Glossary

The terms this wiki uses, precisely.

Alias — any inbound spelling that resolves to a language. Stored normalised; globally unique. Aliases and Normalisation

Base language — a language whose canonical code carries no qualifier: en, de, ceb. Its own code is what providers receive; it needs no mapping. Note this means no qualifier, not two letters.

BCP-47 — the IETF standard for language tags: language-Script-REGION, e.g. zh-Hans-CN. The casing convention this service canonicalises to.

Canonical code — the one spelling of a language this service uses and returns. Store this.

Console — animated-spoon, the legacy Django application this catalog was seeded from. Its cabinet_language, cabinet_languageregioncode, translation_provider and translation_provider_supported_region_codes tables are the source data.

Derived — an answer produced by the fallback ladder rather than an explicit mapping. Correct and usable, but not curated. Anything with match other than exact.

Effective code — what a provider would actually receive for a language, whether explicit or derived. ?effective=true.

Explicit mapping — a configured (provider, language) row. Yields match: exact. Also called pinned.

Fallback policy — per-provider: what to answer for a regional language with no explicit row. base_language, passthrough or strict. Fallback Policies

Homoglyph — a character that renders identically to another, like Cyrillic с (U+0441) and ASCII c. The console contains one in Cebuano's code; normalisation repairs it. Aliases and Normalisation

Lookup key — the normalised form of a code: lowercase, - separated. What aliases are stored and matched as.

match — how an answer was reached: exact, variant, canonical, base_language, passthrough. Resolution Ladder

Provider — a translation engine, identified by the same code provider_credentials uses: deepl_api, openai_api.

Provider alias — another name for a provider: a short name (deepl) or a legacy console service_name (OpenAiGPT4o_2024_11_20_Provider).

Regional language — a language whose canonical code carries a script, a region or both: pt-BR, zh-Hans, sr-Latn-RS. The ones providers disagree about.

Seed — data/seed/*.json, generated from a console dump by scripts/build_seed.py. Committed, reviewable, deterministic. Rebuilding the Seed

service_name — the console's identifier for a provider row. Registered here as a provider alias so console callers keep working.

Vendor — the company behind one or more providers. Google Cloud is the vendor of both google_translate and gemini_api.