How I do use the recent llama.cpp native tools to do web rag a.k.a. web_fetch (or anything else for the matter) directly from inside the llama-server's webui

In three linesA llama.cpp user implements a secure web RAG workflow by enabling native server tools (exec_shell_command) with multi-sandboxing: firejail + dedicated Linux user + Alpine OCI container. Allows Qwen 3.6-35B model to execute wget commands directly from web UI to fetch and analyze content.

Summary generated by Claude — human-verified