RSS Utils for Flow-Like

Flow-Like WASM nodes for RSS and Atom feed utility workflows. The package is intended to cover the practical pieces around feed discovery, fetch request preparation, central parsing, release-date filtering, deduplication, and formatting.

Repository: https://github.com/Rheosoph/node-package-rss-utils

Author: Felix Schultz, Rheosoph GmbH

Store page copy: STORE.md

Implemented Nodes

Node	Category	Purpose
Normalize Feed URL	RSS/Utilities	Trim and normalize a feed URL, optionally adding https:// and removing fragments.
Build Feed Request	RSS/Fetch	Create a typed HTTP request descriptor with RSS/Atom/JSON Feed Accept, User-Agent, ETag, and Last-Modified headers.
Parse Feed	RSS/Parsing	Centrally parse RSS, Atom, RDF, and JSON Feed documents into normalized feed metadata and items.
Extract Feed Links	RSS/Discovery	Find RSS, Atom, and JSON Feed alternate links in an HTML document.
Search Feed Items	RSS/Items	Keep items whose text fields contain a query.
Filter Feed Items by Category	RSS/Items	Keep items with a matching category term.
Filter Feed Items by Date	RSS/Items	Keep items inside an inclusive release-date window.
Dedupe Feed Items	RSS/Items	Remove repeated items by id, link, or title plus date.
Limit Feed Items	RSS/Items	Keep the first N feed items.
Feed Item to Markdown	RSS/Formatting	Format a normalized feed item as Markdown for summaries, digests, and notifications.

Layout

Each Flow-Like node lives in its own file under src/nodes/:

src/nodes/normalize_feed_url.rs
src/nodes/build_feed_request.rs
src/nodes/parse_feed.rs
src/nodes/extract_feed_links.rs
src/nodes/search_feed_items.rs
src/nodes/filter_feed_items_by_category.rs
src/nodes/filter_feed_items_by_date.rs
src/nodes/dedupe_feed_items.rs
src/nodes/limit_feed_items.rs
src/nodes/feed_item_to_markdown.rs

Shared code is split by responsibility: parser.rs, dates.rs, filtering.rs, discovery.rs, formatting.rs, and types.rs.

When parsing a body fetched by another node, pass the original feed URL into Parse Feed as source_url. The parser uses it as the base URL for feeds that contain relative media, enclosure, or content links.

Extract Feed Links only returns absolute HTTP(S) feed URLs. Relative feed links are resolved when base_url is provided; unresolved relative links are ignored so downstream HTTP nodes do not receive invalid request URLs.

Date Handling

dates.rs holds one tolerant parser, used both for the released_from and released_to bounds and - registered as feed-rs' timestamp parser - for every date inside a feed document. Text is always tried as written first, so a well-formed date goes straight to chrono and none of the repair below can change it.

Standard forms: RFC3339, RFC2822, HTTP-date, Unix seconds/milliseconds/microseconds/nanoseconds, YYYY, YYYY-MM, YYYY-MM-DD, YYYYMMDD, ISO basic (20260604T123456Z), and named dates such as June 4, 2026.

Feeds stray from all of those, so the parser also reads:

• Timezone names beyond the short RFC 2822 list: CEST, MESZ, IST, AEST, and around sixty more, plus GMT+2 and UTC+02:00 style offsets.
• Zone database names such as Europe/Berlin or EST5EDT, resolved against the date being parsed so summer and winter offsets both come out right.
• Localized weekday and month names across roughly twenty languages, including Cyrillic, Greek, and Arabic script, and CJK field markers (2026年6月4日).
• Dates stated relative to now: 2 hours ago, yesterday, last week. These work as filter bounds too, so released_from can be 7 days ago.
• Single digit hours, the 24:00 spelling of midnight, (CEST) comments, ISO decimal commas, and dates carrying no timezone at all, which are read as UTC.

Abbreviations that several countries share are settled by the feed URL's country code: IST is +05:30 by default, +01:00 from a .ie feed, +03:00 from .il. A date whose zone name is recognized by neither the abbreviation table nor the zone database is read as UTC rather than dropped, since a shifted release date still sorts and filters while a missing one leaves the item out of every window. A date that still cannot be read is reported through the Parse Feed warnings with the offending text.

RSS has no item update element, so feeds borrow atom:updated or dcterms:modified. feed-rs ignores both; the parser reads them back, which gives date_field: "updated" something of its own to compare.

Date-only values expand to the whole day, month, or year. Filter Feed Items by Date uses Flow-Like Date pins for its bounds; programmatic callers may still pass strings or wrapped Flow-Like date objects.

Zone database size

The zone database is filtered at build time by CHRONO_TZ_TIMEZONE_FILTER in .cargo/config.toml, which keeps the major world zones and trims about 700 KB off the built component. Zone names outside that list still resolve - they fall back to being read as UTC - they just lose their exact offset. Widen the filter if a feed you care about needs one.

Validation Corpus

tests/feed_corpus.csv tracks 200 public feed targets for parser validation: 50 Reddit feeds, 25 AWS feeds, 50 news feeds, 50 technology feeds, and 25 GitHub Atom feeds. Normal tests validate the corpus structure and coverage without making live network calls.

tests/date_corpus.txt pins how date text is read: real values sampled from 300 live feeds alongside one case per parser rule, and inputs that must be refused so article titles never become release dates. Every row is also chopped, padded, and byte-swapped to check the parser cannot panic - the component is built with panic = "abort", so a panic takes down the workflow rather than one item.

Node Ideas

The most useful next nodes fall into a few groups:

Node Idea	Why it is useful
Fetch Feed	Perform the actual HTTP GET, expose status, response headers, body, ETag, and Last-Modified; requires NetworkHttp.
Validate Feed	Report XML, date, URL, enclosure, and required-field issues before downstream processing.
Sort Feed Items	Sort items by published or updated date with stable fallback behavior.
Merge Feeds	Combine multiple normalized item arrays while preserving source metadata.
Configurable Dedupe Feed Items	Selectable key strategy: GUID, canonical link, title/date, or content hash.
Extract Enclosures	Pull podcast/audio/video/image enclosures into typed media structs.
Feed Items to Digest	Produce Markdown, HTML, or plain-text digests grouped by feed, date, author, or category.
Build RSS Feed	Emit valid RSS XML from normalized feed metadata and items.
Build Atom Feed	Emit valid Atom XML from normalized feed metadata and items.
Feed Health Check	Track stale feeds, changed status codes, parse failures, item velocity, and missing conditional headers.
Feed Cache Key	Create deterministic cache keys from URL plus conditional headers for scheduled workflows.
Discover Site Feeds	Fetch a webpage and combine HTML link extraction with common paths like /feed, /rss, and /atom.xml.
OPML Import	Parse OPML subscription lists into feed source records.
OPML Export	Write feed source records back to OPML.

Build

cargo build --release --target wasm32-wasip2

The release WASM component is written to:

target/wasm32-wasip2/release/node_package_rss_utils.wasm

Using mise:

mise run build

This also copies the component to node.wasm.

Test

cargo test --target $(rustc -vV | grep host | awk '{print $2}')

The default test suite runs offline. It validates the 200-feed corpus structure and smoke-tests every node against corpus-derived fixtures. Live third-party feed fetching is intentionally ignored by default:

cargo test --target $(rustc -vV | grep host | awk '{print $2}') -- --ignored live_feed_corpus_fetches_and_parses_all_feeds

Using mise:

mise run test

Publishing

1. Build with mise run build.
2. Open Flow-Like Desktop.
3. Go to Library -> Packages -> Publish.
4. Select node.wasm and flow-like.toml.
5. Submit the package.

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE)
• MIT License (LICENSE-MIT)

at your option.

Copyright (c) 2026 Rheosoph GmbH, Felix Schultz.