commit c9c0207c83bfa6294f9f559c95ae8acfe93c76fa
parent fa2f521343c6c9a2b4a94e7b9871d66926ea8c4b
Author: I Mean I'm Just Saying <imeanimjustsaying@kiwifarms.st>
Date: Sat, 3 Oct 2026 13:20:25 -0400
mcp: a search date bound is normalised, or refused — never silently dropped
date_from/date_to accepted only YYYYMMDD; any other spelling became a footer
warning and the search ran unbounded. normalizeDateArg (mcp/src/dateArg.ts)
accepts YYYYMMDD, YYYY-MM-DD (also / or .) and ISO timestamps, and returns
null for anything else, which parseSearchArgs turns into an error result:
nothing is searched.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Diffstat:
4 files changed, 82 insertions(+), 18 deletions(-)
diff --git a/editor/CHANGELOG.md b/editor/CHANGELOG.md
@@ -1,6 +1,7 @@
# Changelog
## [Unreleased]
+- **The MCP's search tools take `date_from` and `date_to` as `2024-10-26` as well as `20241026`, and refuse a date they cannot read.** `search_transcripts` and `enumerate_matches` used to accept only `YYYYMMDD`: any other spelling was dropped with a footer warning and the search ran with no date bound, so a whole-corpus count could be read as the bounded one. Dashed, slashed and dotted dates and ISO timestamps are now normalised, and anything else is an error and nothing is searched.
- **`pnpm ops transcribe-bucket` transcribes a channel's downloaded-but-untranscribed videos**, as the channel page's **Transcribe N downloaded** button does, on the transcription queue. `"ids"` runs only some of them; each must be in the bucket, and a stray id is refused by name. `retry-bucket` is not the way to do this: it retries downloads, and counts a video whose audio is on disk as complete.
- **`pnpm ops retry-bucket` can run part of a bucket.** Its body takes `"ids"`, a list of video ids, and runs only those, as ticking them on the bucket's card does. Every id must be in the named bucket: one that is not is refused with a 400 naming it, and nothing runs. A job started with `ids` is not replayable, like a checkbox selection in the UI.
- **An X post fetch keeps what it has read when it is cancelled, times out or fails part-way, and the next fetch picks up where it stopped.** The gallery-dl fetcher used to receive an account's posts all at once when gallery-dl finished, so a long fetch that X's rate limit held past the 30-minute limit — or one you cancelled — ended with nothing saved. gallery-dl now hands over each post as it reads it: the fetch saves posts every 200 posts or every minute along with gallery-dl's own resume point, and the next fetch of the channel continues from that point instead of starting again. A fetch of new posts stops once it reaches 100 already-archived posts in a row instead of reading the whole timeline, and a fetch of an account's history may run for up to 3 hours (new-post fetches keep the 30-minute limit). gallery-dl's rate-limit waits now appear in the job's log as they happen.
diff --git a/mcp/src/dateArg.test.ts b/mcp/src/dateArg.test.ts
@@ -0,0 +1,27 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { normalizeDateArg } from "./dateArg";
+
+test("normalizeDateArg: YYYYMMDD passes through", () => {
+ assert.equal(normalizeDateArg("20241026"), "20241026");
+});
+
+test("normalizeDateArg: dashed, slashed, dotted and ISO timestamps normalise", () => {
+ assert.equal(normalizeDateArg("2024-10-26"), "20241026");
+ assert.equal(normalizeDateArg(" 2024/10/26 "), "20241026");
+ assert.equal(normalizeDateArg("2024.10.26"), "20241026");
+ assert.equal(normalizeDateArg("2024-10-26T00:00:00Z"), "20241026");
+ assert.equal(normalizeDateArg("2024-10-26T12:30:00+02:00"), "20241026");
+});
+
+test("normalizeDateArg: absent or blank is no bound", () => {
+ assert.equal(normalizeDateArg(undefined), undefined);
+ assert.equal(normalizeDateArg(null), undefined);
+ assert.equal(normalizeDateArg(" "), undefined);
+});
+
+test("normalizeDateArg: anything else is refused, never read as no bound", () => {
+ for (const bad of ["2024", "202410", "2024-1-5", "Oct 26 2024", "26/10/2024", "20241332", "20240230", 20241026, "yesterday"]) {
+ assert.equal(normalizeDateArg(bad), null, String(bad));
+ }
+});
diff --git a/mcp/src/dateArg.ts b/mcp/src/dateArg.ts
@@ -0,0 +1,29 @@
+// An upload-date bound as a tool caller may spell it, normalised to the index's
+// YYYYMMDD.
+//
+// THE INDEX COMPARES YYYYMMDD LEXICOGRAPHICALLY, so a date in any other shape
+// cannot be used as given. It used to be dropped with a footer warning, and the
+// search ran with NO bound — a complete-set count over the whole corpus that
+// read exactly like the bounded count the caller asked for. Now the common
+// spellings are accepted, and anything else is refused (`null`) so the caller
+// gets an error and nothing is searched.
+//
+// Returns the YYYYMMDD string, `undefined` for an absent or blank value, or
+// `null` for a value that is not a date.
+export function normalizeDateArg(raw: unknown): string | undefined | null {
+ if (raw === undefined || raw === null) return undefined;
+ if (typeof raw !== "string") return null;
+ const v = raw.trim();
+ if (!v) return undefined;
+ const m =
+ /^(\d{4})(\d{2})(\d{2})$/.exec(v) ??
+ /^(\d{4})[-/.](\d{2})[-/.](\d{2})(?:[T ][\d:.]+(?:Z|[+-]\d{2}:?\d{2})?)?$/.exec(v);
+ if (!m) return null;
+ const [, y, mo, d] = m;
+ const month = Number(mo);
+ const day = Number(d);
+ if (month < 1 || month > 12 || day < 1) return null;
+ const last = new Date(Date.UTC(Number(y), month, 0)).getUTCDate();
+ if (day > last) return null;
+ return `${y}${mo}${d}`;
+}
diff --git a/mcp/src/server.ts b/mcp/src/server.ts
@@ -1,3 +1,4 @@
+import { normalizeDateArg } from "./dateArg";
import {
Server,
type Prompt,
@@ -245,11 +246,11 @@ const SEARCH_FILTER_ARGS = {
},
date_from: {
type: "string",
- description: "Keep only videos uploaded on or after this date (YYYYMMDD).",
+ description: "Keep only videos uploaded on or after this date (YYYYMMDD or YYYY-MM-DD; anything else is an error).",
},
date_to: {
type: "string",
- description: "Keep only videos uploaded on or before this date (YYYYMMDD).",
+ description: "Keep only videos uploaded on or before this date (YYYYMMDD or YYYY-MM-DD; anything else is an error).",
},
media_type: {
type: "string",
@@ -1426,6 +1427,7 @@ async function handleSearch(
const includeSnippets = args.include_snippets !== false;
const base = linkStyleOf(args) === "base";
const parsed = parseSearchArgs(args);
+ if (parsed.error) return errorText(parsed.error);
const result = await searchTranscripts(source, {
query,
filters: parsed.filters,
@@ -1604,6 +1606,7 @@ async function handleEnumerateMatches(
const batchSize = batchSizeArg >= 1 ? batchSizeArg : 8;
const parsed = parseSearchArgs(args);
+ if (parsed.error) return errorText(parsed.error);
const result = await searchTranscripts(source, {
query,
filters: parsed.filters,
@@ -1710,6 +1713,9 @@ type ParsedSearchArgs = {
scopes?: LayerScope[];
collapseDuplicates: boolean;
warnings: string[];
+ // Set when an argument must stop the call rather than be dropped (a date
+ // bound that is not a date). Callers return it as an error and search nothing.
+ error?: string;
};
const SCOPE_TOKENS: ReadonlyArray<LayerScope> = [
@@ -1753,20 +1759,20 @@ function parseSearchArgs(args: Record<string, unknown>): ParsedSearchArgs {
warnings.push(`unknown age ignored: ${String(args.age)}`);
}
- const dateFrom = typeof args.date_from === "string" ? args.date_from.trim() : "";
- const dateTo = typeof args.date_to === "string" ? args.date_to.trim() : "";
- for (const [name, v] of [
- ["date_from", dateFrom],
- ["date_to", dateTo],
- ] as const) {
- // The comparison is lexicographic on YYYYMMDD, so a differently-shaped date
- // wouldn't error — it would quietly filter wrongly. Say so instead.
- if (v && !/^\d{8}$/.test(v)) {
- warnings.push(`${name}="${v}" is not YYYYMMDD — it was ignored`);
- }
- }
- const from = /^\d{8}$/.test(dateFrom) ? dateFrom : undefined;
- const to = /^\d{8}$/.test(dateTo) ? dateTo : undefined;
+ // A date bound the caller cannot have meant is an ERROR, not a warning: an
+ // ignored bound widens the search to the whole corpus and the count reads as
+ // the bounded one (see dateArg.ts). The common spellings are normalised.
+ const from = normalizeDateArg(args.date_from);
+ const to = normalizeDateArg(args.date_to);
+ const badDates = [
+ ["date_from", args.date_from, from],
+ ["date_to", args.date_to, to],
+ ]
+ .filter(([, , v]) => v === null)
+ .map(([name, raw]) => `${name}=${JSON.stringify(raw)}`);
+ const error = badDates.length
+ ? `not a date: ${badDates.join(", ")} — use YYYYMMDD or YYYY-MM-DD. Nothing was searched.`
+ : undefined;
// Curated tags. A malformed id is REPORTED, never silently dropped: dropping
// the only tag in the list would widen the search back to the whole corpus
@@ -1794,8 +1800,8 @@ function parseSearchArgs(args: Record<string, unknown>): ParsedSearchArgs {
states !== undefined ||
mediaType !== undefined ||
age !== undefined ||
- from !== undefined ||
- to !== undefined ||
+ Boolean(from) ||
+ Boolean(to) ||
tags !== undefined;
const filters: SearchFilters | null = anyFilter
@@ -1835,6 +1841,7 @@ function parseSearchArgs(args: Record<string, unknown>): ParsedSearchArgs {
scopes,
collapseDuplicates: args.collapse_duplicates !== false,
warnings,
+ ...(error ? { error } : {}),
};
}
