Files
PaperClipAI/ui/src/lib/issue-reference.ts
T
DottaandClaude Opus 5.5 68b7852cd1 fix(ui): keep another company's prefix on task links in markdown (#21026)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - Chat replies, comments and documents render Markdown through
`MarkdownBody`. Task links in that Markdown become company-aware links
> - `parseIssueReferenceFromHref` turned every task link into an
unprefixed `/issues/<id>` path. The router then added the viewer's
active company prefix
> - In chat-eval cohort 4 (`access.other-company`), a reply in company
COM linked `/BIR/issues/BIR-1`. The chat rendered it as
`/COM/issues/BIR-1`, which points into the wrong company
> - This pull request keeps the company that a task link names: an
explicit `/BIR/issues/...` prefix, or an identifier whose prefix belongs
to a known company
> - The benefit is that a link to another company's task opens that
company's task, not a same-numbered task or an error in the current
company

## Linked Issues or Issue Description

**What happened?**
A chat reply in company COM contained `[BIR-1](/BIR/issues/BIR-1)`. The
rendered link pointed to `/COM/issues/BIR-1`.

**Expected behavior**
A task link that names a company keeps that company: `/BIR/issues/BIR-1`
stays `/BIR/issues/BIR-1`. A bare `BIR-1` or `/issues/BIR-1` goes to BIR
when BIR is a company that the viewer knows.

**Steps to reproduce**
1. Have two companies, COM and BIR.
2. In a COM chat or comment, render the Markdown
`[BIR-1](/BIR/issues/BIR-1)`.
3. Inspect the link. Its `href` is `/COM/issues/BIR-1`.

**Paperclip version or commit**
master at f1aef0cb58.

## What Changed

- `ui/src/lib/issue-reference.ts`: a new `issueHref` builds the internal
link. It keeps the company prefix of an explicit `/PREFIX/issues/...`
path. When the path has no prefix, it uses the identifier's prefix if
that prefix is in the known company prefixes. Otherwise the link stays
`/issues/<id>` and resolves in the active company, as before.
- `ui/src/components/MarkdownBody.tsx`: passes the known company
prefixes to the explicit-link path too (the auto-link path already had
them).
- Tests: a red/green unit test with the recorded shape in
`issue-reference.test.ts`, and a rendered test in
`MarkdownBody.test.tsx` (COM and BIR known; explicit, bare and
unprefixed BIR links all render `/BIR/issues/BIR-1`). Existing
expectations that asserted the stripped `/issues/<id>` form now assert
the company-prefixed form.

## Verification

- `cd ui && npx vitest run src/components/MarkdownBody.test.tsx
src/lib/issue-reference.test.ts src/lib/company-routes.test.ts`: 105
tests pass.
- Red/green: with `issue-reference.ts` and `MarkdownBody.tsx` from
`origin/master`, the new tests fail (`/issues/BIR-1` instead of
`/BIR/issues/BIR-1`). With the change, they pass.
- `pnpm check:token-gates` is clean. `tsc -p ui/tsconfig.json` reports
only existing errors in
`ui/storybook/support/feature-map/FeatureOverview.tsx`, which this
change does not touch.
- A paid chat-eval run with this change is linked in a comment below.

## Risks

- Same-company links now carry the explicit prefix (`/COM/issues/COM-1`
instead of `/issues/COM-1` resolved to the same place). The route is the
same.
- An explicit link with a bare identifier as its target (`[x](JIRA-2)`)
is now gated by known company prefixes like auto-linked text. Such a
link to a non-Paperclip key stays an ordinary link.
- Low risk otherwise: no API, data or saved state changes.

## Model Used

- Claude Opus 5.5 (`claude-opus-5-5`), Anthropic, with tool use and code
execution in Claude Code.

## 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
- [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
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-11 23:51:54 -06:00

203 lines
6.9 KiB
TypeScript

import { extractCompanyPrefixFromPath } from "./company-routes.js";
type MarkdownNode = {
type: string;
value?: string;
url?: string;
children?: MarkdownNode[];
};
const BARE_ISSUE_IDENTIFIER_RE = /^[A-Z][A-Z0-9]*-\d+$/i;
const ISSUE_SCHEME_RE = /^issue:\/\/:?([^?#\s]+)([?#].*)?$/i;
const ISSUE_REFERENCE_TOKEN_RE = /issue:\/\/:?[^\s<>()]+|https?:\/\/[^\s<>()]+|\/(?:[^\s<>()/]+\/)*issues\/[A-Z][A-Z0-9]*-\d+(?=$|[\s<>)\],.;!?:])|\b[A-Z][A-Z0-9]*-\d+\b/gi;
export function parseIssuePathIdFromPath(pathOrUrl: string | null | undefined): string | null {
if (!pathOrUrl) return null;
const value = pathOrUrl.trim();
if (!value) return null;
if (/^https?:\/\//i.test(value)) return null;
const pathname = value.split(/[?#]/, 1)[0]!;
const segments = pathname.split("/").filter(Boolean);
const issueIndex = segments.findIndex((segment) => segment === "issues");
if (issueIndex === -1 || issueIndex === segments.length - 1) return null;
const issuePathId = decodeURIComponent(segments[issueIndex + 1] ?? "");
if (!issuePathId || issuePathId.startsWith(":")) return null;
return BARE_ISSUE_IDENTIFIER_RE.test(issuePathId) ? issuePathId.toUpperCase() : issuePathId;
}
export function parseIssueReferenceFromHref(
href: string | null | undefined,
knownPrefixes?: Set<string>,
) {
if (!href) return null;
const trimmed = href.trim();
const issueSchemeMatch = trimmed.match(ISSUE_SCHEME_RE);
if (issueSchemeMatch?.[1]) {
const issuePathId = decodeURIComponent(issueSchemeMatch[1]);
return {
issuePathId,
href: issueHref(issuePathId, issueSchemeMatch[2] ?? "", null, knownPrefixes),
};
}
const pathId = parseIssuePathIdFromPath(href);
if (pathId) {
return {
issuePathId: pathId,
href: issueHref(
pathId,
trimmed.match(/[?#].*$/)?.[0] ?? "",
extractCompanyPrefixFromPath(trimmed.split(/[?#]/, 1)[0]!),
knownPrefixes,
),
};
}
if (!BARE_ISSUE_IDENTIFIER_RE.test(trimmed)) return null;
const normalized = trimmed.toUpperCase();
// Only auto-link a bare IDENT-123 token when its prefix belongs to a known
// company. An empty or omitted set means "prefixes unknown" -> stay permissive
// (preserves behavior during initial load and in provider-less render
// surfaces). The issue:// scheme and /issues/ path forms handled above are
// never gated -- they are deliberate references, not accidental foreign keys.
if (knownPrefixes && knownPrefixes.size > 0) {
const prefix = normalized.split("-")[0];
if (!prefix || !knownPrefixes.has(prefix)) return null;
}
return {
issuePathId: normalized,
href: issueHref(normalized, "", null, knownPrefixes),
};
}
function issueHref(
issuePathId: string,
suffix: string,
pathCompanyPrefix: string | null,
knownPrefixes?: Set<string>,
): string {
const identifierPrefix = BARE_ISSUE_IDENTIFIER_RE.test(issuePathId)
? issuePathId.split("-")[0]!.toUpperCase()
: null;
const companyPrefix =
pathCompanyPrefix ??
(identifierPrefix && knownPrefixes?.has(identifierPrefix) ? identifierPrefix : null);
const path = `/issues/${encodeURIComponent(issuePathId)}${suffix}`;
return companyPrefix ? `/${companyPrefix}${path}` : path;
}
function splitTrailingPunctuation(token: string) {
let core = token;
let trailing = "";
while (core.length > 0) {
const lastChar = core.at(-1);
if (!lastChar || !/[),.;!?:\]]/.test(lastChar)) break;
if (lastChar === ")") {
const openCount = (core.match(/\(/g) ?? []).length;
const closeCount = (core.match(/\)/g) ?? []).length;
if (closeCount <= openCount) break;
}
if (lastChar === "]") {
const openCount = (core.match(/\[/g) ?? []).length;
const closeCount = (core.match(/\]/g) ?? []).length;
if (closeCount <= openCount) break;
}
trailing = `${lastChar}${trailing}`;
core = core.slice(0, -1);
}
return { core, trailing };
}
function createIssueLinkNode(value: string, href: string, childType: "text" | "inlineCode" = "text"): MarkdownNode {
return {
type: "link",
url: href,
children: [{ type: childType, value }],
};
}
function linkifyIssueReferencesInText(value: string, knownPrefixes?: Set<string>): MarkdownNode[] | null {
const nodes: MarkdownNode[] = [];
let cursor = 0;
let matched = false;
for (const match of value.matchAll(ISSUE_REFERENCE_TOKEN_RE)) {
const raw = match[0];
if (!raw) continue;
const start = match.index ?? 0;
const end = start + raw.length;
const { core, trailing } = splitTrailingPunctuation(raw);
const issueRef = parseIssueReferenceFromHref(core, knownPrefixes);
if (!issueRef) continue;
matched = true;
if (start > cursor) {
nodes.push({ type: "text", value: value.slice(cursor, start) });
}
nodes.push(createIssueLinkNode(core, issueRef.href));
if (trailing) {
nodes.push({ type: "text", value: trailing });
}
cursor = end;
}
if (!matched) return null;
if (cursor < value.length) {
nodes.push({ type: "text", value: value.slice(cursor) });
}
return nodes;
}
function rewriteMarkdownTree(node: MarkdownNode, knownPrefixes?: Set<string>) {
if (!Array.isArray(node.children) || node.children.length === 0) return;
if (node.type === "link" || node.type === "linkReference" || node.type === "code" || node.type === "definition" || node.type === "html") {
return;
}
const nextChildren: MarkdownNode[] = [];
for (const child of node.children) {
if (child.type === "inlineCode" && typeof child.value === "string") {
const issueRef = parseIssueReferenceFromHref(child.value, knownPrefixes);
if (issueRef) {
nextChildren.push(createIssueLinkNode(child.value, issueRef.href, "inlineCode"));
continue;
}
}
if (child.type === "text" && typeof child.value === "string") {
const linked = linkifyIssueReferencesInText(child.value, knownPrefixes);
if (linked) {
nextChildren.push(...linked);
continue;
}
}
rewriteMarkdownTree(child, knownPrefixes);
nextChildren.push(child);
}
node.children = nextChildren;
}
export interface RemarkLinkIssueReferencesOptions {
/**
* Company issue prefixes that are eligible for auto-linking. When provided
* and non-empty, a bare IDENT-123 token only becomes an issue link if its
* prefix is in this set -- this keeps foreign tracker keys (e.g. a Jira
* "TREE-604") from linking to non-existent Paperclip issues. Omit or pass an
* empty list to keep the legacy permissive behavior (link any IDENT-123).
*/
knownPrefixes?: string[];
}
export function remarkLinkIssueReferences(options?: RemarkLinkIssueReferencesOptions) {
const knownPrefixes = options?.knownPrefixes && options.knownPrefixes.length > 0
? new Set(options.knownPrefixes.map((prefix) => prefix.toUpperCase()))
: undefined;
return (tree: MarkdownNode) => {
rewriteMarkdownTree(tree, knownPrefixes);
};
}