[TEST] Add test case for lxm service internal API

- Added positive and negative TCs for lxm service internal API

Signed-off-by: hyunil park <[email protected]>

Author: songgot

tests/capi/meson.build

@@ -76,6 +76,22 @@ if get_option('enable-ml-service')
       test('unittest_capi_service_training_offloading', unittest_capi_service_training_offloading, env: testenv, timeout: 100)
     endif
   endif
+
+  # LXM Service Tests
+  # These tests require both ml-service and llamacpp to be enabled.
+  llamacpp_dep = dependency('llama', required: false)
+  if llamacpp_dep.found()
+    # Note: The source file itself is also conditionally compiled with ENABLE_LLAMACPP.
+    unittest_capi_lxm_service = executable('unittest_capi_lxm_service',
+      'unittest_capi_lxm_service.cc',
+      dependencies: service_unittest_deps,
+      install: get_option('install-test'),
+      install_dir: unittest_install_dir
+    )
+    test('unittest_capi_lxm_service', unittest_capi_lxm_service, env: testenv, timeout: 120) # Increased timeout for LLM response
+  else
+    message('LXM Service tests will be skipped because llama dependency was not found.')
+  endif
 endif
 
 if nnfw_dep.found()

tests/capi/unittest_capi_lxm_service.cc

@@ -0,0 +1,312 @@
+/* SPDX-License-Identifier: Apache-2.0 */
+/**
+ * @file        unittest_capi_lxm_service.cc
+ * @date        26 JULY 2025
+ * @brief       Unit test for ml-lxm-service.
+ * @see         https://github.com/nnstreamer/api
+ * @author      Hyunil Park <[email protected]>
+ * @bug         No known bugs
+ */
+
+#include <gtest/gtest.h>
+#include <glib.h>
+#include <ml-api-common.h>
+#include <ml-api-service-private.h>
+#include <ml-api-service.h>
+#include <string.h>
+#include "ml-lxm-service-internal.h"
+#include "unittest_util.h"
+
+#if defined(ENABLE_LLAMACPP)
+
+/**
+ * @brief Internal function to get the model file path.
+ */
+static gchar *
+_get_model_path (const gchar *model_name)
+{
+  const gchar *root_path = g_getenv ("MLAPI_SOURCE_ROOT_PATH");
+
+  /* Supposed to run test in build directory. */
+  if (root_path == NULL)
+    root_path = "..";
+
+  gchar *model_file = g_build_filename (
+      root_path, "tests", "test_models", "models", model_name, NULL);
+
+  return model_file;
+}
+
+/**
+ * @brief Macro to skip testcase if required files are not ready.
+ */
+#define skip_lxm_tc(tc_name)                                                                       \
+  do {                                                                                             \
+    g_autofree gchar *model_file = _get_model_path ("llama-2-7b-chat.Q2_K.gguf");                  \
+    if (!g_file_test (model_file, G_FILE_TEST_EXISTS)) {                                           \
+      g_autofree gchar *msg = g_strdup_printf (                                                    \
+          "Skipping '%s' due to missing model file. "                                              \
+          "Please download model file from https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF.", \
+          tc_name);                                                                                \
+      GTEST_SKIP () << msg;                                                                        \
+    }                                                                                              \
+  } while (0)
+
+/**
+ * @brief Test data structure to pass to the callback.
+ */
+typedef struct {
+  int token_count;
+  GString *received_tokens;
+} lxm_test_data_s;
+
+/**
+ * @brief Callback function for LXM service token streaming.
+ */
+static void
+_lxm_token_cb (ml_service_event_e event, ml_information_h event_data, void *user_data)
+{
+  lxm_test_data_s *tdata = (lxm_test_data_s *) user_data;
+  ml_tensors_data_h data = NULL;
+  void *_raw = NULL;
+  size_t _size = 0;
+  int status;
+
+  switch (event) {
+    case ML_SERVICE_EVENT_NEW_DATA:
+      ASSERT_TRUE (event_data != NULL);
+
+      status = ml_information_get (event_data, "data", &data);
+      EXPECT_EQ (status, ML_ERROR_NONE);
+      if (status != ML_ERROR_NONE)
+        return;
+
+      status = ml_tensors_data_get_tensor_data (data, 0U, &_raw, &_size);
+      EXPECT_EQ (status, ML_ERROR_NONE);
+      if (status != ML_ERROR_NONE)
+        return;
+
+      if (tdata) {
+        if (tdata->received_tokens) {
+          g_string_append_len (tdata->received_tokens, (const char *) _raw, _size);
+        }
+        tdata->token_count++;
+      }
+      g_print ("%.*s", (int) _size, (char *) _raw); // Print received token
+      break;
+    default:
+      // Handle unknown or unimplemented events if necessary
+      g_printerr ("Received unhandled LXM service event: %d\n", event);
+      break;
+  }
+}
+
+/**
+ * @brief Internal function to run a full LXM session test.
+ */
+static void
+_run_lxm_session_test (const gchar *config_path, const gchar *input_text, ml_option_h options)
+{
+  ml_lxm_session_h session = NULL;
+  ml_lxm_prompt_h prompt = NULL;
+  lxm_test_data_s tdata = { 0, NULL };
+  int status;
+
+  tdata.received_tokens = g_string_new ("");
+
+  // 1. Create session with callback
+  status = ml_lxm_session_create (config_path, NULL, _lxm_token_cb, &tdata, &session);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+  ASSERT_TRUE (session != NULL);
+
+  // 2. Create prompt
+  status = ml_lxm_prompt_create (&prompt);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+  ASSERT_TRUE (prompt != NULL);
+
+  status = ml_lxm_prompt_append_text (prompt, input_text);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+
+  // 3. Generate response (callback is already set during session creation)
+  status = ml_lxm_session_respond (session, prompt, options);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+
+  // Wait for the callback to receive data.
+  // 10 seconds should be enough for a simple response.
+  g_usleep (10000000U);
+
+  // 4. Verify results
+  EXPECT_GT (tdata.token_count, 0);
+  EXPECT_GT (tdata.received_tokens->len, 0U);
+
+  g_print ("\nReceived total tokens: %d\n", tdata.token_count);
+  g_print ("Full received text: %s\n", tdata.received_tokens->str);
+
+  // 5. Cleanup
+  status = ml_lxm_prompt_destroy (prompt);
+  EXPECT_EQ (status, ML_ERROR_NONE);
+
+  status = ml_lxm_session_destroy (session);
+  EXPECT_EQ (status, ML_ERROR_NONE);
+
+  if (tdata.received_tokens) {
+    g_string_free (tdata.received_tokens, TRUE);
+  }
+}
+
+/**
+ * @brief Test basic flow of LXM service.
+ */
+TEST (MLLxmService, basicFlow_p)
+{
+  skip_lxm_tc ("basicFlow_p");
+
+  g_autofree gchar *config = get_config_path ("config_single_llamacpp.conf");
+  ASSERT_TRUE (config != NULL);
+
+  const gchar input_text[] = "Hello LXM, how are you?";
+  ml_option_h options = NULL;
+  int status;
+
+  // Create options
+  status = ml_option_create (&options);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+  ASSERT_TRUE (options != NULL);
+
+  // Set temperature option
+  status = ml_option_set (options, "temperature", g_strdup_printf ("%f", 0.8), g_free);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+
+  // Set max_tokens option
+  status = ml_option_set (
+      options, "max_tokens", g_strdup_printf ("%zu", (size_t) 32), g_free);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+
+  _run_lxm_session_test (config, input_text, options);
+
+  // Cleanup options
+  ml_option_destroy (options);
+}
+
+
+/**
+ * @brief Test LXM service with invalid parameters.
+ */
+TEST (MLLxmService, invalidParams_n)
+{
+  ml_lxm_session_h session = NULL;
+  ml_lxm_prompt_h prompt = NULL;
+  int status;
+  ml_option_h options = NULL;
+  g_autofree gchar *valid_config = get_config_path ("config_single_llamacpp.conf");
+
+  // Create options for testing
+  status = ml_option_create (&options);
+  ASSERT_EQ (status, ML_ERROR_NONE);
+  ml_option_set (options, "temperature", g_strdup_printf ("%f", 0.5), g_free);
+  ml_option_set (options, "max_tokens", g_strdup_printf ("%zu", (size_t) 10), g_free);
+
+  // ml_lxm_session_create
+  status = ml_lxm_session_create (valid_config, NULL, NULL, NULL, NULL);
+  EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+  status = ml_lxm_session_create (NULL, NULL, NULL, NULL, &session);
+  EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+  status = ml_lxm_session_create ("non_existent_config.conf", NULL, NULL, NULL, &session);
+  EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+  status = ml_lxm_session_create (valid_config, NULL, NULL, NULL, &session);
+  EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+  status = ml_lxm_session_create (valid_config, NULL, _lxm_token_cb, NULL, &session);
+  if (status == ML_ERROR_NONE) {
+    // ml_lxm_prompt_create
+    status = ml_lxm_prompt_create (NULL);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+    status = ml_lxm_prompt_create (&prompt);
+    ASSERT_EQ (status, ML_ERROR_NONE);
+
+    // ml_lxm_prompt_append_text
+    status = ml_lxm_prompt_append_text (NULL, "text");
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+    status = ml_lxm_prompt_append_text (prompt, NULL);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+    // ml_lxm_prompt_append_instruction
+    status = ml_lxm_prompt_append_instruction (NULL, "instruction");
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+    status = ml_lxm_prompt_append_instruction (prompt, NULL);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+    // ml_lxm_session_set_instructions
+    status = ml_lxm_session_set_instructions (NULL, "new instructions");
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+    status = ml_lxm_session_set_instructions (session, NULL);
+    EXPECT_EQ (status, ML_ERROR_NONE);
+    status = ml_lxm_session_set_instructions (session, "new instructions");
+    EXPECT_EQ (status, ML_ERROR_NONE);
+
+    // ml_lxm_session_respond - callback is already set during session creation
+    status = ml_lxm_session_respond (NULL, prompt, options);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+    status = ml_lxm_session_respond (session, NULL, options);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+
+    // Now ml_lxm_session_respond should succeed with valid parameters
+    status = ml_lxm_session_respond (session, prompt, options);
+    EXPECT_EQ (status, ML_ERROR_NONE);
+
+    // ml_lxm_prompt_destroy
+    status = ml_lxm_prompt_destroy (NULL);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+    status = ml_lxm_prompt_destroy (prompt);
+    EXPECT_EQ (status, ML_ERROR_NONE);
+    prompt = NULL;
+
+    // ml_lxm_session_destroy
+    status = ml_lxm_session_destroy (NULL);
+    EXPECT_EQ (status, ML_ERROR_INVALID_PARAMETER);
+    status = ml_lxm_session_destroy (session);
+    EXPECT_EQ (status, ML_ERROR_NONE);
+    session = NULL;
+  } else {
+    g_print ("Skipping part of invalidParams_n as session creation failed (possibly due to missing models/config).\n");
+  }
+
+  // Cleanup options
+  ml_option_destroy (options);
+}
+
+/**
+ * @brief Main function to run the test.
+ */
+int
+main (int argc, char **argv)
+{
+  int result = -1;
+
+  try {
+    testing::InitGoogleTest (&argc, argv);
+  } catch (...) {
+    g_warning ("catch 'testing::internal::<unnamed>::ClassUniqueToAlwaysTrue'");
+  }
+
+  /* ignore tizen feature status while running the testcases */
+  set_feature_state (ML_FEATURE, SUPPORTED);
+  set_feature_state (ML_FEATURE_INFERENCE, SUPPORTED);
+  set_feature_state (ML_FEATURE_SERVICE, SUPPORTED);
+
+  try {
+    result = RUN_ALL_TESTS ();
+  } catch (...) {
+    g_warning ("catch `testing::internal::GoogleTestFailureException`");
+  }
+
+  set_feature_state (ML_FEATURE, NOT_CHECKED_YET);
+  set_feature_state (ML_FEATURE_INFERENCE, NOT_CHECKED_YET);
+  set_feature_state (ML_FEATURE_SERVICE, NOT_CHECKED_YET);
+
+  return result;
+}
+#endif /* ENABLE_LLAMACPP */