Removed api prefix

main/api.py

@@ -80,62 +80,62 @@ class LLMApi:
             self.logger.error(f"Failed to download model {model_name}: {str(e)}")
             raise
 
-def initialize_model(self, model_name: str) -> None:
-    """
-    Initialize a model and tokenizer for text generation.
-    Handles different platforms (CUDA, MPS, CPU) appropriately.
-    """
-    self.logger.info(f"Initializing generation model: {model_name}")
-    try:
-        self.generation_model_name = model_name
-        local_model_path = self.models_path / model_name.split('/')[-1]
-
-        # Check if model exists locally
-        if local_model_path.exists():
-            self.logger.info(f"Loading model from local path: {local_model_path}")
-            model_path = local_model_path
-        else:
-            self.logger.info(f"Loading model from source: {model_name}")
-            model_path = model_name
-
-        # Check platform and set appropriate configuration
-        if torch.cuda.is_available():
-            self.logger.info("CUDA detected, using GPU with quantization")
-            quantization_config = BitsAndBytesConfig(
-                load_in_8bit=True,
-                llm_int8_threshold=3.0
-            )
-            self.generation_model = AutoModelForCausalLM.from_pretrained(
-                model_path,
-                device_map="auto",
-                quantization_config=quantization_config,
-                torch_dtype=torch.float16
-            )
-        elif torch.backends.mps.is_available():
-            self.logger.info("Apple Silicon detected, using MPS device")
-            self.generation_model = AutoModelForCausalLM.from_pretrained(
-                model_path,
-                device_map="mps",
-                torch_dtype=torch.float16
-            )
-        else:
-            self.logger.info("No GPU detected, falling back to CPU")
-            self.generation_model = AutoModelForCausalLM.from_pretrained(
-                model_path,
-                device_map="cpu",
-                torch_dtype=torch.float32  # Use full precision for CPU
-            )
+    def initialize_model(self, model_name: str) -> None:
+        """
+        Initialize a model and tokenizer for text generation.
+        Handles different platforms (CUDA, MPS, CPU) appropriately.
+        """
+        self.logger.info(f"Initializing generation model: {model_name}")
+        try:
+            self.generation_model_name = model_name
+            local_model_path = self.models_path / model_name.split('/')[-1]
 
-        self.tokenizer = AutoTokenizer.from_pretrained(model_path)
+            # Check if model exists locally
+            if local_model_path.exists():
+                self.logger.info(f"Loading model from local path: {local_model_path}")
+                model_path = local_model_path
+            else:
+                self.logger.info(f"Loading model from source: {model_name}")
+                model_path = model_name
+
+            # Check platform and set appropriate configuration
+            if torch.cuda.is_available():
+                self.logger.info("CUDA detected, using GPU with quantization")
+                quantization_config = BitsAndBytesConfig(
+                    load_in_8bit=True,
+                    llm_int8_threshold=3.0
+                )
+                self.generation_model = AutoModelForCausalLM.from_pretrained(
+                    model_path,
+                    device_map="auto",
+                    quantization_config=quantization_config,
+                    torch_dtype=torch.float16
+                )
+            elif torch.backends.mps.is_available():
+                self.logger.info("Apple Silicon detected, using MPS device")
+                self.generation_model = AutoModelForCausalLM.from_pretrained(
+                    model_path,
+                    device_map="mps",
+                    torch_dtype=torch.float16
+                )
+            else:
+                self.logger.info("No GPU detected, falling back to CPU")
+                self.generation_model = AutoModelForCausalLM.from_pretrained(
+                    model_path,
+                    device_map="cpu",
+                    torch_dtype=torch.float32  # Use full precision for CPU
+                )
 
-        # Update generation config with tokenizer-specific values
-        self.generation_config["eos_token_id"] = self.tokenizer.eos_token_id
-        self.generation_config["pad_token_id"] = self.tokenizer.eos_token_id
+            self.tokenizer = AutoTokenizer.from_pretrained(model_path)
 
-        self.logger.info(f"Successfully initialized generation model: {model_name}")
-    except Exception as e:
-        self.logger.error(f"Failed to initialize generation model {model_name}: {str(e)}")
-        raise
+            # Update generation config with tokenizer-specific values
+            self.generation_config["eos_token_id"] = self.tokenizer.eos_token_id
+            self.generation_config["pad_token_id"] = self.tokenizer.eos_token_id
+
+            self.logger.info(f"Successfully initialized generation model: {model_name}")
+        except Exception as e:
+            self.logger.error(f"Failed to initialize generation model {model_name}: {str(e)}")
+            raise
 
     def initialize_embedding_model(self, model_name: str) -> None:
         """

main/resources/config.yaml

@@ -25,7 +25,7 @@ logging:
 
 api:
   version: "v1"
-  prefix: "/api"
+  prefix: ""
   cors:
     origins: ["*"]
     credentials: true