Websearch

Package websearch is a web search and a page fetch your agents can call.

websearch

GET /v1/websearch/search

Keyless web meta-search, in the SearXNG JSON envelope.

Answers {query, number_of_results, results:[{url, title, content, engine}]} — the exact /search?format=json contract a SearXNG client decodes, so an agent tool configured against SearXNG reaches this with no change. q is the query and language narrows it; both are read from the QUERY STRING.

Served in-process by a Go meta-search over keyless public engines, never a third-party search API and never a search key. The enabled engines run concurrently and their hits are merged, deduplicated by normalised URL (host and path, trailing slash and fragment dropped, query kept — distinct queries are distinct results) and capped at 30. Ranking is deterministic rather than scored: the first configured engine's hits lead.

TWO WAYS IN, one-way equivalent, and no third: a validated principal — the same gate the whole data plane uses — passes straight through, and a caller without one must present the shared service key as X-API-Key, compared in constant time. A deployment with no key configured answers 503 rather than opening the surface to everyone, and a missing or wrong key is 401. It is never an open proxy. There is no tenant scoping beyond that gate, and there is nothing to scope: the results are public web pages, identical for every caller.

It fails SOFT on the engines and closed only on the gate. An engine that errors or is served a bot-challenge page contributes zero results and never fails the request, so an empty results is a real answer — nothing was found — and not an outage. The array is always present, never null.

The one thing to get right: every method answers identically. This is one handler registered for all of them, and it reads only the query string, so a body sent on the write verbs is ignored rather than refused.

POST /v1/websearch/search

Keyless web meta-search, in the SearXNG JSON envelope.

Answers {query, number_of_results, results:[{url, title, content, engine}]} — the exact /search?format=json contract a SearXNG client decodes, so an agent tool configured against SearXNG reaches this with no change. q is the query and language narrows it; both are read from the QUERY STRING.

Served in-process by a Go meta-search over keyless public engines, never a third-party search API and never a search key. The enabled engines run concurrently and their hits are merged, deduplicated by normalised URL (host and path, trailing slash and fragment dropped, query kept — distinct queries are distinct results) and capped at 30. Ranking is deterministic rather than scored: the first configured engine's hits lead.

TWO WAYS IN, one-way equivalent, and no third: a validated principal — the same gate the whole data plane uses — passes straight through, and a caller without one must present the shared service key as X-API-Key, compared in constant time. A deployment with no key configured answers 503 rather than opening the surface to everyone, and a missing or wrong key is 401. It is never an open proxy. There is no tenant scoping beyond that gate, and there is nothing to scope: the results are public web pages, identical for every caller.

It fails SOFT on the engines and closed only on the gate. An engine that errors or is served a bot-challenge page contributes zero results and never fails the request, so an empty results is a real answer — nothing was found — and not an outage. The array is always present, never null.

The one thing to get right: every method answers identically. This is one handler registered for all of them, and it reads only the query string, so a body sent on the write verbs is ignored rather than refused.

PUT /v1/websearch/search

Keyless web meta-search, in the SearXNG JSON envelope.

Answers {query, number_of_results, results:[{url, title, content, engine}]} — the exact /search?format=json contract a SearXNG client decodes, so an agent tool configured against SearXNG reaches this with no change. q is the query and language narrows it; both are read from the QUERY STRING.

Served in-process by a Go meta-search over keyless public engines, never a third-party search API and never a search key. The enabled engines run concurrently and their hits are merged, deduplicated by normalised URL (host and path, trailing slash and fragment dropped, query kept — distinct queries are distinct results) and capped at 30. Ranking is deterministic rather than scored: the first configured engine's hits lead.

TWO WAYS IN, one-way equivalent, and no third: a validated principal — the same gate the whole data plane uses — passes straight through, and a caller without one must present the shared service key as X-API-Key, compared in constant time. A deployment with no key configured answers 503 rather than opening the surface to everyone, and a missing or wrong key is 401. It is never an open proxy. There is no tenant scoping beyond that gate, and there is nothing to scope: the results are public web pages, identical for every caller.

It fails SOFT on the engines and closed only on the gate. An engine that errors or is served a bot-challenge page contributes zero results and never fails the request, so an empty results is a real answer — nothing was found — and not an outage. The array is always present, never null.

The one thing to get right: every method answers identically. This is one handler registered for all of them, and it reads only the query string, so a body sent on the write verbs is ignored rather than refused.

PATCH /v1/websearch/search

Keyless web meta-search, in the SearXNG JSON envelope.

Answers {query, number_of_results, results:[{url, title, content, engine}]} — the exact /search?format=json contract a SearXNG client decodes, so an agent tool configured against SearXNG reaches this with no change. q is the query and language narrows it; both are read from the QUERY STRING.

Served in-process by a Go meta-search over keyless public engines, never a third-party search API and never a search key. The enabled engines run concurrently and their hits are merged, deduplicated by normalised URL (host and path, trailing slash and fragment dropped, query kept — distinct queries are distinct results) and capped at 30. Ranking is deterministic rather than scored: the first configured engine's hits lead.

TWO WAYS IN, one-way equivalent, and no third: a validated principal — the same gate the whole data plane uses — passes straight through, and a caller without one must present the shared service key as X-API-Key, compared in constant time. A deployment with no key configured answers 503 rather than opening the surface to everyone, and a missing or wrong key is 401. It is never an open proxy. There is no tenant scoping beyond that gate, and there is nothing to scope: the results are public web pages, identical for every caller.

It fails SOFT on the engines and closed only on the gate. An engine that errors or is served a bot-challenge page contributes zero results and never fails the request, so an empty results is a real answer — nothing was found — and not an outage. The array is always present, never null.

The one thing to get right: every method answers identically. This is one handler registered for all of them, and it reads only the query string, so a body sent on the write verbs is ignored rather than refused.

DELETE /v1/websearch/search

Keyless web meta-search, in the SearXNG JSON envelope.

Answers {query, number_of_results, results:[{url, title, content, engine}]} — the exact /search?format=json contract a SearXNG client decodes, so an agent tool configured against SearXNG reaches this with no change. q is the query and language narrows it; both are read from the QUERY STRING.

Served in-process by a Go meta-search over keyless public engines, never a third-party search API and never a search key. The enabled engines run concurrently and their hits are merged, deduplicated by normalised URL (host and path, trailing slash and fragment dropped, query kept — distinct queries are distinct results) and capped at 30. Ranking is deterministic rather than scored: the first configured engine's hits lead.

TWO WAYS IN, one-way equivalent, and no third: a validated principal — the same gate the whole data plane uses — passes straight through, and a caller without one must present the shared service key as X-API-Key, compared in constant time. A deployment with no key configured answers 503 rather than opening the surface to everyone, and a missing or wrong key is 401. It is never an open proxy. There is no tenant scoping beyond that gate, and there is nothing to scope: the results are public web pages, identical for every caller.

It fails SOFT on the engines and closed only on the gate. An engine that errors or is served a bot-challenge page contributes zero results and never fails the request, so an empty results is a real answer — nothing was found — and not an outage. The array is always present, never null.

The one thing to get right: every method answers identically. This is one handler registered for all of them, and it reads only the query string, so a body sent on the write verbs is ignored rather than refused.

---

All Hanzo APIs · Interactive reference