feat(search): add Keenable as a search provider

litellm/llms/keenable/__init__.py

@@ -0,0 +1,3 @@
+from litellm.llms.keenable.search.transformation import KeenableSearchConfig
+
+__all__ = ("KeenableSearchConfig",)

litellm/llms/keenable/search/__init__.py

@@ -0,0 +1,3 @@
+from litellm.llms.keenable.search.transformation import KeenableSearchConfig
+
+__all__ = ("KeenableSearchConfig",)

litellm/llms/keenable/search/transformation.py

@@ -0,0 +1,270 @@
+"""
+Calls Keenable's /v1/search endpoint to search the web.
+
+Keenable works without an API key: keyless requests go to /v1/search/public, which is rate
+limited per IP and identifies the calling app by the X-Keenable-Title header. With an API key
+the same body goes to /v1/search, which has higher limits.
+
+Keenable API Reference: https://docs.keenable.ai/api-reference/search
+"""
+
+from __future__ import annotations
+
+from collections.abc import Mapping
+from types import MappingProxyType
+from typing import TYPE_CHECKING, Final
+
+import httpx
+from pydantic import ConfigDict, TypeAdapter, ValidationError
+
+from litellm.llms.base_llm.chat.transformation import BaseLLMException
+from litellm.llms.base_llm.search.transformation import (
+    BaseSearchConfig,
+    SearchResponse,
+    SearchResult,
+)
+from litellm.types.llms.base import LiteLLMBaseModel
+
+if TYPE_CHECKING:
+    from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+
+_KEENABLE_DOCS_URL: Final = "https://docs.keenable.ai/api-reference/search"
+
+# Sent on every request. The keyless endpoint rejects a request without it (400), and it
+# names the calling software only, never the user.
+_APP_TITLE: Final = "litellm"
+
+
+class _KeenableResult(LiteLLMBaseModel):
+    """One entry of Keenable's `results` array. Every field is optional so a single degraded
+    result degrades to empty strings instead of failing the whole call."""
+
+    model_config = ConfigDict(extra="ignore", frozen=True)
+
+    title: str | None = None
+    url: str | None = None
+    # Usually empty. The page text is in `snippet`.
+    description: str | None = None
+    snippet: str | None = None
+    published_at: str | None = None
+    acquired_at: str | None = None
+
+
+class _KeenableSearchResponse(LiteLLMBaseModel):
+    """Keenable's /v1/search response envelope."""
+
+    model_config = ConfigDict(extra="ignore", frozen=True)
+
+    # Required: a search with no hits returns `[]`, so a null or absent `results` means the
+    # body is not a search response and must not be reported as a successful empty search.
+    results: tuple[_KeenableResult, ...]
+
+
+class _ErrorEnvelope(LiteLLMBaseModel):
+    """Keenable reports errors as `{"error": "Invalid parameter", "message": "..."}`."""
+
+    model_config = ConfigDict(extra="ignore", frozen=True)
+
+    error: str | None = None
+    message: str | None = None
+
+
+_DomainListAdapter: Final = TypeAdapter(tuple[str, ...])
+
+_NOTHING: Final[Mapping[str, object]] = MappingProxyType({})
+
+
+def _optional(key: str, value: object) -> Mapping[str, object]:
+    """A one-entry mapping to spread into a payload, or nothing when the value is absent."""
+    return MappingProxyType({key: value}) if value is not None else _NOTHING
+
+
+class KeenableSearchConfig(BaseSearchConfig):
+    KEENABLE_API_BASE = "https://api.keenable.ai/v1"
+
+    @staticmethod
+    def ui_friendly_name() -> str:
+        return "Keenable"
+
+    def _resolve_api_key(self, api_key: str | None, api_base: str | None) -> str | None:
+        """The caller's key, else KEENABLE_API_KEY, else None for a keyless call."""
+        return self.resolve_server_api_key(
+            caller_api_key=api_key,
+            caller_api_base=api_base,
+            key_env_vars=("KEENABLE_API_KEY",),
+            base_env_var=None,
+            default_api_base=self.KEENABLE_API_BASE,
+        )
+
+    def validate_environment(
+        self,
+        headers: dict[str, str],  # mutable-ok: BaseSearchConfig.validate_environment signature
+        api_key: str | None = None,
+        api_base: str | None = None,
+        **kwargs: object,  # kwargs-ok: BaseSearchConfig.validate_environment signature
+    ) -> dict[str, str]:  # mutable-ok: the http handler passes this straight to httpx as headers
+        """
+        Validate environment and return headers. No API key is required.
+
+        Returns a new dict rather than mutating ``headers``: the http handler calls this
+        a second time after ``litellm/search/main.py`` already did, so it has to be idempotent.
+        """
+        resolved_api_key: Final = self._resolve_api_key(api_key, api_base)
+        return {
+            **headers,
+            "Content-Type": "application/json",
+            "X-Keenable-Title": _APP_TITLE,
+            **_optional("Authorization", f"Bearer {resolved_api_key}" if resolved_api_key else None),
+        }
+
+    def get_complete_url(
+        self,
+        api_base: str | None,
+        optional_params: dict[str, object],  # mutable-ok: BaseSearchConfig.get_complete_url signature
+        data: dict[str, object] | list[dict[str, object]] | None = None,  # mutable-ok: base signature
+        **kwargs: object,  # kwargs-ok: BaseSearchConfig.get_complete_url signature
+    ) -> str:
+        """
+        The keyed and keyless endpoints take the same body and differ only by path, so the key
+        decides it: /search with a key, /search/public without. ``api_base`` is the version
+        root (https://api.keenable.ai/v1); a base that already ends in either path is accepted.
+        """
+        caller_api_key: Final = kwargs.get("api_key")
+        resolved_api_key: Final = self._resolve_api_key(
+            caller_api_key if isinstance(caller_api_key, str) else None, api_base
+        )
+        root: Final = (
+            (api_base or self.KEENABLE_API_BASE).rstrip("/").removesuffix("/search/public").removesuffix("/search")
+        )
+        return f"{root}/search" if resolved_api_key else f"{root}/search/public"
+
+    def transform_search_request(
+        self,
+        query: str | list[str],  # mutable-ok: BaseSearchConfig.transform_search_request signature
+        optional_params: dict[str, object],  # mutable-ok: base signature
+        **kwargs: object,  # kwargs-ok: BaseSearchConfig.transform_search_request signature
+    ) -> dict[str, object]:  # mutable-ok: the http handler passes this straight to httpx as the JSON body
+        """
+        Transform Search request to Keenable API format.
+
+        - query -> query (a list is joined with spaces; Keenable takes a single string)
+        - max_results -> max_results (sent unclamped so Keenable's own 1-50 validation reports the error)
+        - search_domain_filter -> `site` for a single domain, `site:` / `-site:` clauses in the query
+          otherwise, following the Perplexity unified spec where a `-` prefix excludes a domain
+        - country, max_tokens_per_page -> dropped (no Keenable equivalent)
+
+        Everything else is forwarded as-is, so the rest of Keenable's surface (date filters,
+        snippet_max_length) stays reachable without LiteLLM tracking it.
+        """
+        unified_params: Final = self.get_supported_perplexity_optional_params()
+        include, exclude = _split_domains(optional_params.get("search_domain_filter"))
+        site: Final = include[0] if len(include) == 1 else None
+
+        # Spread after the derived `site` so an explicitly supplied one wins.
+        passthrough: Final = MappingProxyType(
+            {param: value for param, value in optional_params.items() if param not in unified_params}
+        )
+
+        return {
+            **_optional("site", site),
+            **passthrough,
+            "query": _query_with_domain_clauses(
+                " ".join(query) if isinstance(query, list) else query,
+                include=() if site else include,
+                exclude=exclude,
+            ),
+            **_optional("max_results", optional_params.get("max_results")),
+        }
+
+    def transform_search_response(
+        self,
+        raw_response: httpx.Response,
+        logging_obj: LiteLLMLoggingObj,
+        **kwargs: object,  # kwargs-ok: BaseSearchConfig.transform_search_response signature
+    ) -> SearchResponse:
+        """
+        Transform Keenable API response to LiteLLM unified SearchResponse format.
+
+        `snippet` carries the page text and `description` is usually empty, so the snippet comes
+        first. `date` is the page's `published_at` and `last_updated` is `acquired_at`, when
+        Keenable last fetched the page. A body that does not match the documented schema raises
+        an attributed error rather than being reported as a successful empty search.
+        """
+        try:
+            parsed: Final = _KeenableSearchResponse.model_validate_json(raw_response.content)
+        except ValidationError as e:
+            raise self.get_error_class(
+                error_message=f"response does not match the documented /v1/search schema: {e}",
+                status_code=raw_response.status_code,
+                headers=dict(raw_response.headers),
+            )
+
+        return SearchResponse(
+            results=[
+                SearchResult(
+                    title=result.title or "",
+                    url=result.url or "",
+                    snippet=result.snippet or result.description or "",
+                    date=result.published_at,
+                    last_updated=result.acquired_at,
+                )
+                for result in parsed.results
+            ],
+            object="search",
+        )
+
+    def get_error_class(
+        self,
+        error_message: str,
+        status_code: int,
+        headers: dict[str, str],  # mutable-ok: BaseSearchConfig.get_error_class signature
+    ) -> Exception:
+        detail: Final = _unwrap_error_detail(error_message).rstrip(". ")
+        # Keyless calls share one per-IP budget, which a proxy serving many users exhausts first.
+        hint: Final = (
+            " Without an API key, requests share a per-IP limit; set KEENABLE_API_KEY to raise it."
+            if status_code == 429
+            else ""
+        )
+        return BaseLLMException(
+            status_code=status_code,
+            message=f"Keenable Search: {detail}.{hint} See {_KEENABLE_DOCS_URL} for details.",
+            headers=headers,
+        )
+
+
+def _unwrap_error_detail(error_message: str) -> str:
+    """
+    Surface the human-readable message inside Keenable's error envelope.
+
+    Falls back to the raw body for anything else (CDN HTML pages, plain text, other shapes).
+    """
+    try:
+        body: Final = _ErrorEnvelope.model_validate_json(error_message)
+    except ValidationError:
+        return error_message
+    return body.message or body.error or error_message
+
+
+def _split_domains(search_domain_filter: object) -> tuple[tuple[str, ...], tuple[str, ...]]:
+    """
+    Split the unified `search_domain_filter` into included and excluded domains.
+
+    Anything that is not a list of strings is ignored rather than raising, since it only
+    ever narrows a search that is otherwise valid.
+    """
+    try:
+        domains: Final = _DomainListAdapter.validate_python(search_domain_filter)
+    except ValidationError:
+        return (), ()
+    return (
+        tuple(d for d in domains if d and not d.startswith("-")),
+        tuple(d[1:] for d in domains if d.startswith("-") and len(d) > 1),
+    )
+
+
+def _query_with_domain_clauses(query: str, include: tuple[str, ...], exclude: tuple[str, ...]) -> str:
+    """Append `(site:a OR site:b)` and `-site:c` clauses, which Keenable honors in the query."""
+    included: Final = (f"({' OR '.join(f'site:{d}' for d in include)})",) if include else ()
+    excluded: Final = tuple(f"-site:{d}" for d in exclude)
+    return " ".join((query, *included, *excluded))

litellm/model_prices_and_context_window_backup.json

@@ -22865,6 +22865,14 @@
             "notes": "Nimble Search API pay-as-you-go list price: $5 per 1,000 searches, up to 100 results per search. Volume plans price differently."
         }
     },
+    "keenable/search": {
+        "input_cost_per_query": 0.004,
+        "litellm_provider": "keenable",
+        "mode": "search",
+        "metadata": {
+            "notes": "Keenable pay-as-you-go list price: $4 per 1,000 requests, after a free allowance of 100,000 requests a month per organization. Keyless requests to /v1/search/public are free and rate limited per IP."
+        }
+    },
     "elevenlabs/scribe_v1": {
         "input_cost_per_second": 6.11e-05,
         "litellm_provider": "elevenlabs",

litellm/types/utils.py

@@ -4224,6 +4224,7 @@ class SearchProviders(str, Enum):
     DATAFORSEO = "dataforseo"
     FIRECRAWL = "firecrawl"
     FASTCRW = "fastcrw"
+    KEENABLE = "keenable"
     SEARXNG = "searxng"
     LINKUP = "linkup"
     DUCKDUCKGO = "duckduckgo"

litellm/utils.py

@@ -9823,6 +9823,7 @@ def get_provider_search_config(
         from litellm.llms.fastcrw.search.transformation import FastCRWSearchConfig
         from litellm.llms.firecrawl.search.transformation import FirecrawlSearchConfig
         from litellm.llms.google_pse.search.transformation import GooglePSESearchConfig
+        from litellm.llms.keenable.search.transformation import KeenableSearchConfig
         from litellm.llms.linkup.search.transformation import LinkupSearchConfig
         from litellm.llms.nimble.search.transformation import NimbleSearchConfig
         from litellm.llms.parallel_ai.search.transformation import (
@@ -9846,6 +9847,7 @@ def get_provider_search_config(
             SearchProviders.DATAFORSEO: DataForSEOSearchConfig,
             SearchProviders.FIRECRAWL: FirecrawlSearchConfig,
             SearchProviders.FASTCRW: FastCRWSearchConfig,
+            SearchProviders.KEENABLE: KeenableSearchConfig,
             SearchProviders.SEARXNG: SearXNGSearchConfig,
             SearchProviders.LINKUP: LinkupSearchConfig,
             SearchProviders.DUCKDUCKGO: DuckDuckGoSearchConfig,

model_prices_and_context_window.json

@@ -22865,6 +22865,14 @@
             "notes": "Nimble Search API pay-as-you-go list price: $5 per 1,000 searches, up to 100 results per search. Volume plans price differently."
         }
     },
+    "keenable/search": {
+        "input_cost_per_query": 0.004,
+        "litellm_provider": "keenable",
+        "mode": "search",
+        "metadata": {
+            "notes": "Keenable pay-as-you-go list price: $4 per 1,000 requests, after a free allowance of 100,000 requests a month per organization. Keyless requests to /v1/search/public are free and rate limited per IP."
+        }
+    },
     "elevenlabs/scribe_v1": {
         "input_cost_per_second": 6.11e-05,
         "litellm_provider": "elevenlabs",

provider_endpoints_support.json

@@ -2619,6 +2619,13 @@
         "search": true
       }
     },
+    "keenable": {
+      "display_name": "Keenable (`keenable`)",
+      "url": "https://docs.keenable.ai/api-reference/search",
+      "endpoints": {
+        "search": true
+      }
+    },
     "triton": {
       "display_name": "Triton (`triton`)",
       "url": "https://docs.litellm.ai/docs/providers/triton-inference-server",

tests/code_coverage_tests/enforce_llms_folder_style.py

@@ -13,6 +13,7 @@
     "brave",
     "firecrawl",
     "fastcrw",
+    "keenable",
     "searxng",
     "linkup",
     "duckduckgo",