mutateAsync({ interactionId: target.id, verdicts }).then(() => undefined)
}
/>
);
}