feat(proxy): serve the auto-router preset catalog at runtime

litellm/__init__.py

@@ -424,6 +424,10 @@ def _dev_env_hot_reload_enabled() -> bool:
     "LITELLM_ANTHROPIC_BETA_HEADERS_URL",
     "https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/anthropic_beta_headers_config.json",
 )
+autorouter_presets_url: str = os.getenv(
+    "LITELLM_AUTOROUTER_PRESETS_URL",
+    "https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/proxy/public_endpoints/autorouter_presets.json",
+)
 suppress_debug_info: bool = False
 dynamodb_table_name: Optional[str] = None
 s3_callback_params: Optional[Dict] = None

litellm/proxy/public_endpoints/public_endpoints.py

@@ -1,11 +1,13 @@
+import asyncio
 import json
 import os
 import re
-from collections.abc import Awaitable, Mapping, Sequence
+from collections.abc import Awaitable, Callable, Mapping, Sequence
 from importlib.resources import files
 from typing import TYPE_CHECKING, Final, Protocol
 
 from fastapi import APIRouter, HTTPException, Request
+from pydantic import TypeAdapter
 from typing_extensions import ReadOnly, TypedDict
 
 import litellm
@@ -28,6 +30,7 @@
 )
 from litellm.types.proxy.public_endpoints.public_endpoints import (
     AgentCreateInfo,
+    AutoRouterPresetRecord,
     ComplexityScorerDefaults,
     ProviderCreateInfo,
     PublicModelHubInfo,
@@ -464,6 +467,86 @@ async def get_litellm_blog_posts():
     return BlogPostsResponse(posts=posts)
 
 
+_AUTOROUTER_PRESETS_ADAPTER: Final = TypeAdapter(dict[str, AutoRouterPresetRecord])
+
+
+def _load_bundled_autorouter_presets() -> Mapping[str, AutoRouterPresetRecord]:
+    raw: Final = json.loads(
+        files("litellm.proxy.public_endpoints").joinpath("autorouter_presets.json").read_text(encoding="utf-8")
+    )
+    return _AUTOROUTER_PRESETS_ADAPTER.validate_python(raw)
+
+
+async def _fetch_remote_autorouter_presets(url: str) -> Mapping[str, AutoRouterPresetRecord]:
+    from litellm.llms.custom_httpx.http_handler import get_async_httpx_client
+    from litellm.types.llms.custom_http import httpxSpecialProvider
+
+    client: Final = get_async_httpx_client(llm_provider=httpxSpecialProvider.UI)
+    response: Final = await client.get(url, timeout=5.0)
+    response.raise_for_status()
+    presets: Final = _AUTOROUTER_PRESETS_ADAPTER.validate_python(response.json())
+    if not presets:
+        raise ValueError("remote auto-router preset catalog is empty")
+    return presets
+
+
+async def _resolve_autorouter_presets(
+    url: str,
+    fetch: Callable[[str], Awaitable[Mapping[str, AutoRouterPresetRecord]]],
+) -> Mapping[str, AutoRouterPresetRecord]:
+    if os.getenv("LITELLM_LOCAL_AUTOROUTER_PRESETS", "").lower() == "true":
+        return _load_bundled_autorouter_presets()
+    try:
+        return await fetch(url)
+    except Exception as e:
+        verbose_logger.warning(
+            "LiteLLM: failed to fetch auto-router presets from %s: %s. Serving the bundled catalog for the life of this process.",
+            url,
+            str(e),
+        )
+        return _load_bundled_autorouter_presets()
+
+
+class _AutoRouterPresetsCache:
+    presets: Mapping[str, AutoRouterPresetRecord] | None = None
+    lock: asyncio.Lock | None = None
+
+
+async def get_autorouter_presets(
+    url: str,
+    fetch: Callable[[str], Awaitable[Mapping[str, AutoRouterPresetRecord]]] = _fetch_remote_autorouter_presets,
+) -> Mapping[str, AutoRouterPresetRecord]:
+    cached: Final = _AutoRouterPresetsCache.presets
+    if cached is not None:
+        return cached
+    if _AutoRouterPresetsCache.lock is None:
+        _AutoRouterPresetsCache.lock = asyncio.Lock()
+    async with _AutoRouterPresetsCache.lock:
+        held: Final = _AutoRouterPresetsCache.presets
+        if held is not None:
+            return held
+        resolved: Final = await _resolve_autorouter_presets(url=url, fetch=fetch)
+        _AutoRouterPresetsCache.presets = resolved
+        return resolved
+
+
+@router.get(
+    "/public/autorouter_presets",
+    tags=["public", "auto router"],  # mutable-ok: FastAPI route tags take a list
+    response_model=dict[str, AutoRouterPresetRecord],
+)
+async def get_public_autorouter_presets() -> Mapping[str, AutoRouterPresetRecord]:
+    """
+    Return the auto-router preset catalog the dashboard's template picker renders.
+
+    Resolved once per process, like the model cost map: fetched from ``litellm.autorouter_presets_url``
+    (override with ``LITELLM_AUTOROUTER_PRESETS_URL``) on the first request, falling back to the
+    catalog bundled with the package on any failure. Set ``LITELLM_LOCAL_AUTOROUTER_PRESETS=True``
+    to serve the bundled catalog only. A restart picks up a newly published catalog.
+    """
+    return await get_autorouter_presets(url=litellm.autorouter_presets_url)
+
+
 @router.get(
     "/public/endpoints",
     tags=["public"],

litellm/types/proxy/public_endpoints/public_endpoints.py

@@ -1,7 +1,7 @@
-from collections.abc import Mapping
+from collections.abc import Mapping, Sequence
 from typing import Any, Literal
 
-from pydantic import BaseModel
+from pydantic import BaseModel, ConfigDict
 
 
 class PublicModelHubInfo(BaseModel):
@@ -73,6 +73,44 @@ class SupportedEndpointsResponse(BaseModel):
     endpoints: list[SupportedEndpoint]
 
 
+class AutoRouterPresetTiers(BaseModel):
+    """Exactly the four built-in tiers the dashboard's preset prefill can apply.
+
+    extra="forbid" on purpose: a tier name this dashboard cannot apply would grey out or crash the
+    picker, so such a catalog is rejected wholesale and the bundled one serves instead.
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    SIMPLE: Sequence[str]
+    MEDIUM: Sequence[str]
+    COMPLEX: Sequence[str]
+    REASONING: Sequence[str]
+
+
+class AutoRouterPresetConfig(BaseModel):
+    """The complexity_router_config a preset prefills.
+
+    Only tiers is validated, because every dashboard consumer dereferences it; everything else
+    passes through verbatim with unknown fields kept (extra="allow"), so a catalog published after
+    this proxy shipped still serves its new fields intact.
+    """
+
+    model_config = ConfigDict(extra="allow")
+
+    tiers: AutoRouterPresetTiers
+
+
+class AutoRouterPresetRecord(BaseModel):
+    """One auto-router preset as served to the dashboard's template picker."""
+
+    model_config = ConfigDict(extra="allow")
+
+    label: str
+    description: str
+    complexity_router_config: AutoRouterPresetConfig
+
+
 class ComplexityScorerDefaults(BaseModel):
     """The complexity router's shipped heuristic scorer defaults.