mirror of https://github.com/dapr/dapr-agents.git
				
				
				
			
		
			
				
	
	
		
			74 lines
		
	
	
		
			2.6 KiB
		
	
	
	
		
			Python
		
	
	
	
			
		
		
	
	
			74 lines
		
	
	
		
			2.6 KiB
		
	
	
	
		
			Python
		
	
	
	
| from typing import List, Dict, Any
 | |
| from pydantic import BaseModel, Field, ConfigDict
 | |
| from dapr_agents.storage.vectorstores import VectorStoreBase
 | |
| import logging
 | |
| 
 | |
| logger = logging.getLogger(__name__)
 | |
| 
 | |
| 
 | |
| class VectorToolStore(BaseModel):
 | |
|     """
 | |
|     Manages tool information within a vector store, providing methods for adding tools and
 | |
|     retrieving similar tools based on queries.
 | |
|     """
 | |
| 
 | |
|     vector_store: VectorStoreBase = Field(
 | |
|         ..., description="The vector store instance for tool data storage."
 | |
|     )
 | |
| 
 | |
|     model_config = ConfigDict(arbitrary_types_allowed=True)
 | |
| 
 | |
|     def add_tools(self, tools: List[Dict[str, Any]]):
 | |
|         """
 | |
|         Adds tool information to the vector store.
 | |
| 
 | |
|         Args:
 | |
|             tools (List[Dict[str, Any]]): A list of dictionaries representing tools, each containing
 | |
|                 definitions and optional metadata for each tool.
 | |
|         """
 | |
|         logger.info("Adding tools to Vector Tool Store.")
 | |
| 
 | |
|         documents = []
 | |
|         metadatas = []
 | |
| 
 | |
|         for tool in tools:
 | |
|             func_name = tool["definition"]["function"]["name"]
 | |
|             description = tool["definition"]["function"]["description"]
 | |
|             parameters = tool["definition"]["function"]["parameters"]
 | |
| 
 | |
|             # Prepare the document for each tool
 | |
|             documents.append(f"{func_name}: {description}. Args schema: {parameters}")
 | |
| 
 | |
|             # Prepare metadata, ensuring 'name' is always set
 | |
|             metadata = tool.get("metadata", {}).copy()
 | |
|             metadata.setdefault("name", func_name)  # Ensure name is set in metadata
 | |
|             metadatas.append(metadata)
 | |
| 
 | |
|         self.vector_store.add(documents=documents, metadatas=metadatas)
 | |
| 
 | |
|     def get_similar_tools(self, query_texts: str, k: int = 4) -> List[Dict[str, Any]]:
 | |
|         """
 | |
|         Retrieves tools from the vector store similar to the query text.
 | |
| 
 | |
|         Args:
 | |
|             query_texts (str): The query string to find similar tools.
 | |
|             k (int): The number of similar results to return. Defaults to 4.
 | |
| 
 | |
|         Returns:
 | |
|             List[Dict[str, Any]]: List of similar tool entries based on the query.
 | |
|         """
 | |
|         logger.info(f"Searching for tools similar to query: {query_texts}")
 | |
|         similar_docs = self.vector_store.search_similar(query_texts=query_texts, k=k)
 | |
|         return similar_docs
 | |
| 
 | |
|     def get_tool_names(self) -> List[str]:
 | |
|         """
 | |
|         Retrieves the names of all tools stored in the vector store.
 | |
| 
 | |
|         Returns:
 | |
|             List[str]: A list of tool names.
 | |
|         """
 | |
|         logger.info("Retrieving all tool names from Vector Tool Store.")
 | |
|         tools = self.vector_store.get()
 | |
|         return [tool["metadata"]["name"] for tool in tools]
 |