Add LLM configuration and pipeline execution

src/llm/processor.py

@@ -23,6 +23,7 @@ class LLMProcessor:
         api_key: Optional[str] = None,
         base_url: Optional[str] = None,
         model: Optional[str] = None,
+        backend: Optional[str] = None,
     ):
         """
         Initializes the LLMProcessor.
@@ -30,14 +31,16 @@ class LLMProcessor:
         :param api_key: OpenAI API key. If None, it looks for OPENAI_API_KEY in environment variables.
         :param base_url: OpenAI-compatible base URL (e.g., for vLLM).
         :param model: The model to use for processing. If None, it looks for LLM_MODEL in environment variables.
+        :param backend: The LLM backend to use (openai, ollama, or vllm).
         """
-        backend = os.environ.get("LLM_BACKEND", "openai").lower()
+        # Use provided backend or fallback to environment variable
+        backend_env = backend or os.environ.get("LLM_BACKEND", "openai").lower()
 
-        if backend == "ollama":
+        if backend_env == "ollama":
             # Ollama's OpenAI-compatible API
             final_base_url = base_url or "http://localhost:11434/v1"
             final_api_key = api_key or "ollama"
-        elif backend == "vllm":
+        elif backend_env == "vllm":
             # Remote vLLM server
             final_base_url = base_url or os.environ.get("OPENAI_BASE_URL")
             final_api_key = api_key or os.environ.get("OPENAI_API_KEY")
@@ -45,19 +48,19 @@ class LLMProcessor:
             final_base_url = base_url or os.environ.get("OPENAI_BASE_URL")
             final_api_key = api_key or os.environ.get("OPENAI_API_KEY")
 
-        logger.info(f"Using LLM backend: {backend}")
+        logger.info(f"Using LLM backend: {backend_env}")
         try:
             self.client = OpenAI(
                 api_key=final_api_key,
                 base_url=final_base_url,
             )
             # Simple connectivity check for local backends
-            if backend == "ollama":
+            if backend_env == "ollama":
                 # We can't easily check connectivity without making a call,
                 # but we can ensure the client is initialized.
                 pass
         except Exception as e:
-            logger.error(f"Error initializing LLM client for backend {backend}: {e}")
+            logger.error(f"Error initializing LLM client for backend {backend_env}: {e}")
             raise
 
         self.model = model or os.environ.get("LLM_MODEL", "gpt-4o")

src/pipeline/orchestrator.py

@@ -39,14 +39,15 @@ logger = logging.getLogger(__name__)
 
 
 class PipelineOrchestrator:
-    def __init__(self, loop: asyncio.AbstractEventLoop):
+    def __init__(self, loop: asyncio.AbstractEventLoop, llm_config: Optional[dict] = None):
         self.loop = loop
+        self.llm_config = llm_config or {}
 
         # Modules
         self.listener = AudioListener(loop=self.loop)
         self.transcriber = Transcriber(model_size="base", device="cuda")
-        self.processor = LLMProcessor()
-        self.rag_manager = RAGManager()
+        self.processor = LLMProcessor(**self.llm_config)
+        self.rag_manager = RAGManager(llm_config=self.llm_config)
 
         # Queues
         self.stt_to_clean_queue = asyncio.Queue()

src/rag/manager.py

@@ -12,8 +12,9 @@ from src.llm.processor import LLMProcessor
 
 
 class RAGManager:
-    def __init__(self, persist_dir: str = "data/rag_index"):
+    def __init__(self, persist_dir: str = "data/rag_index", llm_config: Optional[dict] = None):
         self.persist_dir = persist_dir
+        self.llm_config = llm_config or {}
         self.db = chromadb.PersistentClient(path=self.persist_dir)
         self.collection_name = "phb_collection"
 
@@ -110,7 +111,7 @@ class RAGManager:
         if not nodes:
             return []
 
-        processor = LLMProcessor()
+        processor = LLMProcessor(**self.llm_config)
 
         # Construct the context from retrieved nodes
         context_text = "\n\n".join(