| 12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879 |
- """Exa web search + content extraction via the ``exa-py`` SDK (lazy-installed).
- Env: ``EXA_API_KEY`` (https://exa.ai). Both methods are sync — Exa's SDK is
- sync-only; the dispatcher threads extract when the caller is async.
- """
- from __future__ import annotations
- import logging
- from typing import Any, Dict, List
- from plugins.web._common import (
- BaseWebSearchProvider, cached_sdk_client, document, exa_crawl_status, keyless_extract, keyless_search, keyless_variant_schema,
- provider_env, run_extract, run_search, search_ok, use_keyless, vendor_error_entry, web_hit,
- )
- logger = logging.getLogger(__name__)
- _MISSING_KEY = "EXA_API_KEY environment variable not set. Get your API key at https://exa.ai"
- def _get_exa_client() -> Any:
- def _factory(api_key: str) -> Any:
- from exa_py import Exa # deliberately lazy
- client = Exa(api_key=api_key)
- client.headers["x-exa-integration"] = "hermes-agent"
- return client
- return cached_sdk_client("_exa_client", "EXA_API_KEY", _MISSING_KEY, "exa", _factory)
- def _crawl_status(error: Any) -> Any:
- """The origin status of a failed ``statuses[]`` entry: ``http_status_code``, else the one its tag names."""
- return getattr(error, "http_status_code", None) or exa_crawl_status(getattr(error, "tag", None))
- class ExaWebSearchProvider(BaseWebSearchProvider):
- """Exa search + extract provider."""
- NAME = "exa"
- DISPLAY_NAME = "Exa"
- KEY_ENV = "EXA_API_KEY"
- EXTRACT = True
- KEYLESS = True
- def search(self, query: str, limit: int = 5) -> dict[str, Any]:
- def _body() -> dict[str, Any]:
- if use_keyless("exa", provider_env("EXA_API_KEY")):
- return keyless_search("Exa", "exa", query, limit, logger)
- logger.info("Exa search: '%s' (limit=%d)", query, limit)
- response = _get_exa_client().search(query, num_results=limit, contents={"highlights": True})
- return search_ok([
- web_hit(r.url or "", r.title or "", " ".join(r.highlights or []), i + 1)
- for i, r in enumerate(response.results or [])
- ])
- return run_search("Exa", logger, _body, sdk=True)
- def extract(self, urls: list[str], **kwargs: Any) -> list[dict[str, Any]]:
- def _body() -> list[dict[str, Any]]:
- if use_keyless("exa", provider_env("EXA_API_KEY")):
- return keyless_extract("Exa", "exa", urls, logger)
- logger.info("Exa extract: %d URL(s)", len(urls))
- response = _get_exa_client().get_contents(urls, text=True)
- results = [document(r.url or "", r.title or "", r.text or "") for r in response.results or []]
- # A URL Exa could not fetch is absent from ``results``; its reason is in ``statuses``.
- return results + [
- vendor_error_entry(s.id or "", status=_crawl_status(s.error), reason=getattr(s.error, "tag", None))
- for s in getattr(response, "statuses", None) or [] if s.status == "error"
- ]
- return run_extract("Exa", logger, urls, _body, sdk=True)
- def get_setup_schema(self) -> dict[str, Any]:
- return keyless_variant_schema(
- "Exa", "EXA_API_KEY", "https://exa.ai",
- free_tag="Semantic + neural web search with content extraction on Exa's anonymous free tier. Rate-limited under burst load.",
- paid_tag="Semantic + neural web search with content extraction via the Exa SDK. Unthrottled, guaranteed service.",
- )
|