メニュー
LiteLLM AI gateway proxy for unified LLM API access. Use when deploying via Docker/Helm/Kubernetes, config.yaml, virtual API keys, OIDC/SSO, RBAC, Langfuse, Open WebUI, vLLM, or fallback routing.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Reviews code comments across languages before a PR opens. Use when finishing a development branch, before merging, or when a pr create was blocked pending comment review.
Create, promote, categorize, and validate skills for the kettleofskills marketplace. Use when syncing dotfiles skills into the marketplace, adding skills, assigning categories, or preparing a release.
Build Go terminal UIs (TUIs) and CLIs with charmbracelet — Bubble Tea runtime, Lip Gloss styling, Bubbles components, Glamour markdown, Harmonica animation, BubbleZone mouse. For any bubbletea TUI.
Copier templating — author copier.yml templates (questions, Jinja, _tasks, migrations), scaffold via `copier copy`, sync via `copier update`. Use for copier.yml, .copier-answers.yml, or .jinja.
Copyparty portable file server (HTTP/WebDAV/FTP/SFTP/SMB). This skill should be used when deploying via Docker/pip/NixOS, configuring accounts/volumes, indexer, hooks, or forward-auth IdP integration.
Sway tiling Wayland compositor on NixOS. This skill should be used when configuring programs.sway, GPU rendering (Intel/AMD/NVIDIA), Wails/webkit2gtk fixes, xdg portals, home-manager, or wallpapers.
| name | litellm |
| last_updated | "2026-04-05T00:00:00.000Z" |
| version | v1.82.3-stable (latest stable), v1.83.1-nightly (latest nightly) |
| description | LiteLLM AI gateway proxy for unified LLM API access. Use when deploying via Docker/Helm/Kubernetes, config.yaml, virtual API keys, OIDC/SSO, RBAC, Langfuse, Open WebUI, vLLM, or fallback routing. |
OpenAI-compatible API gateway routing to 100+ LLM providers with unified auth, spend tracking, rate limiting, and observability.
# Docker (stable)
docker run -v $(pwd)/config.yaml:/app/config.yaml \
-e LITELLM_MASTER_KEY=sk-1234 \
-e DATABASE_URL=postgresql://user:pass@host:5432/litellm \
-p 4000:4000 \
docker.litellm.ai/berriai/litellm:main-stable \
--config /app/config.yaml
# Helm
helm install litellm oci://docker.litellm.ai/berriai/litellm-helm
| Tag | Use |
|---|---|
main-stable | Production (recommended) |
main-latest | Bleeding edge nightly |
main-v1.82.3 | Pinned version |
litellm-non_root:main-stable | Non-root container |
Registries: docker.litellm.ai/berriai/litellm, ghcr.io/berriai/litellm-database
Five top-level sections:
model_list: # Model deployments (provider, api_base, api_key)
router_settings: # Load balancing, retries, Redis, routing strategy
litellm_settings: # Callbacks, caching, fallbacks, timeouts
general_settings: # master_key, database_url, JWT auth, health checks
environment_variables: # Runtime env vars
hosted_vllm/ provider| Variable | Purpose |
|---|---|
LITELLM_MASTER_KEY | Admin key (must start with sk-) |
LITELLM_SALT_KEY | Encryption key for stored credentials (immutable after setup) |
DATABASE_URL | PostgreSQL connection string (required for key management) |
JWT_PUBLIC_KEY_URL | OIDC JWKS endpoint (comma-separated for multiple) |
PROXY_BASE_URL | Public URL of proxy (required for SSO callback) |
| Endpoint | Auth | Purpose |
|---|---|---|
/health | Yes | Full model health (makes real API calls) |
/health/readiness | No | DB connectivity check |
/health/liveliness | No | Alive check |
/health/services | Yes | Integration health (langfuse, datadog) |
--num_workers 8 for scaling, --max_requests_before_restart 10000 for memory recycling--detailed_debug in production/health/liveliness, readiness=/health/readiness, initialDelaySeconds: 120