feat: add persistent multi-model workers and orchestration budgets

include/cst_agent.h

@@ -111,6 +111,13 @@ typedef struct {
   char tools_off[1024]; /* ",name,name," tools switched off (/tools off) */
   /* Sub-agents (task tool). */
   int depth;              /* 0 = top-level; sub-agents run at 1 */
+  int orchestrate;        /* coordinator delegates write/shell work */
+  int orchestrate_override; /* CLI: 1 force on, -1 force off */
+  int subagent_steps, subagent_tasks, subagent_total;
+  void *worker_budget;    /* owned by the top-level agent */
+  int (*request_gate)(void *); /* reserve budget before each HTTP attempt */
+  void *request_gate_ud;
+  char cache_namespace[128];
   int read_only;          /* no write / shell tools */
   char tools_only[1024];  /* ",a,b," allowlist; empty = every tool */
   const char *role_prompt; /* role instructions (not owned) */
@@ -267,6 +274,7 @@ typedef struct {
   long http_status;
   char finish_reason[32];
   long prompt_tokens;
+  int cache_reported;
   long completion_tokens;
   long cached_tokens;      /* prompt tokens read from the provider's cache */
   long cache_miss_tokens;  /* prompt tokens billed at the full input price */
@@ -283,6 +291,8 @@ typedef struct {
   const char *api_key;
   const char *model;
   const char *reasoning_effort; /* NULL = not sent */
+  int (*request_gate)(void *);
+  void *request_gate_ud;
   const cst_message_t *msgs;
   int nmsgs;
   cst_llm_on_delta_fn on_delta;
@@ -433,6 +443,7 @@ long cst_agent_last_prompt_tokens(const cst_agent_t *a);
 /* Token use since the session was opened, and of the latest turn. */
 typedef struct {
   int requests;
+  int cache_known_requests;
   long prompt, cached, miss, write, out;
   double cost;      /* in currency; valid when priced */
   int priced;       /* every request had a known price */