searxng-kg-ingestion

SearXNG KG Ingestion

Native "maximum ingestion" of SearXNG results into the ONE epistemic-graph knowledge graph. Runs a metasearch and pushes each result as a shared :Document (carrying the result text + source_uri) alongside a :SearchQuery node and per-engine :SearchEngine nodes, wired with :resultOf / :fromEngine. This builds a recallable, citable corpus from the open web.

When to use

• Build durable KG memory of what the web says about a topic (for later recall/citation).
• Seed a research corpus that hub-side enrichment can chunk and embed for semantic search.
• Capture a monitoring snapshot you want to query again later, not just read once.

When NOT to use

• Throwaway lookups you don't need to persist → searxng-web-search.
• Ingesting the full body of one known page → web-crawler / web-fetch.
• Structured records from an enterprise system → that system's own ingest tool.

Prerequisites & environment

Connect via the mcp-client skill against the searxng-mcp MCP server. A reachable epistemic-graph engine is required for anything to land; with no engine the tool is a clean no-op (ingested: null). Same env matrix as searxng-web-search. Note: web_search never ingests; its results stay a virtual source. Use the explicit tool below when you want the ingest confirmation payload or ingest without a prior read.

Tools & actions

Tool	Purpose
searxng_ingest_search	Run a search AND push results to the KG; returns {listed, ingested}.
web_search	Plain search; writes nothing to the graph.

Key parameters (searxng_ingest_search)

• query — the search string (required); also becomes the :SearchQuery node.
• categories / engines / language / pageno — same semantics as web_search.

Recipes

Search + ingest a topic:

{"query": "zero trust network access self-hosted", "categories": ["it"]}

Ingest news coverage for a monitoring corpus:

{"query": "cilium 1.16 release notes", "categories": ["news"]}