Public: anyone can read this; the crew can change it.
/explore takes `updated:7d` (or 1d, 30d) in its one line, and a chip for each narrows to what moved in that window #5887
/explore now takes updated:1d, updated:7d or updated:30d, and the page has three chips (Today, This week, This month) that set it. I ran bun test test/explore test/archive-web, which passes, bun scripts/design-lint.ts, which passes, and the type check, which reports only the missing bun type definitions.
stopped parked by the crew deploy c0d1d89
3 earlier comments
/explore now takes updated:1d, updated:7d or updated:30d, and the page has three chips (Today, This week, This month) that set it. I ran bun test test/explore test/archive-web, which passes, bun scripts/design-lint.ts, which passes, and the type check, which reports only the missing bun type definitions. I did not run the full suite.
- Parsing:
exploreLinewrites the term back. A value other than those three is not ignored: it falls through as search text, the same waysort:nonsensealready does. That is what the existing parser does with unknown values, so I kept it. The issue's wording ("ignored, like an unknown sort") could be read the other way. - Filter: a row is kept only if its
shippedAtis inside the window.shippedAtis the row's last merged pull, which is the freshness number rows already carry. It does not read separate push, release or profile-change times, so a push that was never merged does not count as movement. A person's row uses their latest app'sshippedAt. Rows with noshippedAtare dropped when a window is set, and spawned-forge rows (admin only) are left out too, since they carry no time. - Chips: they are the existing
seg()parts, so there is no new component class. The active chip is marked and the chips keep the other terms in the line. Clicking the active chip does not clear it; you clear it by editing the line. - MCP and API: both go through
parseExploreQuery, so they take the same line. I addedupdated:1d|7d|30dto the MCPexploretool's description. - Tests: they cover parsing and write-back (including an unknown window), filtering seeded apps updated 2 and 20 days ago across all three windows, the canonical query string from
/api/v1/explore, and the chip links with the active one marked. I also addedupdated: "all"to the two existingtoEqualassertions on the parsed query.
To try it, open /explore?q=updated:7d, or GET /api/v1/explore?q=updated:7d.
I merged main into this branch and pushed. The explore and MCP tests pass (59 of 59), but I didn't run the full suite. bun x tsc --noEmit can't find the bun type definitions in this sandbox, so I only checked that it reports no other errors.
The only conflict was in src/actions/explore.ts, where main had added owner:<name> and this branch had added updated:1d|7d|30d. The merged query now carries both:
- Parsing and links: it parses both filters, and
exploreLinewrites both back out. - Spawned forges: they carry no time and belong to no account, so either filter leaves them out.
- Row filter: it applies the
ownerfilter and theupdatedwindow together.
I also changed the owner link on a person's row in src/web/studio/explore.ts. It builds an explore query by hand, and since updated is now a required field it came out as updated:undefined. It now sets updated: "all", so the link reads kind:app owner:bob again. No test was changed.
I merged main into the branch and pushed it. The only conflict was src/http/mcp.ts. I ran the explore and MCP tests, and they pass. The full suite timed out here, so I didn't see it finish. bun x tsc --noEmit failed on a missing bun type definition, so there is no type check.
- Conflict:
mainsplitsrc/http/mcp.tsinto one file per tool. I tookmain's version and moved my one change, theupdated:mention in thequerydescription, intosrc/http/mcp/tools/explore.ts. - Size budget: the
tools/listsize test failed by 17 bytes with that wording. I shortened it to "Words; kind: sort: updated:7d", and the test passes. The description now lists only the7dvalue, not1dor30d.
The pull adds an updated: window filter to Explore. It is reachable in the explore action, the studio Explore page and the MCP explore tool's query.
To try it, search Explore with updated:7d (or 1d, 30d).
🤖 Generated with Claude Code
Parked by the crew: the check stayed red through 2 automatic tries
/explorenow takesupdated:1d,updated:7dorupdated:30d, and the page has three chips (Today,This week,This month) that set it. I ranbun test test/explore test/archive-web, which passes,bun scripts/design-lint.ts, which passes, and the type check, which reports only the missingbuntype definitions. I did not run the full suite.exploreLinewrites the term back. A value other than those three is not ignored: it falls through as search text, the same waysort:nonsensealready does. That is what the existing parser does with unknown values, so I kept it. The issue's wording ("ignored, like an unknown sort") could be read the other way.shippedAtis inside the window.shippedAtis the row's last merged pull, which is the freshness number rows already carry. It does not read separate push, release or profile-change times, so a push that was never merged does not count as movement. A person's row uses their latest app'sshippedAt. Rows with noshippedAtare dropped when a window is set, and spawned-forge rows (admin only) are left out too, since they carry no time.seg()parts, so there is no new component class. The active chip is marked and the chips keep the other terms in the line. Clicking the active chip does not clear it; you clear it by editing the line.parseExploreQuery, so they take the same line. I addedupdated:1d|7d|30dto the MCPexploretool's description./api/v1/explore, and the chip links with the active one marked. I also addedupdated: "all"to the two existingtoEqualassertions on the parsed query.To try it, open
/explore?q=updated:7d, orGET /api/v1/explore?q=updated:7d.Closes #5874