diff --git a/litellm/__init__.py b/litellm/__init__.py index a611b90bc37..34f08da3c89 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -440,6 +440,10 @@ autorouter_presets_url: str = os.getenv( "LITELLM_AUTOROUTER_PRESETS_URL", "https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/proxy/public_endpoints/autorouter_presets.json", ) +whats_new_url: str = os.getenv( + "LITELLM_WHATS_NEW_URL", + "https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/proxy/public_endpoints/whats_new.json", +) suppress_debug_info: bool = False dynamodb_table_name: Optional[str] = None s3_callback_params: Optional[Dict] = None diff --git a/litellm/proxy/public_endpoints/public_endpoints.py b/litellm/proxy/public_endpoints/public_endpoints.py index 47caacdcfff..75e8a5af6a5 100644 --- a/litellm/proxy/public_endpoints/public_endpoints.py +++ b/litellm/proxy/public_endpoints/public_endpoints.py @@ -6,8 +6,9 @@ from collections.abc import Awaitable, Callable, Mapping, Sequence from importlib.resources import files from typing import TYPE_CHECKING, Final, Protocol +import httpx from fastapi import APIRouter, HTTPException, Request -from pydantic import TypeAdapter +from pydantic import TypeAdapter, ValidationError from typing_extensions import ReadOnly, TypedDict import litellm @@ -37,6 +38,7 @@ from litellm.types.proxy.public_endpoints.public_endpoints import ( ProviderCreateInfo, PublicModelHubInfo, SupportedEndpointsResponse, + WhatsNewResponse, ) from litellm.types.utils import LlmProviders @@ -505,14 +507,18 @@ def _load_bundled_autorouter_presets() -> Mapping[str, AutoRouterPresetRecord]: ) -async def _fetch_remote_autorouter_presets(url: str) -> Mapping[str, AutoRouterPresetRecord]: +async def _fetch_remote_bytes(url: str) -> bytes: 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()) + return response.content + + +async def _fetch_remote_autorouter_presets(url: str) -> Mapping[str, AutoRouterPresetRecord]: + presets: Final = _AUTOROUTER_PRESETS_ADAPTER.validate_json(await _fetch_remote_bytes(url)) if not presets: raise ValueError("remote auto-router preset catalog is empty") return presets @@ -575,6 +581,70 @@ async def get_public_autorouter_presets() -> Mapping[str, AutoRouterPresetRecord return await get_autorouter_presets(url=litellm.autorouter_presets_url) +def _load_bundled_whats_new() -> WhatsNewResponse: + return WhatsNewResponse.model_validate_json( + files("litellm.proxy.public_endpoints").joinpath("whats_new.json").read_text(encoding="utf-8") + ) + + +async def _fetch_remote_whats_new(url: str) -> WhatsNewResponse: + return WhatsNewResponse.model_validate_json(await _fetch_remote_bytes(url)) + + +async def _resolve_whats_new(url: str, fetch: Callable[[str], Awaitable[WhatsNewResponse]]) -> WhatsNewResponse: + if os.getenv("LITELLM_LOCAL_WHATS_NEW", "").lower() == "true": + return _load_bundled_whats_new() + try: + return await fetch(url) + except (httpx.HTTPError, ValidationError, OSError) as e: + verbose_logger.warning( + "LiteLLM: failed to fetch What's new launches from %s: %s. Serving the bundled list for the life of this process.", + url, + str(e), + ) + return _load_bundled_whats_new() + + +class _WhatsNewCache: + launches: WhatsNewResponse | None = None + lock: asyncio.Lock | None = None + + +async def get_whats_new( + url: str, + fetch: Callable[[str], Awaitable[WhatsNewResponse]] = _fetch_remote_whats_new, +) -> WhatsNewResponse: + cached: Final = _WhatsNewCache.launches + if cached is not None: + return cached + if _WhatsNewCache.lock is None: + _WhatsNewCache.lock = asyncio.Lock() + async with _WhatsNewCache.lock: + held: Final = _WhatsNewCache.launches + if held is not None: + return held + resolved: Final = await _resolve_whats_new(url=url, fetch=fetch) + _WhatsNewCache.launches = resolved + return resolved + + +@router.get( + "/public/whats_new", + tags=["public"], + response_model=WhatsNewResponse, +) +async def get_public_whats_new() -> WhatsNewResponse: + """ + Return the launches the dashboard Home page shows under What's new. + + Resolved once per process: fetched from ``litellm.whats_new_url`` (override with ``LITELLM_WHATS_NEW_URL``) + on the first request, falling back to the list bundled with the package on any failure, so an airgapped + proxy serves the bundled list without retrying. Set ``LITELLM_LOCAL_WHATS_NEW=True`` to skip the fetch. + A restart picks up a newly published list. + """ + return await get_whats_new(url=litellm.whats_new_url) + + @router.get( "/public/endpoints", tags=["public"], diff --git a/litellm/proxy/public_endpoints/whats_new.json b/litellm/proxy/public_endpoints/whats_new.json new file mode 100644 index 00000000000..40affb71e98 --- /dev/null +++ b/litellm/proxy/public_endpoints/whats_new.json @@ -0,0 +1,25 @@ +{ + "launches": [ + { + "icon": "box", + "title": "Claude Haiku 5.5", + "description": "Anthropic's newest small model, with day 0 pricing on Anthropic, Bedrock and Vertex AI.", + "href": "https://docs.litellm.ai/blog/claude-haiku-5-5", + "published_on": "2026-10-07" + }, + { + "icon": "scale", + "title": "Decision Models", + "description": "Call decision models at /v1/decisions and try them in the Decisions playground.", + "href": "https://docs.litellm.ai/docs/decisions", + "published_on": "2026-10-07" + }, + { + "icon": "zap", + "title": "OpenAI Ultrafast", + "description": "service_tier: ultrafast on GPT-6 Astra and GPT-6.1 Sol, including /ultrafast in Codex.", + "href": "https://docs.litellm.ai/docs/providers/openai/ultrafast", + "published_on": "2026-10-06" + } + ] +} diff --git a/litellm/types/proxy/public_endpoints/public_endpoints.py b/litellm/types/proxy/public_endpoints/public_endpoints.py index c487ce8b6c5..6e4adb4bbe7 100644 --- a/litellm/types/proxy/public_endpoints/public_endpoints.py +++ b/litellm/types/proxy/public_endpoints/public_endpoints.py @@ -1,4 +1,5 @@ from collections.abc import Mapping, Sequence +from datetime import date from typing import Any, Literal from pydantic import ConfigDict @@ -113,6 +114,24 @@ class AutoRouterPresetRecord(LiteLLMBaseModel): complexity_router_config: AutoRouterPresetConfig +class WhatsNewLaunch(LiteLLMBaseModel): + """One launch card in the dashboard Home page's What's new section.""" + + model_config = ConfigDict(frozen=True) + + icon: str + title: str + description: str + href: str + published_on: date + + +class WhatsNewResponse(LiteLLMBaseModel): + model_config = ConfigDict(frozen=True) + + launches: tuple[WhatsNewLaunch, ...] + + class ComplexityScorerDefaults(LiteLLMBaseModel): """The complexity router's shipped heuristic scorer defaults. diff --git a/pyproject.toml b/pyproject.toml index 133fc735df1..0ed8b8757ac 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -321,6 +321,7 @@ include = [ "litellm/router_strategy/complexity_router/artifacts/*.json", "litellm/router_strategy/complexity_router/fuse_presets.json", "litellm/proxy/model_insights_tasks.json", + "litellm/proxy/public_endpoints/whats_new.json", "litellm/proxy/common_utils/codex_base_instructions.md", "litellm/proxy/common_utils/codex_bundled_models_0.159.3.json", ] diff --git a/tests/unit/proxy/public_endpoints/test_public_endpoints.py b/tests/unit/proxy/public_endpoints/test_public_endpoints.py index b720ba24c3e..0779f692b17 100644 --- a/tests/unit/proxy/public_endpoints/test_public_endpoints.py +++ b/tests/unit/proxy/public_endpoints/test_public_endpoints.py @@ -1,6 +1,8 @@ import json import re from datetime import datetime, timezone +from importlib.resources import files +from collections.abc import Iterator from typing import Final from unittest.mock import AsyncMock, MagicMock, patch @@ -17,7 +19,7 @@ from litellm.router_strategy.complexity_router.fuse_presets import get_fuse_pres from litellm.types.proxy.management_endpoints.model_management_endpoints import ( ModelGroupInfoProxy, ) -from litellm.types.proxy.public_endpoints.public_endpoints import ProviderCreateInfo +from litellm.types.proxy.public_endpoints.public_endpoints import ProviderCreateInfo, WhatsNewResponse from litellm.types.utils import LlmProviders @@ -67,13 +69,10 @@ def test_get_provider_create_fields(): assert isinstance(first_provider["credential_fields"], list) has_detailed_fields = any( - provider.get("credential_fields") - and len(provider.get("credential_fields", [])) > 0 + provider.get("credential_fields") and len(provider.get("credential_fields", [])) > 0 for provider in response_data ) - assert ( - has_detailed_fields - ), "Expected at least one provider to have detailed credential fields" + assert has_detailed_fields, "Expected at least one provider to have detailed credential fields" def test_get_litellm_model_cost_map_catalog_only_excludes_runtime_registered_entries( @@ -124,10 +123,7 @@ def test_get_litellm_model_cost_map_returns_cost_map(): sample_model_data = payload[sample_model] assert isinstance(sample_model_data, dict) # Check for common cost fields that should be present - assert ( - "input_cost_per_token" in sample_model_data - or "output_cost_per_token" in sample_model_data - ) + assert "input_cost_per_token" in sample_model_data or "output_cost_per_token" in sample_model_data def test_public_ai_hub_info_is_public_by_default(monkeypatch): @@ -218,9 +214,9 @@ def test_anthropic_provider_fields_support_byok(): "Anthropic api_key must be optional so admins can configure BYOK models " "without entering a key. See BYOK tutorial." ) - assert fields_by_key["api_key"].get( - "tooltip" - ), "Anthropic api_key must have a tooltip explaining the BYOK use case." + assert fields_by_key["api_key"].get("tooltip"), ( + "Anthropic api_key must have a tooltip explaining the BYOK use case." + ) assert "api_base" in fields_by_key, ( "Anthropic provider form must expose api_base so cloud customers " "can override the upstream URL without env var access." @@ -228,16 +224,14 @@ def test_anthropic_provider_fields_support_byok(): api_base_field = fields_by_key["api_base"] assert api_base_field["required"] is False assert api_base_field["field_type"] == "text" - assert api_base_field.get( - "tooltip" - ), "api_base should have a tooltip explaining it is optional." + assert api_base_field.get("tooltip"), "api_base should have a tooltip explaining it is optional." # UI forms render fields in credential_fields order; api_base should come first # so an admin sees the URL override before the key field. field_order = [f["key"] for f in anthropic["credential_fields"]] - assert field_order.index("api_base") < field_order.index( - "api_key" - ), "api_base must appear before api_key in credential_fields (matches AI21 and ANTHROPIC_TEXT convention)." + assert field_order.index("api_base") < field_order.index("api_key"), ( + "api_base must appear before api_key in credential_fields (matches AI21 and ANTHROPIC_TEXT convention)." + ) def test_bedrock_mantle_provider_fields(): @@ -554,9 +548,7 @@ def test_google_ai_studio_provider_fields_expose_api_base(): assert response.status_code == 200 providers = response.json() - google_ai = next( - (p for p in providers if p["provider"] == "Google_AI_Studio"), None - ) + google_ai = next((p for p in providers if p["provider"] == "Google_AI_Studio"), None) assert google_ai is not None, "Google_AI_Studio provider entry not found" assert google_ai["litellm_provider"] == "gemini" @@ -576,18 +568,15 @@ def test_google_ai_studio_provider_fields_expose_api_base(): # placeholder shows the canonical URL so users still get the visual hint. # (See greptileai threads on PR #30419.) assert api_base_field["default_value"] is None - assert ( - api_base_field["placeholder"] - == "https://generativelanguage.googleapis.com/v1beta" - ) + assert api_base_field["placeholder"] == "https://generativelanguage.googleapis.com/v1beta" # UI forms render fields in credential_fields order; api_base should come # first so an admin sees the URL override before the key field (matches # OpenAI and Anthropic conventions). field_order = [f["key"] for f in google_ai["credential_fields"]] - assert field_order.index("api_base") < field_order.index( - "api_key" - ), "api_base must appear before api_key in credential_fields." + assert field_order.index("api_base") < field_order.index("api_key"), ( + "api_base must appear before api_key in credential_fields." + ) def test_public_model_hub_with_healthy_model(): @@ -615,20 +604,15 @@ def test_public_model_hub_with_healthy_model(): mock_llm_router = MagicMock() mock_prisma = MagicMock() - mock_prisma.get_all_latest_health_checks = AsyncMock( - return_value=[mock_health_check] - ) + mock_prisma.get_all_latest_health_checks = AsyncMock(return_value=[mock_health_check]) with ( patch("litellm.public_model_groups", ["gpt-3.5-turbo"]), patch("litellm.proxy.proxy_server.get_model_group_info") as mock_get_info, patch("litellm.proxy.proxy_server.llm_router", mock_llm_router), patch("litellm.proxy.proxy_server.prisma_client", mock_prisma), - patch( - "litellm.proxy.health_endpoints._health_endpoints.convert_health_check_to_dict" - ) as mock_convert, + patch("litellm.proxy.health_endpoints._health_endpoints.convert_health_check_to_dict") as mock_convert, ): - mock_get_info.return_value = [mock_model_group] mock_convert.return_value = { "status": "healthy", @@ -673,20 +657,15 @@ def test_public_model_hub_with_unhealthy_model(): mock_llm_router = MagicMock() mock_prisma = MagicMock() - mock_prisma.get_all_latest_health_checks = AsyncMock( - return_value=[mock_health_check] - ) + mock_prisma.get_all_latest_health_checks = AsyncMock(return_value=[mock_health_check]) with ( patch("litellm.public_model_groups", ["gpt-4"]), patch("litellm.proxy.proxy_server.get_model_group_info") as mock_get_info, patch("litellm.proxy.proxy_server.llm_router", mock_llm_router), patch("litellm.proxy.proxy_server.prisma_client", mock_prisma), - patch( - "litellm.proxy.health_endpoints._health_endpoints.convert_health_check_to_dict" - ) as mock_convert, + patch("litellm.proxy.health_endpoints._health_endpoints.convert_health_check_to_dict") as mock_convert, ): - mock_get_info.return_value = [mock_model_group] mock_convert.return_value = { "status": "unhealthy", @@ -732,7 +711,6 @@ def test_public_model_hub_without_health_check(): patch("litellm.proxy.proxy_server.llm_router", mock_llm_router), patch("litellm.proxy.proxy_server.prisma_client", mock_prisma), ): - mock_get_info.return_value = [mock_model_group] response = client.get( @@ -789,9 +767,7 @@ def test_public_model_hub_mixed_health_statuses(): mock_llm_router = MagicMock() mock_prisma = MagicMock() - mock_prisma.get_all_latest_health_checks = AsyncMock( - return_value=[healthy_check, unhealthy_check] - ) + mock_prisma.get_all_latest_health_checks = AsyncMock(return_value=[healthy_check, unhealthy_check]) def convert_side_effect(check): if check.model_name == "gpt-3.5-turbo": @@ -813,11 +789,8 @@ def test_public_model_hub_mixed_health_statuses(): patch("litellm.proxy.proxy_server.get_model_group_info") as mock_get_info, patch("litellm.proxy.proxy_server.llm_router", mock_llm_router), patch("litellm.proxy.proxy_server.prisma_client", mock_prisma), - patch( - "litellm.proxy.health_endpoints._health_endpoints.convert_health_check_to_dict" - ) as mock_convert, + patch("litellm.proxy.health_endpoints._health_endpoints.convert_health_check_to_dict") as mock_convert, ): - mock_get_info.return_value = [ healthy_model, unhealthy_model, @@ -1066,9 +1039,7 @@ def test_get_supported_endpoints_provider_fields(reset_endpoints_cache): def test_get_supported_endpoints_paths_start_with_slash(reset_endpoints_cache): endpoints = _make_client().get("/public/endpoints").json()["endpoints"] for item in endpoints: - assert item["endpoint"].startswith( - "/" - ), f"Expected path starting with /, got: {item['endpoint']}" + assert item["endpoint"].startswith("/"), f"Expected path starting with /, got: {item['endpoint']}" def test_get_supported_endpoints_chat_completions_present(reset_endpoints_cache): @@ -1092,9 +1063,9 @@ def test_get_supported_endpoints_display_names_have_no_slug_suffix( endpoints = _make_client().get("/public/endpoints").json()["endpoints"] for item in endpoints: for provider in item["providers"]: - assert not suffix_re.search( - provider["display_name"] - ), f"display_name still contains slug suffix: {provider['display_name']!r}" + assert not suffix_re.search(provider["display_name"]), ( + f"display_name still contains slug suffix: {provider['display_name']!r}" + ) def test_get_supported_endpoints_is_cached(reset_endpoints_cache): @@ -1320,7 +1291,6 @@ def test_public_mcp_hub_does_not_expose_upstream_url(): app.dependency_overrides.clear() - @pytest.fixture def reset_autorouter_presets_cache(): from litellm.proxy.public_endpoints.public_endpoints import _AutoRouterPresetsCache @@ -1332,9 +1302,7 @@ def reset_autorouter_presets_cache(): _AutoRouterPresetsCache.lock = None -def test_get_autorouter_presets_local_mode_serves_bundled_catalog( - monkeypatch, reset_autorouter_presets_cache -): +def test_get_autorouter_presets_local_mode_serves_bundled_catalog(monkeypatch, reset_autorouter_presets_cache): monkeypatch.setenv("LITELLM_LOCAL_AUTOROUTER_PRESETS", "True") app = FastAPI() app.include_router(router) @@ -1362,9 +1330,7 @@ def test_get_autorouter_presets_local_mode_serves_bundled_catalog( @pytest.mark.asyncio -async def test_get_autorouter_presets_fetches_once_per_process( - monkeypatch, reset_autorouter_presets_cache -): +async def test_get_autorouter_presets_fetches_once_per_process(monkeypatch, reset_autorouter_presets_cache): from litellm.proxy.public_endpoints.public_endpoints import ( _AUTOROUTER_PRESETS_ADAPTER, get_autorouter_presets, @@ -1376,7 +1342,9 @@ async def test_get_autorouter_presets_fetches_once_per_process( "remote_only": { "label": "Remote Only", "description": "from the remote catalog", - "complexity_router_config": {"tiers": {"SIMPLE": ["m1"], "MEDIUM": ["m2"], "COMPLEX": ["m3"], "REASONING": ["m4"]}}, + "complexity_router_config": { + "tiers": {"SIMPLE": ["m1"], "MEDIUM": ["m2"], "COMPLEX": ["m3"], "REASONING": ["m4"]} + }, } } ) @@ -1411,7 +1379,9 @@ async def test_get_autorouter_presets_single_flight_on_concurrent_cold_start( "remote_only": { "label": "Remote Only", "description": "from the remote catalog", - "complexity_router_config": {"tiers": {"SIMPLE": ["m1"], "MEDIUM": ["m2"], "COMPLEX": ["m3"], "REASONING": ["m4"]}}, + "complexity_router_config": { + "tiers": {"SIMPLE": ["m1"], "MEDIUM": ["m2"], "COMPLEX": ["m3"], "REASONING": ["m4"]} + }, } } ) @@ -1507,9 +1477,7 @@ async def test_autorouter_presets_adapter_rejects_wrong_shapes(): ) -def test_get_autorouter_presets_passes_unknown_catalog_fields_through( - monkeypatch, reset_autorouter_presets_cache -): +def test_get_autorouter_presets_passes_unknown_catalog_fields_through(monkeypatch, reset_autorouter_presets_cache): from litellm.proxy.public_endpoints.public_endpoints import ( _AUTOROUTER_PRESETS_ADAPTER, _AutoRouterPresetsCache, @@ -1551,12 +1519,14 @@ async def test_fetch_remote_autorouter_presets_parses_and_rejects_empty(monkeypa "remote_only": { "label": "Remote Only", "description": "from the remote catalog", - "complexity_router_config": {"tiers": {"SIMPLE": ["m1"], "MEDIUM": ["m2"], "COMPLEX": ["m3"], "REASONING": ["m4"]}}, + "complexity_router_config": { + "tiers": {"SIMPLE": ["m1"], "MEDIUM": ["m2"], "COMPLEX": ["m3"], "REASONING": ["m4"]} + }, } } response = MagicMock() response.raise_for_status = MagicMock() - response.json = MagicMock(return_value=catalog) + response.content = json.dumps(catalog).encode() client = MagicMock() client.get = AsyncMock(return_value=response) monkeypatch.setattr(http_handler_module, "get_async_httpx_client", lambda llm_provider: client) @@ -1565,6 +1535,121 @@ async def test_fetch_remote_autorouter_presets_parses_and_rejects_empty(monkeypa assert presets["remote_only"].label == "Remote Only" response.raise_for_status.assert_called_once() - response.json = MagicMock(return_value={}) + response.content = b"{}" with pytest.raises(ValueError, match="empty"): await _fetch_remote_autorouter_presets("https://example.test/presets.json") + + +@pytest.fixture +def reset_whats_new_cache() -> Iterator[None]: + from litellm.proxy.public_endpoints.public_endpoints import _WhatsNewCache + + _WhatsNewCache.launches = None + _WhatsNewCache.lock = None + yield + _WhatsNewCache.launches = None + _WhatsNewCache.lock = None + + +def _remote_whats_new() -> WhatsNewResponse: + return WhatsNewResponse.model_validate( + { + "launches": [ + { + "icon": "sparkles", + "title": "Remote Launch", + "description": "from the remote list", + "href": "https://docs.litellm.ai/blog/remote", + "published_on": "2026-10-11", + } + ] + } + ) + + +def test_get_whats_new_local_mode_serves_bundled_list( + monkeypatch: pytest.MonkeyPatch, reset_whats_new_cache: None +) -> None: + monkeypatch.setenv("LITELLM_LOCAL_WHATS_NEW", "True") + app = FastAPI() + app.include_router(router) + client = TestClient(app) + + response = client.get("/public/whats_new") + + assert response.status_code == 200 + served: Final = WhatsNewResponse.model_validate_json(response.content) + bundled: Final = WhatsNewResponse.model_validate_json( + files("litellm.proxy.public_endpoints").joinpath("whats_new.json").read_text(encoding="utf-8") + ) + assert served == bundled + assert len(served.launches) > 0 + + +@pytest.mark.asyncio +async def test_get_whats_new_fetches_once_per_process( + monkeypatch: pytest.MonkeyPatch, reset_whats_new_cache: None +) -> None: + from litellm.proxy.public_endpoints.public_endpoints import get_whats_new + + monkeypatch.delenv("LITELLM_LOCAL_WHATS_NEW", raising=False) + remote = _remote_whats_new() + calls: Final[list[str]] = [] + + async def fake_fetch(url: str) -> WhatsNewResponse: + calls.append(url) + return remote + + first = await get_whats_new(url="https://example.test/whats_new.json", fetch=fake_fetch) + second = await get_whats_new(url="https://example.test/whats_new.json", fetch=fake_fetch) + + assert first == remote + assert second == remote + assert calls == ["https://example.test/whats_new.json"] + + +@pytest.mark.asyncio +async def test_get_whats_new_caches_bundled_fallback_when_remote_is_unreachable( + monkeypatch: pytest.MonkeyPatch, reset_whats_new_cache: None +) -> None: + from litellm.proxy.public_endpoints.public_endpoints import _load_bundled_whats_new, get_whats_new + + monkeypatch.delenv("LITELLM_LOCAL_WHATS_NEW", raising=False) + calls: Final[list[str]] = [] + + async def unreachable_fetch(url: str) -> WhatsNewResponse: + calls.append(url) + raise OSError("network is unreachable") + + first = await get_whats_new(url="https://example.test/whats_new.json", fetch=unreachable_fetch) + second = await get_whats_new(url="https://example.test/whats_new.json", fetch=unreachable_fetch) + + assert first == _load_bundled_whats_new() + assert second == first + assert len(calls) == 1 + + +@pytest.mark.asyncio +async def test_fetch_remote_whats_new_rejects_a_malformed_list_so_the_bundled_one_is_served( + monkeypatch: pytest.MonkeyPatch, reset_whats_new_cache: None +) -> None: + import httpx + + from litellm.proxy.public_endpoints import public_endpoints + + monkeypatch.delenv("LITELLM_LOCAL_WHATS_NEW", raising=False) + request = httpx.Request("GET", "https://example.test/whats_new.json") + client = MagicMock() + client.get = AsyncMock( + return_value=httpx.Response( + 200, + request=request, + content=b'{"launches": [{"icon": "box", "title": "No date", "description": "d", "href": "https://x"}]}', + ) + ) + + with patch("litellm.llms.custom_httpx.http_handler.get_async_httpx_client", return_value=client): + served = await public_endpoints.get_whats_new(url="https://example.test/whats_new.json") + + assert served == public_endpoints._load_bundled_whats_new() + client.get.assert_awaited_once_with("https://example.test/whats_new.json", timeout=5.0) diff --git a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.integration.test.tsx index ad3f80b7d78..a17a18beee5 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.integration.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.integration.test.tsx @@ -6,7 +6,6 @@ import type { DailyActivityAggregatedResponse } from "@/components/UsagePage/dai import { EMPTY_DAILY_ACTIVITY_METADATA } from "@/components/UsagePage/dailyActivityApi"; import { all_admin_roles } from "@/utils/roles"; import HomePage from "./HomePage"; -import { WHATS_NEW_ITEMS } from "./homeContent"; const auth = vi.hoisted(() => ({ accessToken: "sk-test", @@ -31,6 +30,28 @@ vi.mock("@/components/networking", async (importOriginal) => ({ const blogPosts = { posts: [{ title: "Post one", description: "d", date: "2026-10-09", url: "https://x/1" }] }; +const launch = (title: string, icon: string, published_on: string) => ({ + icon, + title, + description: `${title} description`, + href: `https://docs.litellm.ai/blog/${title}`, + published_on, +}); + +const whatsNew = vi.hoisted(() => ({ launches: [] as unknown[] })); + +const requestUrl = (input: RequestInfo | URL): string => { + if (typeof input === "string") return input; + return input instanceof URL ? input.href : input.url; +}; + +const fakeFetch = (input: RequestInfo | URL) => { + const body = requestUrl(input).endsWith("/public/whats_new") ? whatsNew : blogPosts; + return Promise.resolve( + new Response(JSON.stringify(body), { status: 200, headers: { "Content-Type": "application/json" } }), + ); +}; + const dayResult = (date: string, model: string, tokens: number) => { const metrics = { spend: 1, @@ -86,7 +107,8 @@ describe("HomePage", () => { network.aggregated.mockReset().mockResolvedValue(activity); network.gateway.mockReset().mockResolvedValue({ total_successful_requests: 0, total_failed_requests: 0 }); network.userInfo.mockReset().mockResolvedValue({ user_info: { max_budget: null } }); - vi.stubGlobal("fetch", vi.fn().mockResolvedValue(new Response(JSON.stringify(blogPosts), { status: 200 }))); + whatsNew.launches = [launch("mid", "box", "2026-10-06"), launch("new", "zap", "2026-10-11")]; + vi.stubGlobal("fetch", vi.fn(fakeFetch)); }); afterEach(() => { vi.unstubAllGlobals(); @@ -121,13 +143,32 @@ describe("HomePage", () => { expect(network.aggregated).toHaveBeenCalledTimes(2); }); - it("lists the curated launches newest first with their links", () => { + it("renders the launches served by /public/whats_new newest first, whatever order the JSON lists them in", async () => { + whatsNew.launches = [ + launch("mid", "box", "2026-10-06"), + launch("new", "an-icon-from-a-newer-list", "2026-10-11"), + launch("old", "scale", "2026-09-30"), + ]; renderHome(); - const section = screen.getByTestId("home-whats-new"); - const links = within(section).getAllByRole("link"); - expect(links.map((link) => link.getAttribute("href"))).toEqual(WHATS_NEW_ITEMS.map((item) => item.href)); - const dates = WHATS_NEW_ITEMS.map((item) => item.publishedOn); - expect(dates).toEqual([...dates].sort().reverse()); + const section = await screen.findByTestId("home-whats-new"); + expect(await within(section).findByText("new")).toBeInTheDocument(); + expect( + within(section) + .getAllByRole("link") + .map((link) => link.getAttribute("href")), + ).toEqual([ + "https://docs.litellm.ai/blog/new", + "https://docs.litellm.ai/blog/mid", + "https://docs.litellm.ai/blog/old", + ]); + expect(within(section).getByText("Oct 11")).toBeInTheDocument(); + }); + + it("hides What's new when /public/whats_new has no launches", async () => { + whatsNew.launches = []; + renderHome(); + expect(await screen.findByRole("link", { name: /post one/i })).toBeInTheDocument(); + expect(screen.queryByTestId("home-whats-new")).not.toBeInTheDocument(); }); it("ranks model groups by token share on the leaderboard", async () => { diff --git a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.tsx b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.tsx index f4519958e9d..bcae30795c6 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/HomePage.tsx @@ -4,21 +4,24 @@ import { ArrowRight, Plus, Sparkles } from "lucide-react"; import useAuthorized from "@/app/(dashboard)/hooks/useAuthorized"; import { Button } from "@/components/ui/button"; import { Card } from "@/components/ui/card"; +import { Skeleton } from "@/components/ui/skeleton"; import { Page } from "@/components/shared/Page"; import { PageHeader, PageHeaderControls, PageHeaderTitle } from "@/components/shared/PageHeader"; import { uiHref } from "@/utils/uiHref"; import { Leaderboard, useBreakdown } from "@/app/(dashboard)/usage/_components/components/overview/BreakdownChart"; import { Panel } from "@/app/(dashboard)/usage/_components/components/overview/Primitives"; import UsageStatStrip from "@/app/(dashboard)/usage/_components/components/overview/UsageStatStrip"; -import { formatPublishedOn, WHATS_NEW_ITEMS } from "./homeContent"; +import { formatPublishedOn, launchIcon } from "./homeContent"; import UpdatesFeed from "./UpdatesFeed"; import { HOME_USAGE_DAYS, useHomeUsage } from "./useHomeUsage"; +import { useWhatsNew } from "./useWhatsNew"; const LEADERBOARD = { metric: "tokens", dimension: "model_groups" } as const; export default function HomePage() { const { isViewOnly } = useAuthorized(); const usage = useHomeUsage(); + const whatsNew = useWhatsNew(); const { series, ranking } = useBreakdown(usage.results, LEADERBOARD, 8); return ( @@ -56,34 +59,40 @@ export default function HomePage() {
-
-
-

- - What's new in LiteLLM? -

-
-
- {WHATS_NEW_ITEMS.map((item) => { - const Icon = item.icon; - return ( - - -
- -
-
{item.title}
-

{item.description}

-
- {formatPublishedOn(item.publishedOn)} - -
- - - ); - })} -
-
+ {(whatsNew.isLoading || (whatsNew.data?.length ?? 0) > 0) && ( +
+
+

+ + What's new in LiteLLM? +

+
+
+ {whatsNew.isLoading && + [0, 1, 2].map((i) => ( + + ))} + {whatsNew.data?.map((item) => { + const Icon = launchIcon(item.icon); + return ( + + +
+ +
+
{item.title}
+

{item.description}

+
+ {formatPublishedOn(item.published_on)} + +
+ + + ); + })} +
+
+ )} ({ + icon: "box", + title, + description: "d", + href: `https://docs.litellm.ai/${title}`, + published_on, +}); + +describe("latestLaunches", () => { + it("orders launches by published_on, newest first, without mutating the input", () => { + const launches = [launch("old", "2026-09-01"), launch("new", "2026-10-11"), launch("mid", "2026-10-01")]; + expect(latestLaunches(launches).map((l) => l.title)).toEqual(["new", "mid", "old"]); + expect(launches.map((l) => l.title)).toEqual(["old", "new", "mid"]); + }); + + it("keeps only the three newest so the cards fill one row", () => { + const launches = [ + launch("oldest", "2026-08-01"), + launch("a", "2026-10-03"), + launch("b", "2026-10-02"), + launch("c", "2026-10-01"), + ]; + expect(latestLaunches(launches).map((l) => l.title)).toEqual(["a", "b", "c"]); + }); +}); + +describe("launchIcon", () => { + it("maps a known icon name to its lucide icon", () => { + expect(launchIcon("box")).toBe(Box); + }); + + it("falls back to Sparkles for an icon name this dashboard version does not know", () => { + expect(launchIcon("hologram")).toBe(Sparkles); + expect(launchIcon("constructor")).toBe(Sparkles); + }); +}); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/homeContent.ts b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/homeContent.ts index 41d1aac25de..c3eb4b2c838 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/homeContent.ts +++ b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/homeContent.ts @@ -1,40 +1,26 @@ -import { Box, Scale, Zap, type LucideIcon } from "lucide-react"; +import { Bot, Box, Gauge, Plug, Rocket, Scale, Shield, Sparkles, Zap, type LucideIcon } from "lucide-react"; +import type { components } from "@/lib/http/schema"; -export interface Launch { - readonly icon: LucideIcon; - readonly title: string; - readonly description: string; - readonly href: string; - readonly publishedOn: string; -} +export type WhatsNewLaunch = components["schemas"]["WhatsNewLaunch"]; -const LAUNCHES: readonly Launch[] = [ - { - icon: Box, - title: "Claude Haiku 5.5", - description: "Anthropic's newest small model, with day 0 pricing on Anthropic, Bedrock and Vertex AI.", - href: "https://docs.litellm.ai/blog/claude-haiku-5-5", - publishedOn: "2026-10-07", - }, - { - icon: Scale, - title: "Decision Models", - description: "Call decision models at /v1/decisions and try them in the Decisions playground.", - href: "https://docs.litellm.ai/docs/decisions", - publishedOn: "2026-10-07", - }, - { - icon: Zap, - title: "OpenAI Ultrafast", - description: "service_tier: ultrafast on GPT-6 Astra and GPT-6.1 Sol, including /ultrafast in Codex.", - href: "https://docs.litellm.ai/docs/providers/openai/ultrafast", - publishedOn: "2026-10-06", - }, -]; +const LAUNCH_ICONS: ReadonlyMap = new Map([ + ["bot", Bot], + ["box", Box], + ["gauge", Gauge], + ["plug", Plug], + ["rocket", Rocket], + ["scale", Scale], + ["shield", Shield], + ["sparkles", Sparkles], + ["zap", Zap], +]); -export const WHATS_NEW_ITEMS: readonly Launch[] = [...LAUNCHES].sort((a, b) => - b.publishedOn.localeCompare(a.publishedOn), -); +export const launchIcon = (name: string): LucideIcon => LAUNCH_ICONS.get(name) ?? Sparkles; + +export const WHATS_NEW_MAX = 3; + +export const latestLaunches = (launches: readonly WhatsNewLaunch[]): readonly WhatsNewLaunch[] => + [...launches].sort((a, b) => b.published_on.localeCompare(a.published_on)).slice(0, WHATS_NEW_MAX); export const formatPublishedOn = (isoDate: string): string => new Date(`${isoDate}T00:00:00`).toLocaleDateString("en-US", { month: "short", day: "numeric" }); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/home/_components/useWhatsNew.ts b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/useWhatsNew.ts new file mode 100644 index 00000000000..c2171f3349a --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/home/_components/useWhatsNew.ts @@ -0,0 +1,10 @@ +import { $api } from "@/lib/http/api"; +import { latestLaunches } from "./homeContent"; + +export const useWhatsNew = () => + $api.useQuery( + "get", + "/public/whats_new", + {}, + { staleTime: 60 * 60 * 1000, select: (data) => latestLaunches(data.launches) }, + ); diff --git a/ui/litellm-dashboard/src/lib/http/schema.d.ts b/ui/litellm-dashboard/src/lib/http/schema.d.ts index c56a4b3786c..61354f8d9fa 100644 --- a/ui/litellm-dashboard/src/lib/http/schema.d.ts +++ b/ui/litellm-dashboard/src/lib/http/schema.d.ts @@ -13658,6 +13658,31 @@ export interface paths { patch?: never; trace?: never; }; + "/public/whats_new": { + parameters: { + query?: never; + header?: never; + path?: never; + cookie?: never; + }; + /** + * Get Public Whats New + * @description Return the launches the dashboard Home page shows under What's new. + * + * Resolved once per process: fetched from ``litellm.whats_new_url`` (override with ``LITELLM_WHATS_NEW_URL``) + * on the first request, falling back to the list bundled with the package on any failure, so an airgapped + * proxy serves the bundled list without retrying. Set ``LITELLM_LOCAL_WHATS_NEW=True`` to skip the fetch. + * A restart picks up a newly published list. + */ + get: operations["get_public_whats_new_public_whats_new_get"]; + put?: never; + post?: never; + delete?: never; + options?: never; + head?: never; + patch?: never; + trace?: never; + }; "/queue/chat/completions": { parameters: { query?: never; @@ -49245,6 +49270,30 @@ export interface components { type: "web_search" | "web_search_2025_08_26"; user_location?: components["schemas"]["openai__types__responses__web_search_tool_param__UserLocation"] | null; }; + /** + * WhatsNewLaunch + * @description One launch card in the dashboard Home page's What's new section. + */ + WhatsNewLaunch: { + /** Description */ + description: string; + /** Href */ + href: string; + /** Icon */ + icon: string; + /** + * Published On + * Format: date + */ + published_on: string; + /** Title */ + title: string; + }; + /** WhatsNewResponse */ + WhatsNewResponse: { + /** Launches */ + launches: components["schemas"]["WhatsNewLaunch"][]; + }; /** WorkerRegistryEntry */ WorkerRegistryEntry: { /** Name */ @@ -68630,6 +68679,26 @@ export interface operations { }; }; }; + get_public_whats_new_public_whats_new_get: { + parameters: { + query?: never; + header?: never; + path?: never; + cookie?: never; + }; + requestBody?: never; + responses: { + /** @description Successful Response */ + 200: { + headers: { + [name: string]: unknown; + }; + content: { + "application/json": components["schemas"]["WhatsNewResponse"]; + }; + }; + }; + }; async_queue_request_queue_chat_completions_post: { parameters: { query?: {