@@ -37,3 +37,15 @@ test("a whole timeline's latest review is read on the commit the sidebar names",
   serve([approval, comment]);
   expect(JSON.parse((await pageReads().answer(`/repos/${REPO}/pulls/1/reviews`))!)).toEqual([{ user: { login: "x" }, state: "APPROVED", submitted_at: "2026-10-01T00:00:00Z", commit_id: HEAD }]);
 });
+
+test("open pull requests into a base are listed from the dashboard filtered on it, and none open reads as none", async () => {
+  const asked: string[] = [];
+  globalThis.fetch = async function dashboard(url) {
+    asked.push(decodeURIComponent(String(url)));
+    const route = String(url).includes("base") ? { currentPage: 1, openCount: 0, closedCount: 0 } : { results: [{ number: 4, title: "T", displayState: "REPO_PULL_REQUEST_DISPLAY_STATE_OPEN", headSha: HEAD, labels: [] }], totalPages: 1, openCount: 1 };
+    return new Response(JSON.stringify({ payload: { repoPullsDashboardContentRoute: route } }));
+  } as typeof fetch;
+  expect(JSON.parse((await pageReads().answer(`/repos/${REPO}/pulls?state=open&base=main&per_page=100`))!)).toEqual([]);
+  expect(JSON.parse((await pageReads().answer(`/repos/${REPO}/pulls?state=open&per_page=100`))!)).toMatchObject([{ number: 4 }]);
+  expect(asked[0]).toContain('q=is:pr is:open base:"main"');
+});
@@ -203,11 +203,14 @@ export function pageReads(): PageReads {
     }));
   }
 
-  /** Every open pull request, or only the newest page of GitHub's list when `newest` asks for no more than one page. */
-  async function openPulls(repo: string, newest: boolean) {
+  /** Every open pull request, or only those into `base` when it is set, or only the newest page of GitHub's list when `newest` asks for no more than one page. */
+  async function openPulls(repo: string, newest: boolean, base: string | null) {
     const rows: Array<{ number: number; title: string }> = [];
+    const query = base === null ? "is:pr is:open" : `is:pr is:open base:"${base}"`;
     for (let page = 1; ; page++) {
-      const route = (await json(`/${repo}/pulls?q=${encodeURIComponent("is:pr is:open")}&page=${page}`)).payload.repoPullsDashboardContentRoute;
+      const route = (await json(`/${repo}/pulls?q=${encodeURIComponent(query)}&page=${page}`)).payload.repoPullsDashboardContentRoute;
+      // With nothing open the route leaves out its results and pages.
+      if (route.openCount === 0) return rows;
       rows.push(...(route.results as any[]).map((row) => ({ number: row.number, title: row.title, draft: row.displayState === "REPO_PULL_REQUEST_DISPLAY_STATE_DRAFT", head: { sha: row.headSha }, labels: row.labels })));
       if (newest || page >= route.totalPages) return rows;
     }
@@ -254,7 +257,7 @@ export function pageReads(): PageReads {
     if ((match = COMMITS.exec(pathname))) return commits(match[1]!, Number(match[2]));
     if ((match = ISSUE_COMMENTS.exec(pathname))) return issueComments(match[1]!, Number(match[2]));
     // The dashboard filters by neither branch, so a stack's filtered lists are left to the API.
-    if ((match = OPEN_PULLS.exec(pathname)) && searchParams.get("state") === "open" && !searchParams.has("head") && !searchParams.has("base")) return openPulls(match[1]!, Number(searchParams.get("per_page") ?? 30) <= DASHBOARD_PAGE);
+    if ((match = OPEN_PULLS.exec(pathname)) && searchParams.get("state") === "open" && !searchParams.has("head")) return openPulls(match[1]!, Number(searchParams.get("per_page") ?? 30) <= DASHBOARD_PAGE, searchParams.get("base"));
     if ((match = CHECK_RUNS.exec(pathname))) return checkRuns(match[1]!, match[2]!);
     if ((match = FOLDER.exec(pathname)) && !searchParams.has("ref")) return folder(match[1]!, decodeURIComponent(match[2] ?? ""));
     return null;
 
@@ -3,7 +3,8 @@ import { expect, test } from "bun:test";
 import { randomBytes } from "node:crypto";
 import { sql } from "drizzle-orm";
 import { createDB, runMigrations, schema } from "@intx/db";
-import { diffHash, type PullRequestToReview } from "./pull-request-reader.js";
+import { diffHash } from "@corbits/common/diff-hash";
+import type { PullRequestToReview } from "./pull-request-reader.js";
 import { createCodeReviewRuns } from "./code-review-runs.js";
 import { recordEvent, type PullRequestRef } from "./manager-state.js";
 import { migrateTriageTables } from "./triage-db.js";
 
@@ -5,7 +5,8 @@ import { eq } from "drizzle-orm";
 import { type } from "arktype";
 import { schema, type DB } from "@intx/db";
 import { CodeReviewComment, DiffCut, REPO_SOURCES, repoSourceOf, StageResult, type RepoSource } from "@corbits/triage-contracts";
-import { diffHash, type PullRequestToReview } from "./pull-request-reader.js";
+import { diffHash } from "@corbits/common/diff-hash";
+import type { PullRequestToReview } from "./pull-request-reader.js";
 import { repoRecords, sourceOf, triageNs, type RepoRecord } from "./github/tenant-config.js";
 import type { Member } from "./integrations/app.js";
 import { codeReviews, type CodeReviewRecord } from "./manager-state.js";
 
@@ -1,5 +1,6 @@
 import { expect, test } from "bun:test";
-import { createPullRequestReader, cutDiff, diffHash, reviewRequestOf, type PullRequestToReview } from "./pull-request-reader.js";
+import { diffHash } from "@corbits/common/diff-hash";
+import { createPullRequestReader, cutDiff, reviewRequestOf, type PullRequestToReview } from "./pull-request-reader.js";
 import type { RepoRecord } from "./github/tenant-config.js";
 
 const RECORD: RepoRecord = { name: "acme/widgets", connected: true };
 
@@ -1,12 +1,12 @@
 // Reads a pull request as its host shows it, alone or with its diff, and builds the code review request a manager's
 // review reads: the diff cut to the limit and hashed whole, so a later head with the same diff can take a review over
 // (code-review-runs.ts).
-import { createHash } from "node:crypto";
 import * as bitbucket from "@corbits/bitbucket-datacenter-reads";
 import { readPr, readPrDiff, type GithubFetch } from "@corbits/github-reads";
 import { shownBody, visibleText, type RepoSource } from "@corbits/triage-contracts";
 import type { TenantBitbucket } from "./bitbucket-datacenter/tenant-bitbucket.js";
 import { filesFrom } from "@corbits/common/diff-files";
+import { diffHash } from "@corbits/common/diff-hash";
 import { forInstallation } from "./github/open-pulls.js";
 import { sourceOf, type RepoRecord } from "./github/tenant-config.js";
 
@@ -21,15 +21,6 @@ export type PullRequestHead = { title: string; description: string; headSha: str
 
 export type PullRequestToReview = PullRequestHead & { diff: string };
 
-/**
- * The diff's hash, its `index` lines' blob ids included, so a base change to a file the diff touches changes it; undefined
- * for a diff with no `index` line, such as one rebuilt from the file list, which can't show such a change.
- */
-export function diffHash(diff: string): string | undefined {
-  if (!diff.split("\n").some((line) => line.startsWith("index "))) return undefined;
-  return createHash("sha256").update(diff).digest("hex");
-}
-
 function lineCount(text: string): number {
   return text.split("\n").length - (text.endsWith("\n") ? 1 : 0);
 }
 
@@ -389,6 +389,21 @@ label off that head, so it never stands on a head not yet triaged.
    own changes and focused runs on them, the comment noting what it is
    stacked on. A branch other pull requests were merged into holds their
    changes too, so focused is skipped and the comment says why.
+   With a duplicate rule on, it compares the open pull requests opened before
+   this one on its base branch, drafts left out (`list_open_prs` with `base`,
+   one call per base branch); past 100 of them it reads none and both rules
+   are unconfirmed. It reads the first page of each one's changed paths
+   (`list_pr_files` without patches, ten at a time), so its cost is one call
+   per earlier open pull request. One with the same head commit repeats this
+   one with no further read. For no-duplicate it reads the diff hash
+   (`pr_diff_hash`, the hash the hub's code review takeover uses,
+   `@corbits/common/diff-hash`) and commits only of those whose first page
+   lists no path this one does not change, and of this one when any does: two
+   calls each, usually none. For duplicates-open-pull-request it keeps only
+   the shortlist the judge is asked about (`overlapping`,
+   `logic/duplicates.ts`): up to five that share a path on that first page,
+   lockfiles left out, or two title words, most shared paths first
+   (`openPulls`).
 2. **rules**: runs the hard rules (built-in rules over host data, text
    rules and code-pattern rules) and, unless one failed or the pull request is a draft or closed,
    builds the judged rules' questions, or the verdict question when there are