Enhance the flexibility of the c streamer (#1940)

This PR is aimed at increasing the flexibility of the c streamer for llm
generation.
In C++, a streamer can add additional functionality through lambda
captures (e.g.
https://github.com/openvinotoolkit/model_server/blob/f513c50d45874392f0c6f356484c89003fb47dcc/src/llm/servable.cpp#L66).
Therefore, an extra `void*` parameter is added to the C streamer to
align with the C++ streamer.

Author: apinge

samples/c/text_generation/chat_sample_c.c

@@ -8,17 +8,16 @@
 #include "openvino/genai/c/llm_pipeline.h"
 
 #define MAX_PROMPT_LENGTH 64
-#define MAX_OUTPUT_LENGTH 1024
 
 #define CHECK_STATUS(return_status)                                                      \
     if (return_status != OK) {                                                           \
         fprintf(stderr, "[ERROR] return status %d, line %d\n", return_status, __LINE__); \
         goto err;                                                                        \
     }
-
-ov_genai_streamming_status_e print_callback(const char* args) {
-    if (args) {
-        fprintf(stdout, "%s", args);
+ov_genai_streamming_status_e print_callback(const char* str, void* args) {
+    if (str) {
+        // If args is not null, it needs to be cast to its actual type.
+        fprintf(stdout, "%s", str);
         fflush(stdout);
         return OV_GENAI_STREAMMING_STATUS_RUNNING;
     } else {
@@ -37,7 +36,8 @@ int main(int argc, char* argv[]) {
 
     ov_genai_generation_config* config = NULL;
     ov_genai_llm_pipeline* pipeline = NULL;
-    stream_callback streamer = &print_callback;
+    streamer_callback streamer;
+    streamer.callback_func = print_callback;
     char prompt[MAX_PROMPT_LENGTH];
 
     CHECK_STATUS(ov_genai_llm_pipeline_create(models_path, device, &pipeline));

src/c/include/openvino/genai/c/llm_pipeline.h

@@ -96,9 +96,17 @@ typedef enum {
 } ov_genai_streamming_status_e;
 
 /**
- * @brief Callback function for streaming output.
+ * @brief Structure for streamer callback functions with arguments.
+ *
+ * The callback function takes two parameters:
+ * - `const char* str`: A constant string extracted from the decoded result for processing
+ * - `void* args`: A pointer to additional arguments, allowing flexible data passing.
  */
-typedef ov_genai_streamming_status_e(OPENVINO_C_API_CALLBACK* stream_callback)(const char*);
+typedef struct {
+    ov_genai_streamming_status_e(
+        OPENVINO_C_API_CALLBACK* callback_func)(const char* str, void* args);  //!< Pointer to the callback function
+    void* args;  //!< Pointer to the arguments passed to the callback function
+} streamer_callback;
 
 /**
  * @brief Generate results by ov_genai_llm_pipeline
@@ -114,7 +122,7 @@ typedef ov_genai_streamming_status_e(OPENVINO_C_API_CALLBACK* stream_callback)(c
 OPENVINO_GENAI_C_EXPORTS ov_status_e ov_genai_llm_pipeline_generate(ov_genai_llm_pipeline* pipe,
                                                                     const char* inputs,
                                                                     const ov_genai_generation_config* config,
-                                                                    const stream_callback* streamer,
+                                                                    const streamer_callback* streamer,
                                                                     ov_genai_decoded_results** results);
 /**
  * @brief Start chat with keeping history in kv cache.

src/c/src/llm_pipeline.cpp

@@ -90,7 +90,7 @@ void ov_genai_llm_pipeline_free(ov_genai_llm_pipeline* pipe) {
 ov_status_e ov_genai_llm_pipeline_generate(ov_genai_llm_pipeline* pipe,
                                            const char* inputs,
                                            const ov_genai_generation_config* config,
-                                           const stream_callback* streamer,
+                                           const streamer_callback* streamer,
                                            ov_genai_decoded_results** results) {
     if (!pipe || !(pipe->object) || !inputs || !(streamer || results)) {
         return ov_status_e::INVALID_C_PARAM;
@@ -102,7 +102,7 @@ ov_status_e ov_genai_llm_pipeline_generate(ov_genai_llm_pipeline* pipe,
         ov::genai::StringInputs input = {input_str};
         if (streamer) {
             auto callback = [streamer](std::string word) -> ov::genai::StreamingStatus {
-                return static_cast<ov::genai::StreamingStatus>((*streamer)(word.c_str()));
+                return static_cast<ov::genai::StreamingStatus>((streamer->callback_func)(word.c_str(), streamer->args));
             };
             *(_results->object) = (config && config->object)
                                       ? pipe->object->generate(input, *(config->object), callback)