~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.
Python academiaanthropicarxivbravedeep-researchencryptionhome-automationhomeserverlocallocal-deep-researchlocal-llmmistralollamaopenaipubmedresearchresearch-toolretrieval-augmented-generationsearxngself-hosted
/**
* Shared per-test fixture helpers for library / collection / news-subscription
* UI tests.
*
* Seed and clean up a collection or a news subscription via the synchronous API
* so tests don't depend on pre-existing DB state. These mirror the helpers proven in
* test_crud_operations_ci.js (#4174/#4180/#4187); they live here now that a
* second test file needs them. Each cleanup wraps page.evaluate
* in a Node-side try/catch so a torn-down page during teardown can never throw
* and mask a test result.
*
* All helpers read the CSRF token from the current page's meta tag, so the page
* must already be on a same-origin app page before calling them.
*/
async function seedCollection(page, { agentEnabled } = {}) {
const r = await page.evaluate(async (agentEnabledArg) => {
const csrf = document.querySelector('meta[name="csrf-token"]')?.content;
// Best-effort uniqueness (not guaranteed): timestamp + random suffix.
const name = `ldr-ui-test-collection-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;
const body = { name, description: 'UI test fixture', type: 'user_uploads' };
// Only send agent_enabled when the caller explicitly asked for
// it — omitting the key keeps the historical default (True) for
// every existing call site.
if (agentEnabledArg !== undefined && agentEnabledArg !== null) {
body.agent_enabled = agentEnabledArg;
}
const res = await fetch('/library/api/collections', {
method: 'POST',
credentials: 'same-origin',
headers: { 'Content-Type': 'application/json', 'X-CSRFToken': csrf || '' },
body: JSON.stringify(body),
});
const respBody = await res.json().catch(() => ({}));
return { ok: res.ok, name, success: respBody?.success === true, id: respBody?.collection?.id };
}, agentEnabled);
return r.ok && r.success && r.id ? { id: r.id, name: r.name } : null;
}
async function deleteCollection(page, collectionId) {
if (!collectionId) return;
try {
await page.evaluate(async (id) => {
const csrf = document.querySelector('meta[name="csrf-token"]')?.content;
try {
await fetch(`/library/api/collections/${id}`, {
method: 'DELETE',
credentials: 'same-origin',
headers: { 'X-CSRFToken': csrf || '' },
});
} catch { /* swallow fetch errors */ }
}, collectionId);
} catch { /* swallow page.evaluate errors so cleanup never masks the test result */ }
}
async function seedSubscription(page, { isActive = true } = {}) {
const r = await page.evaluate(async (active) => {
const csrf = document.querySelector('meta[name="csrf-token"]')?.content;
// Best-effort uniqueness (not guaranteed): timestamp + random suffix.
const query = `ldr-ui-test-sub-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;
const res = await fetch('/news/api/subscribe', {
method: 'POST',
credentials: 'same-origin',
headers: { 'Content-Type': 'application/json', 'X-CSRFToken': csrf || '' },
body: JSON.stringify({ query, subscription_type: 'search', is_active: active }),
});
const body = await res.json().catch(() => ({}));
return { ok: res.ok, status: res.status, query, id: body?.subscription_id };
}, isActive);
return r.ok && r.id ? { id: r.id, query: r.query } : null;
}
async function deleteSubscription(page, subscriptionId) {
if (!subscriptionId) return;
try {
await page.evaluate(async (id) => {
const csrf = document.querySelector('meta[name="csrf-token"]')?.content;
try {
await fetch(`/news/api/subscriptions/${id}`, {
method: 'DELETE',
credentials: 'same-origin',
headers: { 'X-CSRFToken': csrf || '' },
});
} catch { /* swallow fetch errors */ }
}, subscriptionId);
} catch { /* swallow page.evaluate errors so cleanup never masks the test result */ }
}
// NOTE: a document-upload helper lives in test_crud_operations_ci.js. It is
// intentionally NOT exported here yet: the only consumer (documentCardActions)
// is deferred until it can clean up the uploaded doc (collection delete doesn't
// cascade to docs — see test_library_documents_ci.js). When that lands, add
// uploadFixtureDocument here returning the document id(s) so callers can delete
// the doc, not just the collection.
module.exports = { seedCollection, deleteCollection, seedSubscription, deleteSubscription };