yagonode

Package adminui serves the yago operator console as a server-rendered surface styled with IBM Carbon design tokens and progressively enhanced with htmx.

Package boltvault is the bbolt implementation of the vault Engine.

Package bootstrap fetches the configured YaCy seed lists and returns the seeds they advertise.

Package cachedpage serves read-only cached copies of locally stored pages on the public listener: the text this node extracted at crawl time, rendered escaped, so a searcher can inspect a result without visiting the origin site.

Package contentquality gates crawled text before it is stored and indexed, using the deterministic token-based subset of the Gopher (Rae et al., arXiv:2112.11446, Appendix A) and C4 (Raffel et al., arXiv:1910.10683) quality rules that FineWeb (arXiv:2406.17557) validated at web scale.

Package crawlbroker is the node's gRPC edge to the crawl fleet.

Package crawldispatch turns an operator's request to crawl seed URLs into a CrawlOrder and hands it to the crawl fleet.

Package crawlformats persists the operator's shared document-format toggles: which format families every crawl parses (YaCy TextParser parity).

Package crawling owns the crawlReceipt endpoint, which rejects remote crawl receipts until an operator policy enables remote crawl work.

Package crawlresults absorbs ingest batches returned by the crawl fleet, storing their URL metadata and then their postings through the node's existing receivers.

Package crawlruns is the node's in-memory registry of crawl runs.

Package crawlschedule stores recurring crawl definitions and decides when each is due — YaCy's Automation_p (recorded API calls with a recurring schedule) parity for the one action worth repeating: re-crawling a site (UI-19).

Package crossorigin applies a configurable, default-deny CORS policy to an HTTP handler.

Package dhtexchange builds outbound YaCy DHT exchange batches.

Package documentsearch finds documents containing wanted terms, orders them by relevance, and reports how many documents matched each term.

Package events records recent structured node events in a bounded in-memory ring so operators and the admin UI can review important activity without scraping logs.

Package eventstore is a durable, bounded log of structured node events so the admin event log survives restarts.

Package eviction frees storage when the vault nears its quota.

Package faviconproxy serves result favicons through this node, so the hosts behind search results never see the searcher's browser before a click (the privacy property YaCy gets from its ViewImage proxy).

Package filetypeclass classifies a crawled document's file type from its Content-Type and URL, so the filetype: search operator and the filetype navigation facet recognise a document by what it actually is rather than only by a file extension in its URL.

Package httpguard enforces the HTTP request invariants every module endpoint shares: method set, body limit, and request timeout.

Package indextransfer owns outbound YaCy index transfer HTTP calls.

Package judgments persists operator-curated relevance judgments — a query paired with graded result URLs — as the human-labelled training and regression set the ranking learner (YagoRank, ADR-0035) fits weights against.

Package landing serves the node's static landing page on GET and HEAD.

Package landiscovery finds peers on the local network the way Syncthing's local discovery does: a small UDP beacon broadcast on a fixed port announces this node's hash and peer port, and received announcements from other nodes hand a candidate endpoint to the greeter, which runs the ordinary hello exchange — the beacon itself is never trusted for anything but «someone at this address claims to be a peer», so a forged packet can at most trigger one verified greet.

Package memvault is the in-memory implementation of the vault Engine.

Package metrichistory samples a curated set of the node's Prometheus metrics into a bounded in-memory ring, so the admin Performance page can render a short operational history (request throughput, error rate, latency, DHT transfer rate, queue depths) without an external metrics stack.

Package metrics translates domain results into Prometheus series.

Package modifierhint supplies the guidance shown when a search carried a filter operator (filetype:, site:, tld:, inurl:) but no free-text term and so matched nothing.

Package nodeidentity holds this node's self-description: the hash, network, and seed attributes that identify it, and the rule for whether a peer request addresses it.

Package nodestatus owns the node's runtime status: its self-seed, the version/uptime header every endpoint echoes, and the query.html capacity answers.

Package peeradmission answers inbound hello requests: it classifies the calling peer by probing it back, and returns a bounded freshest set of reachable peers.

Package peerannouncement greets known peers on an interval: it announces this node to them and reports their reachability to the peer roster.

Package peerbirth keeps this peer's birth date: the moment the node first started with its data directory.

Package peerblock keeps a durable operator-managed blocklist of peer hashes.

Package peeridentity keeps this peer's stable identity — its hash and name — so a node bootstraps without the operator having to supply them.

Package peernews keeps the YaCy peer news pool: news records this peer publishes to the network and news records that arrive attached to other peers' seeds.

Package peerroster owns the set of network peers this node knows.

Package portaltheme stores and renders the operator-editable public portal theme (ADR-0033): Handlebars templates for the search and results pages plus a shared styles block, persisted in the durable vault and rendered server-side.

Package publicportal serves the anonymous public search portal: a minimal, server-rendered search page on the node's public HTTP listener that works without JavaScript and in legacy browsers.

Package publicratelimit throttles the anonymous public search surfaces the way YaCy's search.public.max.access tiers do: per-client sliding windows at three second, one minute, and ten minute horizons, with raised limits for authenticated callers (a valid bearer key or the local operator).

Package publicrobots serves the node's own /robots.txt — YaCy's ConfigRobotsTxt_p parity.

Package rankfit fits the ranking weight profile to measured relevance (YagoRank, ADR-0035).

Package rankingprofile persists the node's default search ranking weights and serves them live to the local searcher, so an operator can retune ranking without a restart and the setting survives one.

Package readpriority defers a bulk writer while interactive reads are in flight, so searches and admin pages reach the disk ahead of ingest writes and their fsyncs (IO-PRIO-01, made a configurable, observable gate for PERF-PRIO-02).

Package recrawlfrontier keeps a durable, node-side recrawl schedule: for each crawled URL it records when the page was last fetched and when it is next due for a recrawl, so pages are refetched on their profile's cadence and the schedule survives node and crawler restarts.

Package redirectpurge sweeps already-indexed search-engine click-tracking documents out of the live corpus (SEARCH-29).

Package remotecrawl owns opt-in YaCy remote crawl delegation.

Package rwi owns RWI posting intake, storage, search, and eviction.

Package searchactivity keeps a bounded in-memory journal of recent search requests for the admin console — YaCy's AccessTracker_p parity (UI-16).

Package searcheval scores a searcher's ranking quality with NDCG@k so a test can gate ranking changes against regression, and turns query→relevant-URL labels (including corpus-derived pseudo-labels) into judgments.

Package searchsession keeps paging stable by caching assembled result prefixes and extending them in bounded windows when a deeper page needs more results.

Package seedimport keeps a durable per-URL record of the last seed-list import so the admin console can show when each configured seed list was last refreshed and whether it succeeded.

Package settingsstore is a durable, admin-writable key/value store for the node's operator-overridable runtime settings.

Package shardvault is the sharded, compressed bbolt implementation of the vault Engine (ADR-0025).

Package snippetmark highlights query terms in result snippets for the human search surfaces.

Package spellcheck corrects query spelling against the node's own indexed vocabulary with the Symmetric Delete algorithm (SymSpell, Garbe 2012): the dictionary is precomputed once so a lookup only deletes characters from the query term, never the dictionary, making correction fast enough to run on every zero-result query.

Package stopwords carries a compact multilingual function-word list for query-time decisions: which query words may be required of documents, verified against peer results, anchor snippets, or drive highlighting.

Package tracectx implements the W3C Trace Context header (traceparent) for the node's own request legs — enough to correlate one public search with the peer fan-out it caused and to attach exemplars to latency histograms (OPS-10).

Package transfertally keeps the running and persisted totals of this peer's YaCy index exchange for the advertised seed statistics.

Package urldenylist keeps a durable, operator-managed denylist of URLs and whole domains.

Package urlmeta owns the transferURL endpoint, URL intake, and URL metadata storage and lookup.

Package urlmetastaleness orders url metadata from stalest to freshest so eviction can name the stalest urls without decompressing a single row.

Package urlreferences knows, for every url, the words that reference it, so eviction can drop a url's postings without scanning the word index.

Package vault lends each module a transactionally-scoped, codec-typed collection over its own ordered byte buckets.

Package vaulttest holds the storage-contract suite every vault Engine driver runs against itself.

Package websearch provides optional, admin-gated web retrieval after a local and federated miss or alongside those sources in always mode.

Package wordforms expands a query word into the inflected surface forms that actually occur in the node's index.