Skip to content

The ultimate harness for coding agents and beyond. All your agents. All your machines. One command center. Start with code, then follow your curiosity and build across disciplines: CAD, circuits, robots, games and music.

fix(jev-browser): stop walking past the list, and never arrive empty-handed (#154)

Dee45c3c828cbcf
diff --git a/store/agents/jev-browser/viewer/crawl.mjs b/store/agents/jev-browser/viewer/crawl.mjs
index 39d78ee54..ae3b26703 100644
--- a/store/agents/jev-browser/viewer/crawl.mjs
+++ b/store/agents/jev-browser/viewer/crawl.mjs
@@ -150,20 +150,26 @@ export async function findThePage({ chrome, ask, start, want, search = '', maxSt
for (const l of links) {
q[`g${l.n}`] = jev.noul(`The person asked for: "${asked}". This page has a link reading "${l.label}" going to ${l.path}. Would following it get closer to what they asked for?`,
{ true: 'it leads towards what they asked for', false: 'it leads away: an unrelated section, a login, a legal page, or an advert' })
+ // The difference between "Travel" and "It's Only the Himalayas": one is a shelf, one is a book.
+ q[`s${l.n}`] = jev.noul(`Is the link "${l.label}" (${l.path}) a section, category or search that would show MANY of them, rather than one single one?`,
+ { true: 'it opens a page listing many of them', false: 'it opens one single one, in detail' })
}
if (box && !searched && search) q.usesearch = jev.noul(`This page has a search box. The person asked for: "${asked}". Is searching this site the best way to find them, rather than following a link?`)
const res = await ask({ state: pageState(page, { what_they_want: asked }), questions: q })
const here = res.answers.here?.noul ?? 0
const one = res.answers.one?.noul ?? 0
- const ranked = links.map((l) => ({ l, p: res.answers[`g${l.n}`]?.noul ?? 0 })).sort((a, b) => b.p - a.p)
+ const ranked = links.map((l) => ({ l, p: res.answers[`g${l.n}`]?.noul ?? 0, section: res.answers[`s${l.n}`]?.noul ?? 0 })).sort((a, b) => b.p - a.p)
const best = ranked[0]
+ // On a page that already lists them, only a NARROWER section is worth following. Following an
+ // individual one from here lands on a single product with nothing left to collect.
+ const narrower = ranked.find((r) => r.p >= 0.7 && r.section >= 0.6)
say('judged', { url: page.url, judged: links.length, found: 0, latencyMs: res.latencyMs,
links: ranked.map(({ l, p }) => ({ n: l.n, label: l.label, p })).slice(0, 40) })
// A front page of a bookshop IS a list of books, so "does this page show them" says yes and the
// walk stops one click short of the travel section. A link the person's own words point
// straight at beats a page that merely qualifies.
- const obvious = best && best.p >= 0.75 && here < 0.85
+ const obvious = !!narrower && here < 0.85
if ((here >= 0.5 && !obvious) || one >= 0.6) { say('arrived', { url: page.url, listing: here >= 0.5, steps: step + 1 }); return { url: page.url, page, steps: step + 1, isOne: one >= 0.6 && here < 0.5 } }
if (search && box && !searched && (res.answers.usesearch?.noul ?? 0) >= 0.5) {
@@ -173,13 +179,14 @@ export async function findThePage({ chrome, ask, start, want, search = '', maxSt
url = await chrome.url()
continue
}
- if (!best || best.p < 0.5) {
+ const go = here >= 0.5 ? narrower : (narrower ?? best)
+ if (!go || go.p < 0.5) {
// Nothing points onward. If this page at least shows the right sort of thing, work from it.
if (here >= 0.5) { say('arrived', { url: page.url, listing: true, steps: step + 1 }); return { url: page.url, page, steps: step + 1 } }
return { url: page.url, page, steps: step + 1, gaveUp: 'no link on this page looked like the way to it' }
}
- say('step', { label: best.l.label, p: best.p })
- url = best.l.url
+ say('step', { label: go.l.label, p: go.p })
+ url = go.l.url
}
return { url, page, steps: maxSteps, gaveUp: `it was still looking after ${maxSteps} pages` }
}
@@ -230,7 +237,10 @@ export async function proposeJob({ chrome, ask, start, want = '', search = '', p
// CALL 1 — what does this page list, and which of its links are the things?
const wanted = want ? ` The person asked for: "${want}".` : ''
const links = list.links.slice(0, LIMITS.linksPerCall)
- const q1 = { kind: jev.choice(THINGS, `This page may list many of one kind of thing.${wanted} What is it listing?`) }
+ const q1 = {
+ kind: jev.choice(THINGS, `This page may list many of one kind of thing.${wanted} What is it listing?`),
+ single: jev.noul(`${wanted} Is THIS page one single one of those, shown in detail, rather than a list of many of them?`),
+ }
for (const l of links) {
q1[`l${l.n}`] = jev.noul(`"${l.label}" is the text of a link on this page, pointing at ${l.path}. Is "${l.label}" the title of one of the things this page lists, rather than part of the site's own menu?`)
}
@@ -242,14 +252,20 @@ export async function proposeJob({ chrome, ask, start, want = '', search = '', p
const surest = [...items].sort((a, b) => b.p - a.p)[0]
const item = ITEM_WORDS[kind] ?? ITEM_WORDS.other
say('judged', { url: list.url, kind, judged: links.length, found: items.length, latencyMs: r1.latencyMs, links: links.map((l) => ({ n: l.n, label: l.label, p: r1.answers[`l${l.n}`]?.noul ?? 0 })).sort((a, b) => b.p - a.p).slice(0, 40) })
- if (!items.length) return { kind, item, columns: [], from: list.url, none: 'no link on this page looks like one of the things' }
+ const single = (r1.answers.single?.noul ?? 0) >= 0.6
+ if (!items.length && !single) return { kind, item, columns: [], from: list.url, none: 'no link on this page looks like one of the things' }
- // CALL 2 — on one of them, which pieces of text deserve a column, and what is each one?
- say('going', { url: surest.url, what: `one ${item}, to see what it says` })
- await chrome.go(surest.url)
- const one = await readPage(chrome)
- const itemWall = wallReason(one)
- if (itemWall) return { kind, item, columns: [], from: one.url, walled: `${new URL(one.url).hostname}: ${itemWall}` }
+ // CALL 2 — which pieces of text deserve a column. Usually on one of the things this page links
+ // to; but a walk can land ON one of them, and then this page is the one to read.
+ let one = list, onlyThis = true
+ if (items.length) {
+ onlyThis = false
+ say('going', { url: surest.url, what: `one ${item}, to see what it says` })
+ await chrome.go(surest.url)
+ one = await readPage(chrome)
+ const itemWall = wallReason(one)
+ if (itemWall) return { kind, item, columns: [], from: one.url, walled: `${new URL(one.url).hostname}: ${itemWall}` }
+ } else say('judged', { url: list.url, kind, judged: 0, found: 1, latencyMs: r1.latencyMs, links: [] })
const q2 = {}
for (const b of one.blocks) {
const text = b.text.slice(0, 140)
@@ -296,7 +312,7 @@ export async function proposeJob({ chrome, ask, start, want = '', search = '', p
}
columns.sort((a, b) => b.p - a.p)
say('proposed', { columns: columns.length, latencyMs: r1.latencyMs + r2.latencyMs })
- return { kind, item, columns: columns.slice(0, LIMITS.fields), from: one.url, ms: Math.round(r1.latencyMs + r2.latencyMs) }
+ return { kind, item, columns: columns.slice(0, LIMITS.fields), from: one.url, onlyThis, ms: Math.round(r1.latencyMs + r2.latencyMs) }
}
/**