Discover RSS, Atom, JSON Feed, podcast, and feed-like sources by keyword using a provider-based LogicSRC plugin.

This repository includes the plugin as a normal LogicSRC workspace plugin:

npm --workspace @logicsrc/plugin-feed-discovery run build

Consumers can import the runtime API:

import { discoverFeeds, renderRss } from "@logicsrc/plugin-feed-discovery";

const result = await discoverFeeds({ q: "microsaas", limit: 25 });
console.log(renderRss(result));

logicsrc feeds discover "microsaas" --format json
logicsrc feeds discover "microsaas" --format opml
logicsrc feeds discover "microsaas" --format rss
logicsrc feeds discover "ai agents" --type podcast
logicsrc feeds validate https://example.com/feed.xml
logicsrc feeds probe https://example.com
logicsrc feeds providers

Validation is enabled by default. Use --include-unvalidated to inspect raw provider candidates without network validation.

MVP providers:

• manual-curated: local high-trust starter feeds.
• opml-directory: local OPML files configured with LOGICSRC_FEEDS_OPML_PATHS.
• web-feed-probe: probes direct URL queries and configured candidate homepages from LOGICSRC_FEEDS_CANDIDATE_URLS.
• itunes-podcast: public iTunes podcast search.
• podcastindex: optional PodcastIndex search when PODCASTINDEX_API_KEY and PODCASTINDEX_API_SECRET are set.

Provider failures are isolated and returned as providerErrors; one failed provider does not fail the whole discovery request.

The CommandBoard reference API exposes:

GET /api/feeds/discover?q=microsaas&type=all&limit=50
GET /api/feeds/providers
GET /rss/discover/microsaas.xml
GET /api/rss/discover?q=microsaas

BitTorrented can consume the plugin package directly and map its public routes to the same runtime calls:

• /api/rss/discover?q=:keyword -> discoverFeeds()
• /rss/discover/:keyword.xml -> discoverFeeds() plus renderRss()
• /rss/discover/:keyword.opml -> discoverFeeds() plus renderOpml()

LOGICSRC_FEEDS_CACHE_TTL_SECONDS=86400
LOGICSRC_FEEDS_MAX_PROVIDERS=10
LOGICSRC_FEEDS_MAX_PROBES=50
LOGICSRC_FEEDS_REQUEST_TIMEOUT_MS=8000
LOGICSRC_FEEDS_MAX_BODY_BYTES=1000000
LOGICSRC_FEEDS_USER_AGENT="LogicSrcFeedDiscovery/0.1"
LOGICSRC_FEEDS_OPML_PATHS="./data/feeds.opml,./data/podcasts.opml"
LOGICSRC_FEEDS_CANDIDATE_URLS="https://example.com|microsaas,https://another.example|ai agents"

PODCASTINDEX_API_KEY=
PODCASTINDEX_API_SECRET=

validateFeed() and probeSite() use guarded fetches:

• Only http and https URLs are allowed.
• localhost, loopback, private, link-local, carrier-grade NAT, and common metadata targets are blocked.
• DNS-resolved private/internal addresses are blocked.
• Redirects are limited and checked through the same guard.
• Request timeouts and response body size limits are enforced.

Supabase-compatible schema SQL is included at src/db/schema.sql. The current LogicSRC repo has no shared Supabase migration convention, so the plugin ships the schema for host applications to apply.

• YouTube, Reddit, GitHub, and RSSHub adapters.
• Persistent cache and refresh jobs.
• Atom and JSON Feed public endpoints in host apps.
• Provider health dashboard and admin tooling.
• Paid provider adapters such as RSS.app, Feedly, Inoreader, Twingly, and Listen Notes.