working

tools/cli/cli-context.h

@@ -1,9 +1,3 @@
-// controller for llama-cli (the "controller" in MVC)
-//
-// owns the chat state, drives the view and talks to llama-server through
-// cli_client; when no --server-base is given it also manages a local
-// llama-server child process via cli_server
-
 #pragma once
 
 #include "common.h"
@@ -20,25 +14,6 @@ struct cli_timings {
     double predicted_per_second = 0.0;
 };
 
-struct cli_command_info {
-    std::string usage;       // e.g. "/read <file>"
-    std::string description; // e.g. "add a text file"
-};
-
-// properties of the connected server, shown on startup
-struct cli_server_info {
-    std::string build_info;
-    std::string model_name;
-    std::string server_base;
-    bool is_local_server   = false; // server is spawned and managed by llama-cli
-    bool has_system_prompt = false;
-    bool has_vision        = false;
-    bool has_audio         = false;
-    bool has_video         = false;
-
-    std::vector<cli_command_info> commands;
-};
-
 // set by the SIGINT handler; cleared once the interrupt has been handled
 extern std::atomic<bool> g_cli_interrupted;
 
@@ -52,6 +27,7 @@ struct cli_context {
     json pending_media = json::array(); // staged multimodal content parts
 
     // properties of the connected server
+    // will be populated by fetch_server_props()
     std::string model_name;
     std::string build_info;
     bool has_vision = false;