Compiles and runs

Author: BrandonWeng

pyproject.toml

@@ -69,9 +69,6 @@ known-first-party = ["fluid_server"]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
 
-[tool.hatch.build.targets.wheel]
-packages = ["src"]
-
 # Entry point for the module
 [project.scripts]
 fluid-server = "fluid_server.__main__:main"

src/fluid_server/api/v1/embeddings.py

@@ -4,7 +4,7 @@
 
 import logging
 import time
-from typing import Annotated
+from typing import Annotated, List
 
 from fastapi import APIRouter, Depends, HTTPException, Request
 from fastapi.responses import JSONResponse
@@ -39,7 +39,7 @@ async def create_embeddings(
 ) -> EmbeddingResponse:
     """
     Create embeddings for text inputs (OpenAI-compatible)
-    
+
     This endpoint is compatible with OpenAI's embeddings API and can be used
     as a drop-in replacement.
     """
@@ -53,15 +53,15 @@ async def create_embeddings(
 
         # Ensure input is a list
         inputs = request.input if isinstance(request.input, list) else [request.input]
-        
+
         # Generate embeddings
         start_time = time.time()
         embeddings = await embedding_manager.get_text_embeddings(
             texts=inputs,
             model_name=request.model
         )
         processing_time = time.time() - start_time
-        
+
         # Create response data
         embedding_data = []
         for i, embedding in enumerate(embeddings):
@@ -71,28 +71,28 @@ async def create_embeddings(
                     index=i
                 )
             )
-        
+
         # Calculate usage statistics (approximate)
         total_tokens = sum(len(text.split()) for text in inputs)
         usage = EmbeddingUsage(
             prompt_tokens=total_tokens,
             total_tokens=total_tokens
         )
-        
+
         # Create response
         response = EmbeddingResponse(
             data=embedding_data,
             model=request.model,
             usage=usage
         )
-        
+
         logger.info(
             f"Generated embeddings for {len(inputs)} inputs "
             f"using model '{request.model}' in {processing_time:.2f}s"
         )
-        
+
         return response
-        
+
     except Exception as e:
         logger.error(f"Error generating embeddings: {e}")
         if isinstance(e, HTTPException):
@@ -114,15 +114,15 @@ async def create_embeddings_batch(
                 status_code=503,
                 detail="Embeddings functionality is disabled"
             )
-        
+
         responses = []
         for request in requests:
             # Process each request individually but return as batch
             response = await create_embeddings(request, embedding_manager)
             responses.append(response)
-        
+
         return responses
-        
+
     except Exception as e:
         logger.error(f"Error in batch embeddings: {e}")
         if isinstance(e, HTTPException):
@@ -140,7 +140,7 @@ async def list_embedding_models(
     """
     try:
         info = embedding_manager.get_embedding_info()
-        
+
         models = []
         for model_type, model_list in info["available_models"].items():
             for model_name in model_list:
@@ -151,12 +151,12 @@ async def list_embedding_models(
                     "owned_by": "fluid-server",
                     "model_type": f"embedding_{model_type}"
                 })
-        
+
         return JSONResponse(content={
             "object": "list",
             "data": models
         })
-        
+
     except Exception as e:
         logger.error(f"Error listing embedding models: {e}")
         raise HTTPException(status_code=500, detail=str(e))
@@ -172,7 +172,7 @@ async def get_embedding_info(
     try:
         info = embedding_manager.get_embedding_info()
         return JSONResponse(content=info)
-        
+
     except Exception as e:
         logger.error(f"Error getting embedding info: {e}")
         raise HTTPException(status_code=500, detail=str(e))

src/fluid_server/runtimes/base_embedding.py

@@ -12,8 +12,6 @@
 class EmbeddingType(Enum):
     """Supported embedding types"""
     TEXT = "text"
-    IMAGE = "image"
-    AUDIO = "audio"
 
 
 class BaseEmbeddingRuntime(ABC):
@@ -52,14 +50,14 @@ def get_info(self) -> dict[str, Any]:
     @abstractmethod
     async def embed(
         self, 
-        inputs: Union[str, List[str], bytes], 
+        inputs: Union[str, List[str]], 
         embedding_type: EmbeddingType
     ) -> List[List[float]]:
         """
         Generate embeddings for the given inputs
         
         Args:
-            inputs: Text string(s), image bytes, or audio bytes
+            inputs: Text string(s)
             embedding_type: Type of embedding to generate
             
         Returns:

src/fluid_server/runtimes/llamacpp_embedding.py

@@ -13,6 +13,8 @@
 
 logger = logging.getLogger(__name__)
 
+DEFAULT_EMBEDDING_REPO_ID = "unsloth/embeddinggemma-300m-GGUF"
+DEFAULT_EMBEDDING_FILENAME = "embeddinggemma-300M-BF16.gguf"
 
 class LlamaCppEmbeddingRuntime(BaseEmbeddingRuntime):
     """llama-cpp runtime for generating text embeddings from GGUF models"""
@@ -47,6 +49,12 @@ def _load_sync(self) -> None:
         self._Llama = Llama
         repo_id, filename, model_file = self._resolve_model_sources()
 
+        if repo_id is None and model_file is None:
+            repo_id = DEFAULT_EMBEDDING_REPO_ID
+            filename = DEFAULT_EMBEDDING_FILENAME
+        elif repo_id == DEFAULT_EMBEDDING_REPO_ID and filename is None:
+            filename = DEFAULT_EMBEDDING_FILENAME
+
         load_kwargs: dict[str, Any] = {
             "embedding": True,
             "n_ctx": 0,
@@ -191,4 +199,4 @@ def _resolve_model_sources(self) -> tuple[Optional[str], Optional[str], Optional
                 return repo_id, filename, None
             return self.model_id, None, None
 
-        return None, None, None
+        return DEFAULT_EMBEDDING_REPO_ID, DEFAULT_EMBEDDING_FILENAME, None