update

NOTES

@@ -3,3 +3,19 @@
 - Don't give the master any tool, since it will try to delegate smaller work to the code agent, miss context
 - Temperature to 0
 - BeautifulSoup too bad
+
+
+
+TASKS
+- MATH: 6f37996b-2ac7-44b0-8e68-6d28256631b4
+- 305ac316-eef6-4446-960a-92d80d542f82
+- 7bd855d8-463d-4ed5-93ca-5fe35145f733
+- cf106601-ab4f-4af9-b045-5295fe67b37d
+- bda648d7-d618-4883-88f4-3466eabd860e
+- 1f975693-876d-457b-a649-393859e79bf3
+- f918266a-b3e0-4914-865d-4faa564f1aef
+- 99c9cc74-fdc8-46c6-8f8d-3ce2d3bfeea3
+- 5a0c1adf-205e-4841-a666-7c3ef95def9d
+- 1f975693-876d-457b-a649-393859e79bf3
+- 3f57289b-8c60-48be-bd80-01f8099ca449
+- 9d191bce-651d-4746-be2d-7ef8ecadb9c2

app.py

@@ -207,6 +207,7 @@ class BasicAgent:
             description="""
                  This is a powerful agent, it specializes in: 
                     - Writing code to solve problem. 
+                    - Solving Maths problems. 
                     - Browse the web to find information.
                     - Reason across audio, vision, and text, a.k.a multimodal agent. """,
             add_base_tools=True,
@@ -249,20 +250,20 @@ class BasicAgent:
                2. Please answer as concisely as possible.
                3. If the question asks for a number, please return a numerical answer without unit (unless unit is specifically asked for). For example: 3 instead of three, 0 instead of None, 3 instead of $3.
                4. `pandas` package is available for reading table data from HTML content or URL. It is useful for extracting tabular data from web pages (including Wikipedia pages).
+               5. `multimodal_agent` is good at Maths, use it when facing Maths questions.
         """
         result = self.manager_agent.run(prompt)
         print(f"Agent responded with: {result}")
         return result
 
 
-def run_and_submit_all(questions_index: str, profile: gr.OAuthProfile | None):
+def run_and_submit_all(question_id: str, profile: gr.OAuthProfile | None):
     """
     Fetches all questions, runs the BasicAgent on them, submits all answers,
     and displays the results.
     """
     # --- Determine HF Space Runtime URL and Repo URL ---
     space_id = os.getenv("SPACE_ID")  # Get the SPACE_ID for sending link to the code
-    QUESTION_INDEX = int(questions_index)
 
     if profile:
         username = f"{profile.username}"
@@ -291,11 +292,9 @@ def run_and_submit_all(questions_index: str, profile: gr.OAuthProfile | None):
     try:
         response = requests.get(questions_url, timeout=15)
         response.raise_for_status()
-        questions_data = (
-            [response.json()[QUESTION_INDEX]]
-            if QUESTION_INDEX >= 0
-            else response.json()
-        )
+        questions_data = response.json()
+        if question_id:
+            questions_data = [item for item in questions_data if item.get("task_id") == question_id]
         if not questions_data:
             print("Fetched questions list is empty.")
             return "Fetched questions list is empty or invalid format.", None
@@ -422,11 +421,11 @@ with gr.Blocks() as demo:
 
     gr.LoginButton()
 
-    questions_limit = gr.Textbox(
-        label="Question index to solve (-1 to solve all)",
+    question_id = gr.Textbox(
+        label="Question id to solve (empty to solve all)",
         lines=1,
         interactive=True,
-        value="0",
+        value="6f37996b-2ac7-44b0-8e68-6d28256631b4",
     )
     run_button = gr.Button("Run Evaluation & Submit All Answers")
     status_output = gr.Textbox(
@@ -437,7 +436,7 @@ with gr.Blocks() as demo:
 
     run_button.click(
         fn=run_and_submit_all,
-        inputs=[questions_limit],
+        inputs=[question_id],
         outputs=[status_output, results_table],
     )