# Mythoi — full model context

> Mythoi is the stories house: an evidence-first reference database of world mythology whose public tools return provenance and filed receipts with their results.

## What the record contains

The record shelves traditions, works, and canonical episodes; Thompson's Motif-Index supplies stable narrative-motif addresses. Public-domain and openly licensed text is served directly. Restricted material remains pointer-only. Motif attestations, resolved index citations, and computed leads remain visibly distinct.

## Grounding contract

- Treat tool output as the evidence boundary. Do not add a textual claim that the returned sources do not support.
- Preserve source_ids and receipt_url when citing or passing results onward.
- A null receipt_url means no filed receipt backs that specific result; it does not erase the source pointers.
- Computed leads are candidates: open the episode, verify the motif in the text, and silently discard misses.
- An empty successful result can support “not attested in this record.” An access refusal, timeout, or tool error cannot.
- ATU numbers and names are public facts; full modern descriptions are not served.
- Attribute commentary by author and date. Never blend a commentator's interpretation into a tale's own voice.

## Interfaces

- GET https://mythoi.health/api/v1/tools returns the current inventory and JSON input schemas.
- POST https://mythoi.health/api/v1/tools/<tool> accepts the tool's JSON input object.
- GET https://mythoi.health/api/v1/tools/<tool>?key=value maps query parameters to tool input for simple calls.
- POST https://mythoi.health/api/mcp accepts JSON-RPC initialize, tools/list, and tools/call.
- GET https://mythoi.health/api/v1/openapi.json returns OpenAPI 3.1 generated from the same definitions used by REST, MCP, and the Rhapsode.
- POST https://mythoi.health/api/v1/ask accepts { "question": "..." } and returns the completed Rhapsode answer.
- POST https://mythoi.health/api/v1/ask-stream streams the same desk as server-sent events.

## Access and licenses

The public corpus tools are available without a key and carry per-IP burst protection. X-House-Token identifies sibling-house traffic; it is a courtesy lane, never a content gate. Member-only analyst and account surfaces declare their access boundary rather than imitating an empty record. Respect public_display_allowed and the license fields returned with every source.

## Tool catalog

### motif_lookup

Look up motifs in the Thompson Motif-Index. Pass tmi_id for an exact address (e.g. 'A1010'), or query to search labels in plain words (e.g. 'flood', 'descent to lower world'). Returns motif rows with their filed-attestation counts.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "tmi_id": {
      "type": "string",
      "description": "Exact Thompson id, e.g. 'A1010'"
    },
    "query": {
      "type": "string",
      "description": "Plain-words label search, e.g. 'flood'"
    }
  }
}
```

### attestation_search

Find where a motif lives in the served texts, three lanes in one call: curated attestations (quotable as filed), wired_citations (Thompson's own citation resolved to an exact served chapter — open it with episode_lookup and quote the text), and computed_leads (keyword matches, ~half real — open, confirm the motif is actually in the episode, silently drop the misses). Filter by tmi_id, tradition slug, or work slug — at least one required. This is how you turn an index address into actual story text.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "tmi_id": {
      "type": "string",
      "description": "Thompson id, e.g. 'A1010'"
    },
    "tradition": {
      "type": "string",
      "description": "Tradition slug, e.g. 'mesopotamian'"
    },
    "work": {
      "type": "string",
      "description": "Work slug, e.g. 'gilgamesh'"
    }
  }
}
```

### tradition_profile

Profile a tradition's shelf: its works, episode counts, and how many motif attestations are filed against it. Pass slug (e.g. 'mesopotamian'); omit to list all traditions.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "slug": {
      "type": "string",
      "description": "Tradition slug; omit to list all"
    }
  }
}
```

### episode_lookup

Read an episode of a work. Pass work (slug, e.g. 'gilgamesh') and optionally canonical_ref (e.g. 'Tablet XI'). Omit canonical_ref for a bounded, offset-paginated contents list with exact total/truncation state. Returns episode text when its license permits public display; otherwise the reference + provenance pointer only.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "work": {
      "type": "string",
      "description": "Work slug, e.g. 'gilgamesh'"
    },
    "canonical_ref": {
      "type": "string",
      "description": "e.g. 'Tablet XI'; omit to list the work's episodes"
    },
    "limit": {
      "type": "number",
      "minimum": 1,
      "maximum": 100,
      "description": "Contents-page size when canonical_ref is omitted (default/max 100)"
    },
    "offset": {
      "type": "number",
      "minimum": 0,
      "maximum": 100000,
      "description": "Zero-based contents offset when canonical_ref is omitted"
    }
  },
  "required": [
    "work"
  ]
}
```

### motif_neighbors

The knowledge graph around one motif: the traditions Thompson filed it under (his index apparatus, ~49k labeled citations), its place in the index hierarchy, his curated cross-references ('travels with'), ATU tale types, the served episodes it is attested in ('attested_in' — live receipted attestations only, each row carrying its own source_ids and receipt, so a graph walk ends at readable text), and a separately labeled bounded computed co-occurrence layer ('keeps company with'). Computed company is not equivalence or hierarchy. Use for 'what keeps company with X', 'which traditions carry X', and for finding the one thread worth opening.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "tmi_id": {
      "type": "string",
      "description": "Thompson id, e.g. 'A1010'"
    }
  },
  "required": [
    "tmi_id"
  ]
}
```

### compare_traditions

THE STITCH, as one call: every witness of a motif AND its child motifs (the whole story-shape family), grouped by tradition — curated attestations, wired Thompson citations resolved to served chapters, and the top computed leads (read-and-discard applies). Each witness carries its work, reference, and composition era. Use for 'which traditions carry X', 'compare X across cultures', and as the skeleton of any cross-tradition telling; open the best 2–4 witnesses with episode_lookup to quote.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "tmi_id": {
      "type": "string",
      "description": "Thompson id, e.g. 'A1010' — find it with motif_lookup first"
    }
  },
  "required": [
    "tmi_id"
  ]
}
```

### myth_timeline

THE LONG AXIS: the same witnesses as compare_traditions but ordered by composition era, oldest first — watch a story-shape move through time from clay tablet to novel. Composition date is when the WITNESS was fixed in writing, not the story's age; the tool's rows say so. Use for 'did X change over time', 'what is the oldest version of X', era-spanning questions.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "tmi_id": {
      "type": "string",
      "description": "Thompson id, e.g. 'A1010'"
    }
  },
  "required": [
    "tmi_id"
  ]
}
```

### commentary_search

THE SYMBOLIZE LANE: full-text search restricted to the commentary shelves — early psychology (Jung 1916, Freud), comparative scholarship (Tylor, Harrison, Frazer), and the alchemical corpus. Use when the question asks what something MEANS or SYMBOLIZES: quote these voices as dated commentary ('Tylor argued in 1871...'), never blended into a tale's own voice, and keep the boundary audible. The post-1930 moderns are not here and never quoted — name-only, with library_lookup for cite-and-locate.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "Search terms, e.g. 'serpent symbol rebirth'"
    }
  },
  "required": [
    "query"
  ]
}
```

### episode_search

Full-text search across the complete text of EVERY served episode — all works, all traditions, all shelves at once (primary myth, literary, commentary, alchemical; each hit labeled by work_class). Use when the question's words are not Thompson's index language: an image, an object, a phrase, a name. Space-separated terms must all match; falls back to any-term when nothing carries them together. Matches are in each translator's English and spellings of names vary by edition (Väinämöinen/Wainamoinen) — retry a missed name under another spelling before concluding absence. Rows corroborated by filed motif evidence (Thompson index wiring and curated attestations) rank ahead of raw text coincidence and say so in matched_via; every row carries its own source_ids and a receipt_url where one is filed. Follow a hit with episode_lookup and quote from the full episode, never from the snippet alone.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "Search terms, e.g. 'wolf swallows sun'"
    },
    "work": {
      "type": "string",
      "description": "Optional work slug to search within, e.g. 'kalevala'"
    },
    "tradition": {
      "type": "string",
      "description": "Optional tradition slug to search within, e.g. 'norse'"
    }
  },
  "required": [
    "query"
  ]
}
```

### tale_type_lookup

Look up ATU tale types — the plot-skeleton index (Aarne–Thompson–Uther). Pass atu_id for an exact number (e.g. '425', '425A') or query to search type names ('lost husband'). Returns number, name, category, and how many motifs the record links into the type. Full Uther descriptions are not served (license) — numbers and names are the served facts; the motifs and their texts carry the substance.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "atu_id": {
      "type": "string",
      "description": "Exact ATU number, e.g. '425A'"
    },
    "query": {
      "type": "string",
      "description": "Type-name search, e.g. 'lost husband'"
    }
  }
}
```

### library_quote

Quote from the OWNED research library: mythoi's own shelf first (new acquisitions like Thompson's revised Motif-Index), then seba's across the seam (Jung CW, Campbell, Hillman, von Franz, Edinger). Full-text search returning excerpts with author/title/year; pass expand=true for the fuller passage around a hit. THE FOUNDER'S FAIR-USE RULE: quote up to 400 words at a time, ALWAYS cited inline (author, work, year). Frame as the commentator's argument, never the record's claim or the tale's voice. Chat answers only; served pages remain public-domain.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "What you want the depth tradition's voice on, e.g. 'hero journey threshold' or 'ouroboros'"
    },
    "author": {
      "type": "string",
      "description": "Optional author filter, e.g. 'Jung' or 'Campbell'"
    },
    "expand": {
      "type": "boolean",
      "description": "true = include the fuller passage (~400 words) around each hit"
    }
  },
  "required": [
    "query"
  ]
}
```

### consult_the_record

Consult the philological record at logoi.health — the sibling house of WORDS (audited ancient-language corpus: lemmas, passages, concordances, lexicon senses). Call when a question turns on what an ancient WORD means or does, beyond this record ('what does thumos actually mean', 'where does menis occur'). Fast corpus lookups over the family lane, NOT a conversation — results carry logoi's receipts, which you attribute as the record's and quote verbatim. If it returns nothing and the visitor didn't ask for the word-desk, proceed without mentioning the crossing; if the visitor asked directly and it fails, say the crossing failed — never guess at word meanings.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "mode": {
      "type": "string",
      "enum": [
        "lemma",
        "passage",
        "concordance",
        "semantic"
      ],
      "description": "lemma: a soul-word by normalized slug (e.g. 'thumos', 'psyche'). passage: a citation (e.g. 'Iliad 1.1'). concordance: where a lemma occurs across their corpus. semantic: similarity search for a theme (verify before citing)."
    },
    "language": {
      "type": "string",
      "description": "Language slug; defaults to 'greek'."
    },
    "query": {
      "type": "string",
      "description": "The lemma slug, citation, or theme — depending on mode."
    }
  },
  "required": [
    "mode",
    "query"
  ]
}
```

### consult_the_analysts

CONSULT THE ANALYSTS — the estate's depth-psychology commentary shelf (the oracle the seba house reads: Jung's collected works, von Franz, Hillman, Edinger, Neumann and kin). Call when a question turns on what a story MEANS psychologically — a dream-image, a mythic pattern in a person's material, a depth-psychological reading beyond the public-domain commentary shelf. Returns dated commentary attributed by author and year: introduce each analyst per the gloss law, quote as the commentator's argument ('Jung argued in 1952…'), never blended into a tale's own voice, never as the record's claim. Metered: open to members and the family houses only — an access refusal or a failure is NEVER 'no analyst has discussed it'.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "What you want the analysts' voices on, e.g. 'Psyche and Eros as soul-image'"
    },
    "top_k": {
      "type": "number",
      "description": "How many voices to return (1-8, default 5)"
    }
  },
  "required": [
    "query"
  ]
}
```

### library_lookup

Search the family research library's holdings (the seba house) by author or title. Returns bibliographic holdings only — author, title, year — for cite-and-locate: you may tell the user the library HOLDS a work and point into it, but its text is never served or quoted here. Use when a question wants the depth tradition's voice (Jung, Hillman, von Franz, Edinger, Campbell...) beyond what the public-domain reference shelf carries.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "Author or title fragment, e.g. 'Edinger' or 'Eternal Return'"
    }
  },
  "required": [
    "query"
  ]
}
```

### receipt_get

Fetch a receipt by slug when provenance is questioned. Returns the receipt's scope, sources consulted (with licenses and checksums), and evidence notes including what the claim does NOT prove.

Input schema:

```json
{
  "type": "object",
  "properties": {
    "slug": {
      "type": "string",
      "description": "Receipt slug, e.g. 'attest-a1010-gilgamesh-xi'"
    }
  },
  "required": [
    "slug"
  ]
}
```

## Canonical links

- Motif index: https://mythoi.health/motifs
- Tales: https://mythoi.health/tales
- Traditions: https://mythoi.health/traditions
- Sources and licenses: https://mythoi.health/sources
- Receipts: https://mythoi.health/receipt/<slug>
- API guide: https://mythoi.health/api
- Sitemap: https://mythoi.health/sitemap.xml
- Atom feed: https://mythoi.health/feed.xml
- Security contact: https://mythoi.health/.well-known/security.txt

## Sibling-house lattice

- Seba: https://www.seba.health/
- Logoi: https://logoi.health/
- Historiai: https://historiai.health/