37 tok/s from Qwen3-Next-80B (48 GB) on a 32 GB laptop, 24 without drafts vs 2.8 for llama.cpp; reading 62 vs 16 tok/s. Faster than llama.cpp on all 13 models compared in Q4_K_M. LLMs bigger than your RAM, pure C, CPU or GPU.
C avx2ccpu-inferenceggufinference-enginelinuxllmlocal-llmmixture-of-expertsmoenvmequantizationqwen3simdspeculative-decodingvulkan
/* SPDX-License-Identifier: GPL-3.0-or-later
Copyright (C) 2026 Maurizio Cammalleri */
/*
* web_search.h - the search behind the Responses API's web_search tool
* (web_search.c): Wikipedia, in the language asked.
*/
#ifndef JANAS_SERVER_WEB_SEARCH_H
#define JANAS_SERVER_WEB_SEARCH_H
#include <stddef.h>
struct srv_web_hit {
char title[256];
char url[512];
char text[2048]; /* the page's first sentences */
int index; /* its rank */
};
/* The pages of Wikipedia (lang: "en", "it"...; NULL or not a code: "en")
that match query, best first, at most max: how many, or -1 with the
reason in err. */
int srv_web_search(const char *query, const char *lang,
struct srv_web_hit *hits, int max, char *err,
size_t err_len);
#endif