centralizing provider initialization, adding all supported providers,…

… adding additional untested agent modules, migrated to functions for provider init.

app/modules/llm_provider.py

@@ -1,6 +1,13 @@
-import os
 from srt_core.config import Config
 from srt_core.utils.logger import Logger
+from llama_cpp_agent.providers import (
+    LlamaCppServerProvider,
+    VLLMServerProvider,
+    TGIServerProvider,
+    LlamaCppPythonProvider,
+    GroqProvider
+)
+from llama_cpp import Llama
 
 class LLMProvider:
     def __init__(self):
@@ -18,6 +25,8 @@ def load_llm_settings(self, llm_name):
             "agent_provider": llm_config["agent_provider"],
             "server_name": llm_config["server"],
             "max_tokens": llm_config["max_tokens"],
+            "model": llm_config.get("model", ""),
+            "api_key": llm_config.get("api_key", "")
         }
 
     def set_llm_attributes(self, llm_prefix, llm_settings):
@@ -38,49 +47,47 @@ def load_provider_settings(self):
         self.set_llm_attributes('summary', self.summary_llm_settings)
         self.set_llm_attributes('chat', self.chat_llm_settings)
 
-        if "llama_cpp_server" in self.default_llm_settings["agent_provider"]:
-            from llama_cpp_agent.providers import LlamaCppServerProvider
-            self.default_provider = LlamaCppServerProvider(self.default_llm_settings["url"])
-            self.summary_provider = LlamaCppServerProvider(self.summary_llm_settings["url"])
-            self.chat_provider = LlamaCppServerProvider(self.chat_llm_settings["url"])
-        elif "llama_cpp_python" in self.default_llm_settings["agent_provider"]:
-            from llama_cpp import Llama
-            from llama_cpp_agent.providers import LlamaCppPythonProvider
-            python_cpp_llm = Llama(
-                model_path=f"models/{self.default_llm_settings['filename']}",
+        self.default_provider = self._initialize_provider(self.default_llm_settings)
+        self.summary_provider = self._initialize_provider(self.summary_llm_settings)
+        self.chat_provider = self._initialize_provider(self.chat_llm_settings)
+
+    def _initialize_provider(self, llm_settings):
+        self.logger.info(f"Initializing provider with settings: {llm_settings}")
+        provider_type = llm_settings["agent_provider"]
+
+        if provider_type == "vllm_server":
+            return VLLMServerProvider(
+                llm_settings["url"],
+                llm_settings["huggingface"],
+                llm_settings["huggingface"],
+                self.config.openai_compatible_api_key,
+            )
+        elif provider_type == "llama_cpp_server":
+            return LlamaCppServerProvider(llm_settings["url"])
+        elif provider_type == "tgi_server":
+            return TGIServerProvider(server_address=llm_settings["url"])
+        elif provider_type == "llama_cpp_python":
+            llama_model = Llama(
+                model_path=f"models/{llm_settings['filename']}",
                 flash_attn=True,
                 n_threads=40,
                 n_gpu_layers=81,
                 n_batch=1024,
-                n_ctx=self.default_llm_settings["max_tokens"],
-            )
-            self.default_provider = LlamaCppPythonProvider(python_cpp_llm)
-            self.summary_provider = LlamaCppPythonProvider(python_cpp_llm)
-            self.chat_provider = LlamaCppPythonProvider(python_cpp_llm)
-        elif "tgi_server" in self.default_llm_settings["agent_provider"]:
-            from llama_cpp_agent.providers import TGIServerProvider
-            self.default_provider = TGIServerProvider(server_address=self.default_llm_settings["url"])
-            self.summary_provider = TGIServerProvider(server_address=self.summary_llm_settings["url"])
-            self.chat_provider = TGIServerProvider(server_address=self.chat_llm_settings["url"])
-        elif "vllm_server" in self.default_llm_settings["agent_provider"]:
-            from llama_cpp_agent.providers import VLLMServerProvider
-            self.default_provider = VLLMServerProvider(
-                base_url=self.default_llm_settings["url"],
-                model=self.default_llm_settings["huggingface"],
-                huggingface_model=self.default_llm_settings["huggingface"],
-            )
-            self.summary_provider = VLLMServerProvider(
-                base_url=self.summary_llm_settings["url"],
-                model=self.summary_llm_settings["huggingface"],
-                huggingface_model=self.summary_llm_settings["huggingface"],
+                n_ctx=llm_settings["max_tokens"],
             )
-            self.chat_provider = VLLMServerProvider(
-                base_url=self.chat_llm_settings["url"],
-                model=self.chat_llm_settings["huggingface"],
-                huggingface_model=self.chat_llm_settings["huggingface"],
+            return LlamaCppPythonProvider(llama_model)
+        elif provider_type == "groq":
+            return GroqProvider(
+                base_url=llm_settings["url"],
+                model=llm_settings["model"],
+                huggingface_model=llm_settings["huggingface"],
+                api_key=llm_settings["api_key"]
             )
+        elif provider_type == "llama_cpp_python_server":
+            return LlamaCppServerProvider(llm_settings["url"], llama_cpp_python_server=True)
         else:
-            self.logger.error(f"Unsupported llama-cpp-agent provider: {self.default_llm_settings['agent_provider']}, {self.summary_llm_settings['agent_provider']}, {self.chat_llm_settings['agent_provider']}")
+            self.logger.error(f"Unsupported provider: {provider_type}")
+            raise ValueError(f"Unsupported provider: {provider_type}")
 
 # Example usage
 if __name__ == "__main__":