just create a tool for the api

app.py

@@ -5,7 +5,6 @@ import pandas as pd
 
 from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
 
-from langchain.tools import RequestsGetTool
 from langchain_community.tools import DuckDuckGoSearchRun
 from langchain.agents import initialize_agent, AgentType
 from langchain_core.prompts import ChatPromptTemplate
@@ -21,8 +20,8 @@ You are GAIA-Bot, a fully-autonomous problem-solver.
 • After finishing all reasoning, RESPOND with **only** the final answer string—no pre-amble, no labels, no extra text.
 
 You have access to these tools:
-- requests_get        – fetch web pages or files via HTTP GET.  
-- duckduckgo_search   – look up current facts on the public web.
+- gaia_file_fetch    – download the file attached to the current task_id.
+- duckduckgo_search  – look up current facts on the public web.
 
 If you require a file attached to the current task, download it with
 `requests_get("https://agents-course-unit4-scoring.hf.space/files/{task_id}")`
@@ -36,6 +35,32 @@ Adhere strictly to these rules:
 Begin.
 """
 
+class ScoringFileTool(BaseTool):
+    """
+    Download the file attached to a GAIA question.
+
+    Input: the task_id string.
+    Output: the (text) content, truncated to 20 000 chars.
+            If the file is binary (e.g. PDF, image) return a notice instead.
+    """
+    name = "gaia_file_fetch"
+    description = textwrap.dedent(f"""\
+        Use this to download the file for a GAIA task when the question
+        mentions an attachment. Input MUST be the task_id.""")
+
+    # ---- sync path (we don't need async) ------------------
+    def _run(self, task_id: str, run_manager=None) -> str:
+        url = f"https://{GAIA_HOST}/files/{task_id}"
+        resp = requests.get(url, timeout=15)
+        if resp.status_code != 200:
+            return f"ERROR {resp.status_code}: could not fetch file."
+        ctype = resp.headers.get("content-type", "")
+        # naive check: treat anything text/* or json/csv as text
+        if "text" in ctype or "json" in ctype or "csv" in ctype:
+            text = resp.text[:20_000]
+            return text
+        return f"[binary file {ctype} of {len(resp.content)} bytes downloaded]"
+
 
 class BasicAgent:
     def __init__(self):
@@ -52,7 +77,7 @@ class BasicAgent:
         )
 
         tools = [
-            RequestsGetTool(),
+            ScoringFileTool(),
             DuckDuckGoSearchRun(),
         ]
 

requirements.txt

@@ -1,8 +1,8 @@
 langchain
 langchain-huggingface
+langchain-community      # for DuckDuckGoSearchRun
 huggingface-hub
 duckduckgo-search
 gradio
 pandas
 requests
-langchain-community