@loutre-run/core

@loutre-run/core

The engine behind Loutre: turn an OpenAPI 3.x or Swagger 2.0 document, a set of TypeScript types, a Postman, Insomnia or Bruno collection, a HAR file, a json-server db.json, curl commands or a GraphQL schema into a typed route table and a realistic, coherent data set, then answer requests against it. No network, no AI call: the same deterministic generator that runs on loutre.run.

npm install @loutre-run/core

import { compileToResult, resolve } from "@loutre-run/core"

const compiled = await compileToResult(openApiYaml, { seed: "demo" })
if (!compiled.ok) throw new Error(compiled.errors.map((e) => e.message).join("\n"))

const reply = resolve(compiled.mock, {}, { method: "GET", pathname: "/users/42", headers: {} })
console.log(reply.status, reply.body)

What you get:

• compileToResult(input, options) and analyseToResult(input): ingestion, normalisation (Swagger 2.0 converted, $ref resolved, OpenAPI 3.0 upgraded to 3.1), interpretation into a route table, generation of the data. Errors come back as diagnostics with a JSON pointer and, when possible, a suggestion.
• resolve(mock, behavior, request): the pure resolver. It honours the per-request headers x-loutre-status, x-loutre-delay, x-loutre-count, x-loutre-message and x-loutre-seed, per-route behaviour (status, latency, error rate, hand-written bodies, conditional rules) and an optional overlay of persisted collections.
• typescriptToOpenApi(source), postmanToOpenApi(collection), insomniaToOpenApi(export), brunoToOpenApi(collection) bruToOpenApi(text), harToOpenApi(har), jsonServerToOpenApi(db) and curlToOpenApi(text): the converters used at ingestion. sanitizeSpec(text) reduces a HAR to its API calls, without headers or cookies, before it is sent or kept.
• answerGraphql(mock, behavior, request) and executeGraphql(mock, request): a GraphQL mock's endpoint. The schema (SDL or introspection) is read by graphql-js, the only extra dependency, kept out of @loutre-run/core/detect; queries, mutations (written as state operations), fragments, variables and introspection are answered from the generated collections, with consistent links and Relay connections.
• detect(input) (also exported as @loutre-run/core/detect, dependency free): format detection for a browser.
• generateSync(table, { seed }): realistic data by field name and format (French identities, valid SIRET and IBAN, created_at before updated_at, totals that add up), stable for a given seed.

Input formats: OpenAPI 3.0 / 3.1 / 3.2 (YAML or JSON; 3.2 is read as 3.1), Swagger 2.0, TypeScript interface / type / enum declarations (CRUD routes derived per entity, or an explicit route map type), Postman collections v2.0 / v2.1, Insomnia exports (v4 JSON, v5 YAML; a design document is read through its OpenAPI spec), Bruno collections (JSON export, or .bru files one after the other), HAR files (shapes only, data generated), json-server db.json (records served as they are, through x-loutre-fixtures) and curl commands. Requests without a response example answer with the shape the collection sends to the same resource.

Diagnostics and generated error messages are in French, on purpose: Loutre is built for French-speaking teams first. Everything else is language neutral.

This package serves Loutre first. The public API is what the hosted service uses; breaking changes are listed in the changelog. Licence MIT.