Usability pass: tool registry, TUI polish, saved keys and settings

Tools
- Table-driven tool registry with permission classes and default/minimal
  profiles; new tools: edit_file, list_dir, glob, grep, delete_file,
  move_file, todo_write, web_fetch, ask_user (plus read_file paging).
- Built-in Myers diff; write tools return unified diffs.
- Persistent permission rules (.agent/permissions.json), per-turn undo
  snapshots (.agent/undo), AGENTS.md project notes, auto-compact at 80%
  of the model context, cancellable run_shell with process-group kill.

TUI
- Table-driven slash commands: /provider /key /effort /status /tools
  /permissions /yolo /undo /retry /diff /export /copy /rename /delete
  /init /max-steps /todo, with / and @ completion.
- Permission modal with diff and prefix rules, ask_user modal, todo
  panel, header with model/effort/context, bracketed paste, line editing,
  wide-char cursor, NO_COLOR and ASCII fallback; Esc never quits.

Startup and configuration
- API keys are entered in the TUI (hidden) and saved per provider in
  <config>/credentials (0600); env variables are only a fallback.
- Last provider, per-provider model and reasoning effort are remembered
  in <config>/settings.json; --effort/-e and /effort send
  reasoning_effort (low|medium|high|max) to models that accept it.
- DeepSeek preset uses deepseek-flash / deepseek-v4-pro.
- Session flags --new/--continue/--resume/--list-sessions, --yolo,
  --confirm.

Compatibility and tests
- ncurses detection, macOS/Homebrew support, make install, XDG config
  dir, POSIX cleanups; Gitea CI (gcc/clang/asan/musl/macOS).
- Unit tests for diff, fs tools, registry, undo/perms and config;
  e2e fixtures multi_file_refactor and ask_or_todo; vendored cJSON.

src/models.c

@@ -6,6 +6,28 @@
 
 #define MODEL_MAX_LEN 128
 
+static const cst_provider_t PROVIDERS[] = {
+    {"minimax", "https://api.minimax.cn/v1", "MiniMax-M2.7", "MINIMAX_API_KEY"},
+    {"deepseek", "https://api.deepseek.com/v1", "deepseek-flash",
+     "DEEPSEEK_API_KEY"},
+    {"openai", "https://api.openai.com/v1", "gpt-4.1-mini", "OPENAI_API_KEY"},
+};
+
+const cst_provider_t *cst_provider_find(const char *name) {
+  if (!name || !name[0])
+    return NULL;
+  for (size_t i = 0; i < sizeof(PROVIDERS) / sizeof(PROVIDERS[0]); i++) {
+    if (strcmp(PROVIDERS[i].name, name) == 0)
+      return &PROVIDERS[i];
+  }
+  return NULL;
+}
+
+const cst_provider_t *cst_provider_default(void) {
+  const cst_provider_t *p = cst_provider_find(getenv("CST_PROVIDER"));
+  return p ? p : &PROVIDERS[0];
+}
+
 static int already(char out[][MODEL_MAX_LEN], int n, const char *name) {
   for (int i = 0; i < n; i++) {
     if (strcmp(out[i], name) == 0)
@@ -44,7 +66,6 @@ int cst_models_list(const char *current, char out[][128], int max_out) {
         p++;
       char save = *p;
       *p = 0;
-      /* trim trailing spaces */
       char *end = start + strlen(start);
       while (end > start && (end[-1] == ' ' || end[-1] == '\t'))
         *--end = 0;
@@ -54,10 +75,12 @@ int cst_models_list(const char *current, char out[][128], int max_out) {
         p++;
     }
   } else {
+    push(out, &n, max_out, "deepseek-flash");
+    push(out, &n, max_out, "deepseek-v4-pro");
     push(out, &n, max_out, "MiniMax-M2.7");
     push(out, &n, max_out, "MiniMax-M2.5");
+    push(out, &n, max_out, "gpt-4.1-mini");
     push(out, &n, max_out, "gpt-4o");
-    push(out, &n, max_out, "gpt-4o-mini");
   }
 
   if (current && current[0])
@@ -65,3 +88,55 @@ int cst_models_list(const char *current, char out[][128], int max_out) {
 
   return n;
 }
+
+/* Context window in tokens (CST_CTX_TOKENS overrides). Prefix match. */
+long cst_model_context_tokens(const char *model) {
+  const char *e = getenv("CST_CTX_TOKENS");
+  if (e && atol(e) > 0)
+    return atol(e);
+  static const struct {
+    const char *prefix;
+    long tokens;
+  } T[] = {{"deepseek", 128000}, {"MiniMax", 196608}, {"minimax", 196608},
+           {"gpt-4.1", 1047576}, {"gpt-4o", 128000},  {"gpt-5", 400000},
+           {"o3", 200000},       {"o4", 200000},      {"claude", 200000},
+           {"qwen", 131072},     {"glm", 128000},     {"kimi", 131072}};
+  if (model) {
+    for (size_t i = 0; i < sizeof(T) / sizeof(T[0]); i++) {
+      if (strncmp(model, T[i].prefix, strlen(T[i].prefix)) == 0)
+        return T[i].tokens;
+    }
+  }
+  return 128000;
+}
+
+int cst_effort_valid(const char *effort) {
+  static const char *const ok[] = {"low", "medium", "high", "max"};
+  for (size_t i = 0; effort && i < sizeof(ok) / sizeof(ok[0]); i++) {
+    if (strcmp(effort, ok[i]) == 0)
+      return 1;
+  }
+  return 0;
+}
+
+const char *cst_effort_wire(const char *provider, const char *model,
+                            const char *effort) {
+  if (!cst_effort_valid(effort) || !provider || !model)
+    return NULL;
+  if (strcmp(provider, "deepseek") == 0)
+    return effort;
+  /* OpenAI reasoning models take low | medium | high. */
+  if (strcmp(provider, "openai") == 0 &&
+      (strncmp(model, "o1", 2) == 0 || strncmp(model, "o3", 2) == 0 ||
+       strncmp(model, "o4", 2) == 0 || strncmp(model, "gpt-5", 5) == 0))
+    return strcmp(effort, "max") == 0 ? "high" : effort;
+  return NULL;
+}
+
+int cst_provider_count(void) {
+  return (int)(sizeof(PROVIDERS) / sizeof(PROVIDERS[0]));
+}
+
+const cst_provider_t *cst_provider_at(int i) {
+  return i >= 0 && i < cst_provider_count() ? &PROVIDERS[i] : NULL;
+}