Documentation
¶
Overview ¶
Package serper wires Serper's Google Search API into web.Searcher.
Endpoint ¶
POST https://google.serper.dev/search
Authentication uses the X-API-KEY header (not bearer).
Parameter mapping ¶
web.SearchRequest → Serper request:
- Query → q (after Google site:/-site: rewriting via web.SearchRequest.QueryWithSiteOperators)
- MaxResults → num (forwarded as-is)
- AllowedDomains → inlined into q as `(site:foo.com OR site:bar.com)`
- BlockedDomains → inlined into q as `-site:foo.com -site:bar.com`
- Recency → tbs=qdr:h|d|w|m|y
Serper has no native include/exclude domain fields, so the query is rewritten with Google's site:/-site: operators. The trade-off is that the operator count counts against Google's per-query token budget.
Hardcoded: autocorrect=true.
Response mapping ¶
Serper organic[] → []*web.SearchResult:
- title → Title
- link → URL
- snippet → Snippet
- date → PublishedTime (best-effort parse; relative strings like "2 days ago" become zero time)
Knowledge graph, answer box, people-also-ask, and related searches are ignored — only organic results are surfaced.
Serper's transport DTOs stay private; only organic results cross the provider boundary.
Reference ¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Client ¶
type Client struct {
// contains filtered or unexported fields
}
Client implements web.Searcher against the Serper API. It exposes no vendor-specific surface, so a caller can substitute another search backend without touching the tool that consumes it.
func NewClient ¶
NewClient requires the API key at construction so a missing credential fails at wiring rather than as an authorization error on a model's first search.
func (*Client) Search ¶
func (c *Client) Search(ctx context.Context, request *web.SearchRequest) (*web.SearchResponse, error)
Source Files