Collections Feeds API

Collections Feeds is a thin API over Watchlists sources and jobs that ingests RSS/Atom into Collections items. Each subscription is stored as a Watchlists source and polled by a Watchlists job. Items are written to the Collections DB with origin="feed".

This API is ingest-only. It does not publish RSS/Atom or render UI. See Docs/Product/Collections_Feeds_Ingestion.md for the roadmap and Reference implementation below for newsletter-to-Atom details.

Endpoints

• POST /api/v1/collections/feeds - create a feed subscription
• GET /api/v1/collections/feeds - list feed subscriptions
• GET /api/v1/collections/feeds/{feed_id} - get a feed subscription
• PATCH /api/v1/collections/feeds/{feed_id} - update a feed subscription
• DELETE /api/v1/collections/feeds/{feed_id} - delete a feed subscription

Core object: CollectionsFeed

Key fields: - id: watchlists source id - name, url, source_type (rss) - origin: always feed - tags: list of tag strings - active: source/job active flag - settings: optional source settings (sanitized) - last_scraped_at, etag, last_modified, defer_until, status, consec_not_modified: polling state - created_at, updated_at: source timestamps - job_id: watchlists job id - schedule_expr, timezone, job_active, next_run_at, wf_schedule_id: job schedule and scheduler metadata

Create

POST /api/v1/collections/feeds

Request:

{
  "url": "https://example.com/feed.xml",
  "name": "Example Feed",
  "tags": ["news"],
  "active": true,
  "timezone": "UTC",
  "settings": {
    "history": {"strategy": "auto", "max_pages": 3}
  }
}

Response:

{
  "id": 42,
  "name": "Example Feed",
  "url": "https://example.com/feed.xml",
  "source_type": "rss",
  "origin": "feed",
  "tags": ["news"],
  "active": true,
  "settings": {
    "history": {"strategy": "auto", "max_pages": 3}
  },
  "job_id": 101,
  "schedule_expr": "0 * * * *",
  "timezone": "UTC",
  "job_active": true,
  "next_run_at": "2025-01-01T01:00:00+00:00",
  "wf_schedule_id": "sched_abc123"
}

Notes: - If name is omitted, it defaults to the URL hostname. - If schedule_expr is omitted, the subscription defaults to hourly polling (0 * * * *) and auto-promotes to daily after 24 hours (see Scheduling). - timezone accepts IANA names (e.g. America/New_York) or UTC+/- offsets.

List

GET /api/v1/collections/feeds

Query params: q, page (default 1), size (default 20, max 200).

Response:

{
  "items": [
    {
      "id": 42,
      "name": "Example Feed",
      "url": "https://example.com/feed.xml",
      "source_type": "rss",
      "origin": "feed",
      "tags": ["news"],
      "active": true,
      "settings": null,
      "job_id": 101,
      "schedule_expr": "0 * * * *",
      "timezone": "UTC",
      "job_active": true,
      "next_run_at": "2025-01-01T01:00:00+00:00",
      "wf_schedule_id": "sched_abc123"
    }
  ],
  "total": 1
}

Get

GET /api/v1/collections/feeds/{feed_id}

Returns a CollectionsFeed object.

Update

PATCH /api/v1/collections/feeds/{feed_id}

Request:

{
  "name": "Example Feed (Daily)",
  "tags": ["news", "daily"],
  "schedule_expr": "0 0 * * *",
  "timezone": "UTC",
  "active": true,
  "settings": {
    "history": {"strategy": "none"}
  }
}

Response: updated CollectionsFeed.

Notes: - When you set schedule_expr via PATCH, the job switches to manual scheduling and stops auto-promoting. - settings is merged into existing settings; reserved keys are overwritten (see Settings).

Delete

DELETE /api/v1/collections/feeds/{feed_id}

Response:

{ "success": true }

Scheduling behavior

Default behavior: - If schedule_expr is omitted on create, the job starts hourly (0 * * * *). - After 24 hours, the watchlists pipeline promotes the schedule to daily (0 0 * * *) and marks the job as promoted.

Manual behavior: - If you set schedule_expr on create or update, the job is treated as manual and will not auto-promote. - timezone is normalized to UTC when absent.

Settings pass-through

settings is stored in the Watchlists source settings_json and is passed into the RSS fetchers.

Reserved keys are enforced by the server and will be removed or overwritten: - collections_origin - collections_feed_job_id, collections_job_id - collections.origin, collections.job_id

For supported settings fields (history/backfill options, limits), see Docs/API-related/Watchlists_API.md and tldw_Server_API/app/core/Watchlists/fetchers.py.

Collections ingestion behavior

• RSS/Atom sources are fetched via the Watchlists pipeline (ETag/Last-Modified, RFC5005 history, dedupe by guid/url/title).
• Items are upserted into Collections with:
• origin = "feed"
• origin_type = source_type
• origin_id = source id
• Embeddings metadata includes origin for traceability.

Reference implementation: Kill the Newsletter (MIT)

The Kill the Newsletter script is a useful reference for email ingestion, Atom rendering, and UI flows. These features are not implemented in tldw_server yet, but can guide future work.

Data model (SQLite)

Tables and key columns: - feeds: id, publicId, title, icon, emailIcon - feedEntries: id, publicId, feed, createdAt, author, title, content - feedEntryEnclosures: id, publicId, type, length, name - feedEntryEnclosureLinks: feedEntry, feedEntryEnclosure - feedVisualizations: feed, createdAt (rate limiting) - feedWebSubSubscriptions: feed, createdAt, callback, secret

SMTP/email ingestion and attachments

• SMTP server with TLS, AUTH disabled, and a size limit (2 ** 19).
• Validates mailFrom and recipient addresses, blocks known relay domains.
• Parses email with mailparser; chooses HTML or text-as-HTML for content.
• Creates feedEntryEnclosures for attachments and writes files under data/files/<publicId>/<name>.
• Links enclosures to entries; feed rendering exposes them as Atom enclosures.
• Caps feed size by total title + content length (2 ** 19) and deletes oldest entries and links when exceeded.

Atom feed rendering

• Feed id: urn:kill-the-newsletter:<feed publicId>.
• self link: /feeds/<publicId>.xml, hub link: /feeds/<publicId>/websub.
• Optional <icon> from icon or emailIcon.
• updated uses most recent entry createdAt.
• Entries include link to HTML entry view, published/updated, author, title, and <content type="html">.
• Enclosures are emitted as <link rel="enclosure" type="..." length="..." href="...">.
• Footer link in entry content points back to feed settings page.

HTML entry views

• /feeds/<feedPublicId>/entries/<entryPublicId>.html returns raw HTML body.
• CSP: default-src 'self'; img-src *; style-src 'self' 'unsafe-inline'; frame-src 'none'; object-src 'none'; form-action 'self'; frame-ancestors 'none'.

UI and settings pages

• Root page creates a feed, then shows the generated email address and Atom URL.
• Feed settings page includes copy-to-clipboard for email/feed URLs, update title/icon, and delete with title confirmation.
• Uses flash notifications, simple CSS layout, and form-based flows.

WebSub push and rate limiting

• POST /feeds/<publicId>/websub validates callback URL, blocks localhost, and caps new callbacks per day.
• Background verification job calls the hub with hub.challenge.
• Dispatch job POSTs the Atom body to each callback, with Link headers and optional X-Hub-Signature HMAC.
• feedVisualizations limits XML feed requests to 10 per hour and is cleaned up hourly.

Cleanup jobs

• Removes orphaned attachment files and DB rows.
• Deletes stale feedVisualizations (> 1 hour) and feedWebSubSubscriptions (> 24 hours).