Update embedding model configuration and simplify tool initialization in `tools.py`

src/gaia_solving_agent/tools.py

@@ -1,3 +1,6 @@
+from llama_index.core import Settings
+from llama_index.embeddings.openai import OpenAIEmbedding
+
 from llama_index.core.schema import ImageDocument, Document
 from llama_index.core.tools import FunctionTool
 from llama_index.core.tools.ondemand_loader_tool import OnDemandLoaderTool
@@ -15,7 +18,9 @@ from workflows import Context
 
 from gaia_solving_agent import TAVILY_API_KEY, NEBIUS_API_KEY, MISTRAL_API_KEY, OPENAI_API_KEY
 
+# changing the global default
 embedding_model = "text-embedding-3-small"
+Settings.embed_model = OpenAIEmbedding(model=embedding_model, api_key=OPENAI_API_KEY)
 
 
 def load_and_search_tools_from_toolspec(tool_spec: BaseToolSpec, **kwargs) -> list[FunctionTool]:
@@ -161,7 +166,6 @@ async def vllm_ask_image(query: str, images: ImageDocument | list[ImageDocument]
 
 simple_web_page_reader_tool = OnDemandLoaderTool.from_defaults(
     SimpleWebPageReader(html_to_text=True),
-    index_kwargs={"embed_model": embedding_model},
     name="simple_web_page_reader_tool",
     description="""
 Tool for loading content from a web page and return it as text.
@@ -179,12 +183,10 @@ Do not use this tool for:
 )
 simple_web_page_reader_toolspec = LoadAndSearchToolSpec.from_defaults(
     simple_web_page_reader_tool,
-    index_kwargs={"embed_model": embedding_model},
 )
 
 youtube_transcript_reader_tool = OnDemandLoaderTool.from_defaults(
     YoutubeTranscriptReader(),
-    index_kwargs={"embed_model": embedding_model},
     name="youtube_transcript_reader_tool",
     description=r"""
 Tool for loading the audio transcript from a youtube video and return it as text.
@@ -200,13 +202,11 @@ If you are provided with a youtube link in the wrong format, make it fit one the
 )
 youtube_transcript_reader_toolspec = LoadAndSearchToolSpec.from_defaults(
     youtube_transcript_reader_tool,
-    index_kwargs={"embed_model": embedding_model},
 )
 
 
 research_paper_reader_tool = OnDemandLoaderTool.from_defaults(
     ArxivReader(),
-    index_kwargs={"embed_model": embedding_model},
     name="research_paper_reader_tool",
     description=r"""
 Gets a search query, return a list of Documents of the top corresponding scientific papers on Arxiv.
@@ -214,10 +214,8 @@ Gets a search query, return a list of Documents of the top corresponding scienti
 )
 research_paper_reader_toolspec = LoadAndSearchToolSpec.from_defaults(
     research_paper_reader_tool,
-    index_kwargs={"embed_model": embedding_model},
 )
 
 wikipedia_toolspec = load_and_search_tools_from_toolspec(
     WikipediaToolSpec(),
-    index_kwargs={"embed_model": embedding_model},
 )