Add tools

pyproject.toml

@@ -12,6 +12,7 @@ dependencies = [
     "llama-index-llms-nebius>=0.1.2",
     "llama-index-multi-modal-llms-mistralai>=0.4.0",
     "llama-index-multi-modal-llms-nebius>=0.4.0",
+    "llama-index-readers-papers>=0.3.2",
     "llama-index-readers-web>=0.4.2",
     "llama-index-readers-youtube-transcript>=0.3.0",
     "llama-index-tools-duckduckgo>=0.3.0",

src/gaia_solving_agent/tools.py

@@ -1,4 +1,4 @@
-from llama_index.core.schema import ImageDocument
+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
 from llama_index.core.tools.tool_spec.base import BaseToolSpec
@@ -7,6 +7,7 @@ from llama_index.multi_modal_llms.mistralai import MistralAIMultiModal
 from llama_index.multi_modal_llms.nebius import NebiusMultiModal
 from llama_index.readers.web import SimpleWebPageReader
 from llama_index.readers.youtube_transcript import YoutubeTranscriptReader
+from llama_index.readers.papers import ArxivReader
 from tavily import AsyncTavilyClient
 from workflows import Context
 
@@ -28,6 +29,35 @@ async def tavily_search_web(query: str) -> str:
     return str(await client.search(query))
 
 
+async def get_text_representation_of_additional_file(ctx: Context) -> str :
+    """
+    Retrieves the text representation of an additional file stored in the llamaindex context.
+
+    If the retrieved file is not a valid `Document` instance, it produces a
+    message indicating the mismatched type. Otherwise, it retrieves the text content
+    from the `Document`.
+
+    Parameters
+    ----------
+    ctx : Context
+        The llamaindex context object which contains the additional file.
+
+    Returns
+    -------
+    str
+        The text representation of the additional file if it is a `Document` instance;
+        otherwise, a descriptive message indicating the mismatched type.
+    """
+    additional_file = await ctx.store.get("additional_file")
+    if not isinstance(additional_file, Document):
+        return f"The additional file is not a LlamaIndex Document but a {type(additional_file)} object. I cannot get text representation, you should try other tool."
+
+    text_representation = additional_file.get_text_content()
+    if not text_representation:
+        return f"The additional {additional_file.__class__.__name__} file does not have any text content. I cannot get text representation, you should try other tool."
+    return text_representation
+
+
 async def vllm_ask_image_tool(ctx: Context, query: str) -> str:
     """
     Asynchronously processes a visual-linguistic query paired with image data
@@ -119,3 +149,13 @@ 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)
+
+
+research_paper_reader_tool = OnDemandLoaderTool.from_defaults(
+    ArxivReader(),
+    name="research_paper_reader_tool",
+    description=r"""
+Gets a search query, return a list of Documents of the top corresponding scientific papers on Arxiv.
+""",
+)
+research_paper_reader_toolspec = LoadAndSearchToolSpec.from_defaults(research_paper_reader_tool)

uv.lock

@@ -18,6 +18,7 @@ dependencies = [
     { name = "llama-index-llms-nebius" },
     { name = "llama-index-multi-modal-llms-mistralai" },
     { name = "llama-index-multi-modal-llms-nebius" },
+    { name = "llama-index-readers-papers" },
     { name = "llama-index-readers-web" },
     { name = "llama-index-readers-youtube-transcript" },
     { name = "llama-index-tools-duckduckgo" },
@@ -46,6 +47,7 @@ requires-dist = [
     { name = "llama-index-llms-nebius", specifier = ">=0.1.2" },
     { name = "llama-index-multi-modal-llms-mistralai", specifier = ">=0.4.0" },
     { name = "llama-index-multi-modal-llms-nebius", specifier = ">=0.4.0" },
+    { name = "llama-index-readers-papers", specifier = ">=0.3.2" },
     { name = "llama-index-readers-web", specifier = ">=0.4.2" },
     { name = "llama-index-readers-youtube-transcript", specifier = ">=0.3.0" },
     { name = "llama-index-tools-duckduckgo", specifier = ">=0.3.0" },
@@ -298,6 +300,19 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/46/27/be2ead7273ecf3be0c5b3ebf9e4bccedcd49ab45a7909e82588f56189cd2/arize_phoenix_otel-0.12.1-py3-none-any.whl", hash = "sha256:85167dd061d7d4e14c98edd5733afe151da8027d69f2ecfe29a6aa2484906fa3", size = 13871, upload-time = "2025-06-24T15:58:33.702Z" },
 ]
 
+[[package]]
+name = "arxiv"
+version = "2.2.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "feedparser" },
+    { name = "requests" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/0b/16/3d72446400a59d1fbda24fed2289661398994164e07d72cfa85e43ce5e36/arxiv-2.2.0.tar.gz", hash = "sha256:6072a2211e95697092ef32acde0144d7de2cfa71208e2751724316c9df322cc0", size = 16910, upload-time = "2025-04-08T06:16:09.824Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/71/1e/e7f0393e836b5347605fc356c24d9f9ae9b26e0f7e52573b80e3d28335eb/arxiv-2.2.0-py3-none-any.whl", hash = "sha256:545b8af5ab301efff7697cd112b5189e631b80521ccbc33fbc1e1f9cff63ca4d", size = 11696, upload-time = "2025-04-08T06:16:08.844Z" },
+]
+
 [[package]]
 name = "asttokens"
 version = "3.0.0"
@@ -1666,6 +1681,20 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/68/4f/e30d4257fe9e4224f5612b77fe99aaceddae411b2e74ca30534491de3e6f/llama_index_readers_llama_parse-0.4.0-py3-none-any.whl", hash = "sha256:574e48386f28d2c86c3f961ca4a4906910312f3400dd0c53014465bfbc6b32bf", size = 2472, upload-time = "2024-11-18T00:00:07.293Z" },
 ]
 
+[[package]]
+name = "llama-index-readers-papers"
+version = "0.3.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "arxiv" },
+    { name = "defusedxml" },
+    { name = "llama-index-core" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/0a/c9/9340fd0ac123508cbcdd9673f153aa47d5ab39c66b7af192fdd13d1c5b02/llama_index_readers_papers-0.3.2.tar.gz", hash = "sha256:825bbc1875744a7e3e84ed8d2776eaba848c57357af1ec84984325a83a742ac1", size = 5238, upload-time = "2025-04-03T20:00:24.86Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/79/2e/6a5dea06fb2d19b499f8db92e29855d32fa4c529bce5360b8e689b866781/llama_index_readers_papers-0.3.2-py3-none-any.whl", hash = "sha256:9166790e26f968db010f1f87c31ef335d0e78e583348ac749a632cd0c8540bea", size = 6774, upload-time = "2025-04-03T20:00:24.05Z" },
+]
+
 [[package]]
 name = "llama-index-readers-web"
 version = "0.4.2"