freshrss-knowledge-ingestion

FreshRSS Knowledge Ingestion

Push FreshRSS content into the ONE epistemic-graph knowledge graph so feed items become semantically searchable documents and feeds become typed, linkable nodes. Ingestion is best-effort and engine-guarded: with no reachable engine it cleanly no-ops (ingested: null) and the fetch still returns.

When to use

• Build a searchable corpus of feed items (:Document with body + source_uri).
• Materialize the feed topology (:FeedSubscription → :FeedCategory) in the KG.
• Run a delta sweep that both reads and persists what is new.

When NOT to use

• Just reading/triaging a stream without persisting → freshrss-feed-reader.
• Managing which feeds are followed / their tags → freshrss-subscription-management.

Prerequisites & environment

Connect via the mcp-client skill against the freshrss-agent MCP server. A reachable epistemic-graph engine is required for a non-null result.

Variable	Required	Notes
FRESHRSS_URL / FRESHRSS_USER / FRESHRSS_API_PASSWORD	✅	GReader access
FRESHRSS_KG_AUTO_INGEST	optional	Default on; set 0 to disable auto-ingest on every fetch

Tools & actions

Tool	Purpose
freshrss_ingest_items	Fetch a stream, push its items → :Document nodes
freshrss_ingest_subscriptions	List subscriptions, push → :FeedSubscription + :FeedCategory (+ :inCategory)

Node ids follow freshrss:<class>:<externalId> (freshrss:item:<id>, freshrss:subscription:<id>, freshrss:category:<id>).

Recipes (params_json)

Ingest the newest 200 items across everything:

{"stream_id":"user/-/state/com.google/reading-list","count":200,"order":"n"}

Delta ingest since a watermark:

{"stream_id":"user/-/state/com.google/reading-list","newer_than":1719792000,"count":500}

Ingest the whole subscription topology (no params):

{}

Gotchas

• Ingestion is best-effort: a null ingested means no engine was reachable, not a failure — the listing still succeeds.
• freshrss_ingest_items accepts the same params as freshrss_reader stream_contents; page with continuation for a full sweep.
• Feed items land as :Document (type-forced) carrying text, title, source_uri, published, and the originating feed_title/feed_stream_id.
• Fetching via freshrss_reader already auto-ingests items by default; call freshrss_ingest_items explicitly when you want an ingest without the read path, or freshrss_ingest_subscriptions for the feed topology.
• type on every node matches a class federated by the package's feed.ttl ontology leg (:Document, :FeedSubscription, :FeedCategory).