refactor content fetching and embedding processes for improved concurrency and error handling

This commit is contained in:
ImBenji
2026-04-19 23:21:29 +01:00
parent 6bf3a9282f
commit 37d9dfb083
8 changed files with 176 additions and 10 deletions
+4 -4
View File
@@ -34,7 +34,7 @@ function buildArticlesQuery(query) {
conditions.push('content IS NOT NULL AND content != \'\'');
conditions.push('is_index_page = 0');
conditions.push('EXISTS (SELECT 1 FROM article_embeddings WHERE article_id = articles.id)');
conditions.push('has_embedding = 1');
const whereClause = `WHERE ${conditions.join(' AND ')}`;
const limit = Number.parseInt(query.limit, 10);
@@ -48,7 +48,7 @@ function buildArticlesQuery(query) {
SELECT id, title, description, content, ${includeEmbedding ? 'embedding,' : ''} url, normalized_title, source, pub_date, ingested_at
FROM articles
${whereClause}
ORDER BY COALESCE(pub_date, ingested_at) DESC, id DESC
ORDER BY pub_date_effective DESC, id DESC
LIMIT ? OFFSET ?
`,
params,
@@ -76,7 +76,7 @@ function mapNeighborsToArticles(neighbors, excludeIndexPages, limit) {
FROM articles
WHERE id IN (${placeholders})
AND content IS NOT NULL AND content != ''
AND EXISTS (SELECT 1 FROM article_embeddings WHERE article_id = articles.id)
AND has_embedding = 1
${excludeIndexPages ? 'AND is_index_page = 0' : ''}
`).all(...ids);
const byId = new Map(articles.map((article) => [article.id, article]));
@@ -149,7 +149,7 @@ async function articleRoutes(fastify) {
WHERE id = ?
AND content IS NOT NULL AND content != ''
AND is_index_page = 0
AND EXISTS (SELECT 1 FROM article_embeddings WHERE article_id = articles.id)
AND has_embedding = 1
`).get(request.params.id);
if (!article) {
+110
View File
@@ -0,0 +1,110 @@
const db = require('../db');
const { getSourceCatalog } = require('../sources/sourceCatalog');
async function sourcesRoutes(fastify) {
fastify.get('/sources', async () => {
const catalog = getSourceCatalog();
// bucket every article by source string and content_status so we can
// answer "why is this source unusable" without N queries
const statusRows = db.prepare(`
SELECT
source,
COUNT(*) AS total,
SUM(CASE WHEN content_status = 'ready' THEN 1 ELSE 0 END) AS ready,
SUM(CASE WHEN content_status = 'skipped' THEN 1 ELSE 0 END) AS skipped,
SUM(CASE WHEN content_status = 'failed' THEN 1 ELSE 0 END) AS failed,
SUM(CASE WHEN content_status = 'pending' THEN 1 ELSE 0 END) AS pending,
SUM(CASE WHEN content_status IS NULL THEN 1 ELSE 0 END) AS untried,
SUM(CASE
WHEN content IS NOT NULL AND content != ''
AND is_index_page = 0
AND has_embedding = 1
THEN 1 ELSE 0
END) AS usable
FROM articles
GROUP BY source
`).all();
const statusByLabel = new Map();
for (const row of statusRows) {
// source can be "rss:Al Jazeera", "gdelt:Al Jazeera", or just "alphavantage"
const idx = row.source.indexOf(':');
const label = idx >= 0 ? row.source.slice(idx + 1) : row.source;
const feed = idx >= 0 ? row.source.slice(0, idx) : row.source;
if (!statusByLabel.has(label)) statusByLabel.set(label, []);
statusByLabel.get(label).push({ feed, ...row });
}
const policyRows = db.prepare(`
SELECT domain, policy, consecutive_plain_failures, consecutive_browser_failures,
plain_success_count, browser_success_count, expires_at, updated_at
FROM domain_fetch_policy
`).all();
const policyByDomain = new Map(policyRows.map((r) => [r.domain, r]));
return catalog.map((s) => {
const buckets = statusByLabel.get(s.label) || [];
const counts = buckets.reduce(
(acc, b) => {
acc.total += b.total;
acc.ready += b.ready;
acc.skipped += b.skipped;
acc.failed += b.failed;
acc.pending += b.pending;
acc.untried += b.untried;
acc.usable += b.usable;
return acc;
},
{ total: 0, ready: 0, skipped: 0, failed: 0, pending: 0, untried: 0, usable: 0 }
);
const domains = s.website.map((d) => {
const row = policyByDomain.get(d);
if (!row) {
return { domain: d, policy: 'auto' };
}
return {
domain: d,
policy: row.policy,
plainFailures: row.consecutive_plain_failures,
browserFailures: row.consecutive_browser_failures,
plainSuccesses: row.plain_success_count,
browserSuccesses: row.browser_success_count,
expiresAt: row.expires_at,
updatedAt: row.updated_at,
};
});
return {
id: s.id,
label: s.label,
websites: s.website,
backfill: s.backfill,
feeds: s.feedUrls,
counts,
byFeed: buckets.map((b) => ({
feed: b.feed,
total: b.total,
ready: b.ready,
skipped: b.skipped,
failed: b.failed,
pending: b.pending,
untried: b.untried,
usable: b.usable,
})),
domains,
};
});
});
}
module.exports = sourcesRoutes;
+15 -2
View File
@@ -1,8 +1,17 @@
const db = require('../db');
const { getLastIngestionBySource } = require('../state');
let statusCache = null;
let statusCacheAt = 0;
const STATUS_CACHE_TTL_MS = 30 * 1000;
async function statusRoutes(fastify) {
fastify.get('/status', async () => {
const now = Date.now();
if (statusCache && now - statusCacheAt < STATUS_CACHE_TTL_MS) {
return statusCache;
}
const bySourceRows = db.prepare(`
SELECT
source,
@@ -11,7 +20,7 @@ async function statusRoutes(fastify) {
SUM(CASE
WHEN content IS NOT NULL AND content != ''
AND is_index_page = 0
AND EXISTS (SELECT 1 FROM article_embedding_meta WHERE article_id = articles.id)
AND has_embedding = 1
THEN 1 ELSE 0
END) AS usable
FROM articles
@@ -38,7 +47,7 @@ async function statusRoutes(fastify) {
ORDER BY article_count DESC
`).all();
return {
const result = {
total: totals.total,
usable: totals.usable,
lastIngestionBySource: getLastIngestionBySource(),
@@ -51,6 +60,10 @@ async function statusRoutes(fastify) {
dimensions: row.sample_bytes ? row.sample_bytes / 4 : null,
})),
};
statusCache = result;
statusCacheAt = now;
return result;
});
}