Shared LLM provider and model configuration registry.
Used by Entelecheia and Shittim Chest.
简体中文 · 繁體中文 · 日本語 · 한국어 · Français · Español · Русский · العربية
provider-registry/
├── entrypoint/ # Per-provider API configurations (22 providers)
│ ├── anthropic/
│ │ └── default.toml
│ ├── openai/
│ │ ├── default.toml
│ │ └── generation.toml
│ └── ...
├── models/ # Per-model metadata (160+ models)
│ ├── anthropic/
│ │ ├── claude-opus-4.toml
│ │ └── ...
│ └── ...
├── registry.toml # Provider catalog (grouped by category)
├── scripts/
│ ├── update_models.py # Sync model lists from external APIs
│ └── utils/
│ └── cli_format.py
├── justfile # Task runner for common operations
└── .github/workflows/ # CI: daily model list sync
Each provider has one or more entrypoint TOML files:
[entrypoint]
id = "anthropic_default"
provider_id = "anthropic"
[entrypoint.api]
protocol = "anthropic_messages_v1"
base_url = "https://api.anthropic.com/v1"
env_var = "ANTHROPIC_API_KEY"
[entrypoint.name]
en = "Anthropic (Official)"
zhs = "Anthropic(官方)"
zht = "Anthropic(官方)"
ja = "Anthropic(公式)"
ko = "Anthropic(공식)"
fr = "Anthropic (Officiel)"
es = "Anthropic (Oficial)"
ru = "Anthropic (Официальный)"Each model has an individual TOML file:
[model]
id = "claude-opus-4"
name = "Claude Opus 4"
context_window = 200000
max_output_tokens = 16384
supports_vision = true
supports_streaming = true
[model.pricing]
input_per_million = 15.0
output_per_million = 75.0# Install just
cargo install just
# Sync from all sources (recommended)
just sync-all
# Sync from specific source
just sync-openrouter
just sync-modelsdev
# Sync specific providers
just sync all openai,anthropic# In your project
git submodule add https://github.com/celestia-island/provider-registry.git provider-registryA daily GitHub Actions workflow fetches the latest model lists from OpenRouter and models.dev and commits changes to the dev branch.
Manual triggers are available via the Actions tab with configurable source and provider filters.
CC0 1.0 Universal — public domain, no restrictions.
