media:cpp --transform 1/3

packages/media/cpp/CMakeLists.txt

@@ -145,6 +145,7 @@ add_executable(pm-image
     src/core/glob_paths.cpp
     src/core/output_path.cpp
     src/core/resize.cpp
+    src/core/transform.cpp
     src/core/url_fetch.cpp
     src/http/serve.cpp
     src/ipc/ipc_serve.cpp

packages/media/cpp/src/core/url_fetch.cpp

@@ -10,14 +10,14 @@ namespace fs = std::filesystem;
 
 namespace media {
 
-namespace {
-
-std::once_flag g_curl_init;
+static std::once_flag g_curl_init;
 
 void ensure_curl_global() {
     std::call_once(g_curl_init, []() { curl_global_init(CURL_GLOBAL_DEFAULT); });
 }
 
+namespace {
+
 static size_t write_cb(char *ptr, size_t size, size_t nmemb, void *userdata) {
     auto *os = static_cast<std::ostream *>(userdata);
     const size_t n = size * nmemb;

packages/media/cpp/src/core/url_fetch.hpp

@@ -5,6 +5,9 @@
 
 namespace media {
 
+/// Call curl_global_init once (thread-safe via std::once_flag).
+void ensure_curl_global();
+
 bool is_http_url(const std::string &s);
 
 /** Derive filename (last path segment) for single-file output when input is a URL. */

packages/media/cpp/src/main.cpp

@@ -27,6 +27,7 @@ std::string join_src_semicolons(const std::vector<std::string> &v) {
 
 #include "core/output_path.hpp"
 #include "core/resize.hpp"
+#include "core/transform.hpp"
 #include "http/serve.hpp"
 #include "ipc/ipc_serve.hpp"
 #if defined(_WIN32)
@@ -145,6 +146,23 @@ int main(int argc, char **argv) {
                          "Windows: new Win32++ ribbon UI with drag-drop queue and settings panel");
 #endif
 
+    // ── transform (AI image editing) ──────────────────────────────────
+    std::string tf_input;
+    std::string tf_output;
+    std::string tf_prompt;
+    std::string tf_provider = "google";
+    std::string tf_model    = "gemini-2.0-flash-exp";
+    std::string tf_api_key;
+
+    auto *transform_cmd = app.add_subcommand("transform", "AI image editing (Gemini / Google)");
+    transform_cmd->add_option("input", tf_input, "Input image path")->required(true);
+    transform_cmd->add_option("output", tf_output, "Output path (omit = auto from input + prompt)");
+    transform_cmd->add_option("-p,--prompt", tf_prompt, "Editing prompt")->required(true);
+    transform_cmd->add_option("--provider", tf_provider, "AI provider (google)")->default_val("google");
+    transform_cmd->add_option("--model", tf_model, "Model name")->default_val("gemini-2.0-flash-exp");
+    transform_cmd->add_option("--api-key", tf_api_key, "API key (or set IMAGE_TRANSFORM_GOOGLE_API_KEY env)");
+
+    // ── serve ───────────────────────────────────────────────────────────
     std::string host = "127.0.0.1";
     int port = 8080;
     bool serve_no_cache = false;
@@ -383,6 +401,33 @@ int main(int argc, char **argv) {
         return 0;
     }
 
+    if (transform_cmd->parsed()) {
+        media::TransformOptions topts;
+        topts.provider = tf_provider;
+        topts.model    = tf_model;
+        topts.prompt   = tf_prompt;
+        topts.api_key  = tf_api_key;
+
+        if (topts.api_key.empty()) {
+            const char* env_key = std::getenv("IMAGE_TRANSFORM_GOOGLE_API_KEY");
+            if (env_key && env_key[0] != '\0') topts.api_key = env_key;
+        }
+
+        auto progress = [](const std::string& msg) {
+            std::cerr << msg << "\n";
+        };
+
+        auto result = media::transform_image(tf_input, tf_output, topts, progress);
+        if (!result.ok) {
+            std::cerr << "transform error: " << result.error << "\n";
+            return 1;
+        }
+        std::cout << result.output_path << "\n";
+        if (!result.ai_text.empty())
+            std::cerr << "AI: " << result.ai_text << "\n";
+        return 0;
+    }
+
     if (serve_cmd->parsed()) {
         media::CacheServerDefaults cd;
         cd.enabled = !serve_no_cache;