Logoi

Lemma lookup

Look up a lemma (dictionary headword) in the audited corpus: term identity, LSJ entry pointers, and how often it is attested. Call this FIRST whenever the user asks about a word. `lemma` accepts a lowercase transliterated slug (e.g. 'psyche', 'thumos', 'menis') or native script (ψυχή) — the slug is the fast path; script input resolves through the term index.

Passage lookup

Resolve a citation to an audited passage with provenance, including source text only when the passage, source record, license, and MCP output channel all authorize it (text_status reports served / not_loaded / withheld:license / withheld:source_authority). Call when the user names a specific passage or you need the exact wording of one. Accepts stable ids ('iliad-1-1-5', 'homer:odyssey:11.205', 'plato:phaedo:64:s134'), Homeric citations ('Iliad 1.1-5', 'Od. 11.205'), shelved Greek works with the

Concordance search

Paginated concordance: every indexed occurrence of a lemma with surface form and passage pointer. Call when the user wants where/how often a word occurs, usage patterns, or specific attestations. Same lemma rule as lemma_lookup. Pass a canonical `work_id` to scope counts/results to exactly one work (totals are corpus-wide otherwise).

Lexicon lookup

Lexicon entries for a lemma. Call when the user asks about meanings, senses, or dictionary definitions. The authority parameter is advisory — the resolved term's full entry shelf is served, and the payload's authorities_served names what actually answered (each entry also carries its source_id).

Source get

Fetch a source record (edition, license, snapshot checksum, audit state) by source_id. Call when the user questions where the text comes from, or to expand a source_id returned by another tool.

Receipt get

Fetch an audit receipt and its evidence chain by id or slug. Call when the user asks how a claim is verified.

Semantic query

Semantic search over the passage index. Call for meaning/theme/paraphrase questions — 'passages about X' — when no exact lemma is given; returns the semantically nearest passages with scores. Results are similarity candidates, not audited claims: verify by reading them before citing. For a known word or citation, prefer the exact tools. The index spans the estate; unfiltered results skew Greek (it is by far the largest wing) — pass `language` to search one wing (e.g. language='pali' for Buddhist

Reverse dictionary (English doorway)

REVERSE dictionary — the English doorway: given an English sense-word ('fate', 'soul', 'anger'), return the lemmas whose lexicon glosses carry it, ranked by how central the sense is to the entry. Use whenever the user starts from an English word and wants the ancient words behind it. Results are gloss matches, NOT equivalences — confirm with lemma_lookup/lexicon_lookup before asserting identity.

Etymology lookup

Etymology authority pointers for a lemma: which reference dictionaries (Beekes EDG, Chantraine DELG, Frisk GEW) treat the word, with locator and any extracted reconstructed-root tokens. Call for origin/root/cognate questions — 'where does X come from', 'what is the root of X'. Results are POINTERS to scholarly authorities, not full articles: report which authorities treat the word and any root candidates, and never invent etymological content beyond them. Same slug rule as lemma_lookup.

Parse form

Inspect an attested surface form in one explicit language wing. Pass an exact canonical work_id to constrain returned evidence to one work; bare titles, slugs, and substring matching are rejected. A separate corpus-wide exact-surface index preflight enforces the live-query ceiling before any morphology join, so an exceptionally common surface can fail with a typed budget state even when work_id is present. Greek returns morphology features only for rows whose provider, audit, review, and public-

Cooccurrence search

Find passages where two lemmas co-occur (paired mode: lemma_a + lemma_b), or rank one lemma's companions by distinct shared passages with descriptive Dice/Jaccard overlap (collocation mode: lemma_a only). Use for questions such as 'where do thumos and psyche appear together' or 'which terms share passages with menis in the Iliad'. These whole-passage measures do not establish token proximity or order, significance, causation, lexical equivalence, or diachronic change. Both lemmas are lowercase t

Frequency profile

Occurrence count of a lemma per work, normalized to per-10,000 attested tokens. Use for quantitative distribution questions. Pass one exact canonical work_id for a one-work profile, an ordered work_ids set of 2–12 exact same-wing works for a complete research set with real zero-attestation rows, or neither for corpus-wide density ranking. work_id and work_ids are mutually exclusive. Works with no token total loaded show a null per_10k field.

Passage analysis

Interlinear analysis of a passage: returns every token in order with surface form, lemma, part of speech, morphological parse features, and an LSJ entry pointer where available. Use for detailed reading and grammatical analysis of a specific passage. Accepts the same citation forms as passage_lookup: stable ids ('iliad-1-1-5', 'homer:odyssey:11.205', 'plato:phaedo:64:s134'), Homeric book.line citations ('Iliad 1.1-5'), shelved Greek works with canonical locators ('Plato, Phaedo 64'), and biblica

KG neighbors

Knowledge-graph neighborhood: the terms that keep company with a lemma in the graph, ranked. Call for 'what clusters with thumos', 'what words are near psyche', 'the companions of menos'. Returns one-hop neighbors with the relation (CO_OCCURS_WITH for same-passage companions, BELONGS_TO_LANE for concept lanes), a weight (co-occurrences per 10k tokens), and a receipt on every edge. This reads the precomputed graph — cheaper and more stable than recomputing cooccurrence_search live. `lemma` is a l

KG path

Bounded shortest path between two lemmas in the knowledge graph: why are these two words connected? Call for 'how are thumos and logos related', 'trace psyche to daimon'. Returns the first path found (hard cap 4 hops) with the relation and a receipt on every hop, or an honest found:false. Both lemmas are lowercase transliterated slugs and must both be seeded graph nodes.

Lane compare

Compare the member soul-words of one concept lane side by side — the cross-tradition record of the soul as a single payload. Call for 'what words share the breath-soul concept', 'compare the heart-mind lane', 'the life-principle words'. Returns each member term with its relation to the lane (core / partial / contested — never a smoothed equivalence), an editorial note, and a receipt. `lane` is a lane slug (e.g. 'breath-soul', 'heart-mind', 'life-principle'); a miss returns the available lane slu

Commentary search (members + family houses)

Search the estate's secondary-literature discovery shelf for unresolved bibliography candidates about a term or theme. Results preserve only allowlisted bounded scalar author/title/publication/year/locator/URL metadata; retrieved text is withheld, and candidates are NOT approved sources, corpus attestations, or scholarly evidence. Never cite a candidate as authority until catalog identity, rights, locator, and receipt resolution are complete. Distinguish a configured successful empty from an unc

Consult the Rhapsode (mythoi.health)

Consult the story-record at mythoi.health — the estate's mythological sibling (Thompson Motif-Index addresses, episodes with canonical references, cross-tradition witnesses) — over the family lane. Reach for it when a word-question needs the TELLINGS as witnesses: where a soul-word's story-pattern actually appears, which traditions carry a motif, the oldest witness of an image. Modes: motif (find a Thompson address; query or tmi_id), episodes (full-text search the tellings; query), compare (witn