Update app.py

app.py

@@ -1,211 +1,235 @@
-import asyncio
-import json
-import logging
-from typing import AsyncGenerator, Dict, Any
-
+from zoneinfo import ZoneInfo
+from google.adk.agents import Agent,BaseAgent,LlmAgent
+from google.adk.tools import google_search
+from google.adk.runners import Runner
+from google.adk.sessions import InMemorySessionService
+from google.genai import types
 import google.genai.types as types
 import requests
-from google.adk.agents import BaseAgent, LlmAgent
-from google.adk.agents.invocation_context import InvocationContext
 from google.adk.events import Event, EventActions
-from google.adk.runners import Runner
-from google.adk.sessions import InMemorySessionService
-from google.adk.tools import (
-    FunctionTool,
-    ToolContext,
-    agent_tool,
-    # built_in_code_execution,
-    google_search,
-)
+from google.adk.agents.invocation_context import InvocationContext
+from typing import AsyncGenerator
+from google.genai import types as genai_types 
+from google.adk.tools import ToolContext, FunctionTool
+import logging
+#from google.adk.tools import built_in_code_execution
+from google.adk.tools import agent_tool
 
-# Configure logging to suppress verbose output
 logging.basicConfig(level=logging.ERROR)
-
-# --- API Interaction Functions ---
-
-def answer_questions() -> list[str]:
-    """
-    Fetches the full list of evaluation questions from the scoring API.
-    Each question is formatted with its task_id.
-    """
-    print("Attempting to fetch questions from the API...")
+#from google.adk.tools import agent_tool
+url = 'https://agents-course-unit4-scoring.hf.space/questions'
+headers = {'accept': 'application/json'}
+response = requests.get(url, headers=headers)
+
+# class responses_api(BaseAgent):
+#     async def _run_async_impl(self, ctx: InvocationContext)-> AsyncGenerator[Event, None]:
+#         # This method is called when the agent is run
+#         # You can implement your logic here
+#         # For example, you can call an external API or perform some calculations
+#         # and return the result
+#         url = 'https://agents-course-unit4-scoring.hf.space/questions'
+#         headers = {'accept': 'application/json'}
+#         response = requests.get(url, headers=headers)
+#         for i in response.json():
+#             if i['file_name'] != '':
+#                 url_file = f"https://agents-course-unit4-scoring.hf.space/files/{i['task_id']}"
+#                 question = i['question']
+#                 prompt = f"{question} and the file is {url_file}, give the final answer only"
+#             else:
+#                 question = i['question']
+#                 prompt = f"{question} give the final answer only"
+#             existing_responses = ctx.session.state.get("user:responses", [])
+#             existing_responses.append(prompt)
+#             ctx.session_state["user:responses"] = existing_responses
+
+#             # Optionally, yield a single event to indicate completion or provide some output
+#             yield Event(author=self.name, content=types.Content(parts=[types.Part(text=f"Fetched {len(questions_data)} questions."))])
+
+
+
+def answer_questions():
     url = 'https://agents-course-unit4-scoring.hf.space/questions'
     headers = {'accept': 'application/json'}
-    try:
-        response = requests.get(url, headers=headers)
-        response.raise_for_status()  # Raise an exception for bad status codes
-        questions_data = response.json()
-        print(f"Successfully fetched {len(questions_data)} questions.")
-        prompts = []
-        for item in questions_data:
-            task_id = item['task_id']
-            question_text = item['question']
-            if item.get('file_name'):
-                file_url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}"
-                prompt = f"{task_id}:{question_text} The URL for the associated file is: {file_url}"
-            else:
-                prompt = f"{task_id}:{question_text}"
-            prompts.append(prompt)
-        return prompts
-    except requests.exceptions.RequestException as e:
-        print(f"Error fetching questions: {e}")
-        return []
-
-def submit_questions(answers: list[Dict[str, Any]]) -> Dict[str, Any]:
-    """
-    Submits the collected answers to the scoring API.
-
-    Args:
-        answers: A list of dictionaries, where each dictionary contains
-                 a 'task_id' and a 'submitted_answer'.
-    """
-    # !!! IMPORTANT !!!
-    # REPLACE the username and agent_code with your own details.
-    username = "YOUR_HUGGING_FACE_USERNAME"
-    agent_code_url = "https://huggingface.co/spaces/YOUR_USERNAME/YOUR_SPACE_NAME/tree/main"
-    
-    print(f"Attempting to submit {len(answers)} answers for user '{username}'...")
+    response = requests.get(url, headers=headers)
+    prompts = []
+    for i in response.json():
+        task_id = i['task_id']
+        if i['file_name'] != '':
+            url_file = f"https://agents-course-unit4-scoring.hf.space/files/{i['task_id']}"
+            question = i['question']
+            prompt = f"{task_id}:{question} and the file is {url_file}, give the final answer only"
+        else:
+            question = i['question']
+            prompt = f"{task_id}:{question} give the final answer only"
+        prompts.append(prompt)
+    return prompts
+#responses_api = responses_api(name= 'responses_api_1')
+from typing import Dict, Any
+def submit_questions(answers: list[str]) -> Dict[str, Any]:
     url = 'https://agents-course-unit4-scoring.hf.space/submit'
     payload = {
-        "username": username,
-        "agent_code": agent_code_url,
-        "answers": answers
-    }
+    "username": "ashishja",
+    "agent_code": "https://huggingface.co/spaces/ashishja/Agents_Course_Final_Assignment_Ashish/tree/main",
+    "answers": answers}
     headers = {'accept': 'application/json', "Content-Type": "application/json"}
-    
-    try:
-        response = requests.post(url, headers=headers, json=payload)
-        response.raise_for_status()
-        print("Submission successful!")
-        print("Response:", response.json())
+    response = requests.post(url, headers=headers, json =payload)
+    import json
+    print(json.dumps(payload, indent=2))
+    if response.status_code == 200:
         return response.json()
-    except requests.exceptions.RequestException as e:
-        print(f"Error submitting answers: {e}")
-        print(f"Response Body: {e.response.text if e.response else 'No response'}")
-        raise
+    else:
+        response.raise_for_status()
+
 
-# Wrap API functions in ADK Tools - The description is taken from the function's docstring.
-responses_api = FunctionTool(func=answer_questions)
+
+
+responses_api = FunctionTool(func= answer_questions)
 submit_api = FunctionTool(func=submit_questions)
 
-# --- Agent Definitions ---
+# class QuestionAnswerer(LlmAgent):
+#     async def _run_async_impl(self, ctx: InvocationContext) -> AsyncGenerator[Event, None]:
+#         questions_to_answer = ctx.session_service.get('fetched_questions', [])
+#         for q in questions_to_answer:
+#             answer = await self._llm(messages=[types.ChatMessage(role="user", parts=[types.Part(text=q)])])
+#             yield Event(author=self.name, content=answer.content)
+
+# qa = QuestionAnswerer(name = 'qa_1', model="gemini-2.0-flash", description="Question Answerer")
+
+
+
+
+
+
+
+
+APP_NAME="weather_sentiment_agent"
+USER_ID="user1234"
+SESSION_ID="1234"
 
-APP_NAME = "gaia_challenge_agent"
-USER_ID = "test_user"
-SESSION_ID = "main_session"
 
-# A specialized agent for tasks requiring code generation
 code_agent = LlmAgent(
-    name='CodeAgent',
-    model="gemini-2.0-flash-lite",
-    description="Generates Python code to analyze data files (.csv, .xlsx, .json, .py) to answer a question. Responds with only the final, exact answer.",
-    instruction=(
-        "You are an expert in data analysis and code generation. Given a question and a file URL, "
-        "write Python code to find the answer. You cannot execute the code. "
-        "Use pandas for data files. Fetch remote files using requests. "
-        "Your final output must be only the answer to the question, with no extra text or explanation."
+    name='codegaiaAgent',
+    model="gemini-2.5-pro-preview-05-06",
+    description=(
+        "You are a smart agent that can write and execute code and answer any questions provided access the given files and answer"
     ),
-    # tools=[built_in_code_execution],
+    instruction = (
+"if the question contains a file with .py ,Get the code file and depending on the question and the file provided, execute the code and provide the final answer. "
+"If the question contains a spreadsheet file like .xlsx and .csv among others, get the file and depending on the question and the file provided, execute the code and provide the final answer. "
+"use code like import pandas as pd , file = pd.read_csv('file.csv') and then use the file to answer the question. "
+"if the question contains a file with .txt ,Get the code file and depending on the question and the file provided, execute the code and provide the final answer. "
+"if the question contains a file with .json ,Get the code file and depending on the question and the file provided, execute the code and provide the final answer. "
+"If you are writing code or if you get a code file, use the code execution tool to run the code and provide the final answer. "
 )
 
-# A specialized agent for web searches
+,
+    # tools=[built_in_code_execution], 
+    # Add the responses_api agent as a tool
+    #sub_agents=[responses_api]
+)
+
+
 search_agent = LlmAgent(
-    name='SearchAgent',
-    model="gemini-2.0-flash-lite",
-    description="Searches the web to answer questions about current events, facts, or general knowledge. Responds with only the final, exact answer.",
-    instruction=(
-        "You are an expert web researcher. You will be given a question. "
-        "Use your search tool to find the most accurate information. "
-        "Synthesize the findings and provide a concise, direct answer to the question. "
-        "Your final output must be only the answer, with no extra text."
+    name='searchgaiaAgent',
+    model="gemini-2.5-pro-preview-05-06",
+    description=(
+        "You are a smart agent that can search the web and answer any questions provided access the given files and answer"
     ),
-    tools=[google_search],
+    instruction = (
+    "Get the url associated perform a search and consolidate the information provided and answer the provided question "
+)
+
+,
+    tools=[google_search], 
+    # Add the responses_api agent as a tool
+    #sub_agents=[responses_api]
 )
 
-# A specialized agent for image analysis
 image_agent = LlmAgent(
-    name='ImageAgent',
-    model="gemini-2.0-flash-lite",
-    description="Analyzes an image to answer a question about its content. Responds with only the final, exact answer.",
-    instruction=(
-        "You are an expert image analyst. You will be given a question and a URL to an image. "
-        "Analyze the image content to answer the question. "
-        "Your final output must be only the answer, with no extra text."
+    name='imagegaiaAgent',
+    model="gemini-2.5-pro-preview-05-06",
+    description=(
+        "You are a smart agent that can when given a image file and answer any questions related to it"
     ),
+    instruction = (
+    "Get the image file from the link associated in the prompt use Gemini to watch the video and answer the provided question ")
+
+,
+   # tools=[google_search], 
+    # Add the responses_api agent as a tool
+    #sub_agents=[responses_api]
 )
 
-# A specialized agent for video analysis
+
 youtube_agent = LlmAgent(
-    name='YouTubeAgent',
-    model="gemini-2.0-flash-lite",
-    description="Watches a YouTube video to answer a question about its content. Responds with only the final, exact answer.",
-    instruction=(
-        "You are an expert video analyst. You will be given a question and a URL to a YouTube video. "
-        "Analyze the video content to answer the question. "
-        "Your final output must be only the answer, with no extra text."
+    name='youtubegaiaAgent',
+    model="gemini-2.5-pro-preview-05-06",
+    description=(
+        "You are a smart agent that can when given a youtube link watch it and answer any questions related to it"
     ),
+    instruction = (
+    "Get the youtube link associated use Gemini to watch the video and answer the provided question ")
+
+,
+   # tools=[google_search], 
+    # Add the responses_api agent as a tool
+    #sub_agents=[responses_api]
 )
 
-# The main orchestrator agent
 root_agent = LlmAgent(
-    name='OrchestratorAgent',
-    model="gemini-2.0-flash-lite",
-    description="Manages a team of specialized agents to answer a list of questions and submits them for scoring.",
-    instruction=(
-        "You are the project manager. Your goal is to answer a series of questions and submit them. "
-        "1. **FETCH**: Start by using the `answer_questions` tool to get the list of all tasks. "
-        "2. **DELEGATE**: For each task string, which contains a 'task_id:question', extract the task_id and the question. "
-        "   - Determine the best specialized agent for the job (Code, Search, Image, YouTube) based on the question and any file URLs. "
-        "   - Invoke that agent with the question and necessary context (like the file URL). "
-        "3. **COLLECT**: Get the precise answer back from the specialist agent. Create a dictionary: `{'task_id': 'the_id', 'submitted_answer': 'the_answer'}`. The answer must be exact, without any extra formatting or text. "
-        "4. **SUBMIT**: After processing all questions, gather all the answer dictionaries into a single list. Call the `submit_questions` tool with this list to complete the assignment."
+    name='basegaiaAgent',
+    model="gemini-2.5-pro-preview-05-06",
+    description=(
+        "You are a smart agent that can answer any questions provided access the given files and answer"
     ),
-    tools=[
-        responses_api,
-        submit_api,
-        agent_tool.AgentTool(agent=code_agent),
-        agent_tool.AgentTool(agent=search_agent),
-        agent_tool.AgentTool(agent=image_agent),
-        agent_tool.AgentTool(agent=youtube_agent),
-    ],
+    instruction = (
+    "You are a helpful agent. When the user asks to get the questions or makes a similar request, "
+    "invoke your tool 'responses_api' to retrieve the questions. "
+    "Once you receive the list of questions, loop over each question and provide a concise answer for each based on the question and any provided file. "
+    "For every answer, return a dictionary with the keys task_id and submitted_answer, for example: "
+    "{'task_id': 'the-task-id', 'submitted_answer': 'your answer'}. "
+    "Collect all such dictionaries in a list (do not include any backslashes), and pass this list to the 'submit_api' tool to submit the answers."
 )
 
-# --- Application Runner ---
+,
+    tools=[responses_api,submit_api,agent_tool.AgentTool(agent = code_agent),\
+           agent_tool.AgentTool(agent = search_agent), agent_tool.AgentTool(youtube_agent), agent_tool.AgentTool(image_agent)], 
+    # Add the responses_api agent as a tool
+    #sub_agents=[responses_api]
+)
+
+# root_agent = LlmAgent(
+#     name='gaiaAgent',
+#     model="gemini-2.5-pro-preview-05-06",
+#     description=(
+#         "You are a smart agent that can answer any questions provided access the given files and answer"
+#     ),
+#     instruction = (
+#     "You are a helpful agent. When the user asks to get the questions or makes a similar request, "
+#     "invoke base agent. "
+#     "Once you the answers check if are in correct format. "
+#     #"Collect all such dictionaries in a list (do not include any backslashes), and pass this list to the 'submit_api' tool to submit the answers."
+# )
+
+# ,
+#     #tools=[submit_api], 
+#     # Add the responses_api agent as a tool
+#     sub_agents=[base_agent]
+# )
 
 session_service = InMemorySessionService()
+session = session_service.create_session(app_name=APP_NAME, \
+                                        user_id=USER_ID,\
+                                        session_id=SESSION_ID)
+
 runner = Runner(agent=root_agent, app_name=APP_NAME, session_service=session_service)
+# # def send_query_to_agent(root_agent, query, session):
+# #     session = session
+# #     content = types.Content(role='user', parts=[types.Part(text=query)])
+
+# # async def main():
+# #     await process_questions_and_answer()
 
-async def run_agent_process():
-    """Initializes a session and runs the agent's main task."""
-    session = await session_service.create_session(
-        app_name=APP_NAME, user_id=USER_ID, session_id=SESSION_ID
-    )
-    # Corrected the print statement below to use the SESSION_ID constant
-    print(f"===== Agent Process Started for session: {SESSION_ID} =====")
-    
-    initial_prompt = "Get all the questions, answer each one using your specialized agents, and submit the final list of answers for scoring."
-    
-    print(f"\nSending initial prompt to the Orchestrator Agent:\n'{initial_prompt}'")
-    
-    # Corrected the run_async call to use the SESSION_ID constant
-    async for event in runner.run_async(
-        session_id=SESSION_ID,
-        content=types.Content(role="user", parts=[types.Part(text=initial_prompt)]),
-    ):
-        # Optional: Print events for debugging
-        if event.action == EventActions.AGENT_RESPONSE and event.author == root_agent.name:
-            if event.content and event.content.parts:
-                print(f"\n[Orchestrator Response]: {event.content.parts[0].text}")
-        elif event.action == EventActions.TOOL_OUTPUT:
-             if event.content and event.content.parts and event.content.parts[0].tool_output:
-                tool_output = event.content.parts[0].tool_output
-                print(f"\n<-- [Tool Output] from `{tool_output.tool_name}`")
-
-    print("\n===== Agent Process Finished =====")
-
-async def main():
-    """Main entry point for the application."""
-    await run_agent_process()
-
-if __name__ == "__main__":
-    asyncio.run(main())
+# # if __name__ == "__main__":
+# #     import asyncio
+# #     asyncio.run(main())