Upload 3 files

agent.py

@@ -1,11 +1,15 @@
 from typing import TypedDict, Annotated, List
 import operator
 import os
+import base64
+import requests
 from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain_core.messages import BaseMessage, HumanMessage
+from langchain_core.messages import BaseMessage, HumanMessage, SystemMessage
 from langgraph.graph import StateGraph, END, START
 from langgraph.prebuilt import ToolNode
 from langchain_tavily import TavilySearch
+from langchain_core.tools import tool
+from langchain_community.document_loaders import YoutubeLoader
 from dotenv import load_dotenv
 
 load_dotenv()
@@ -15,11 +19,21 @@ class AgentState(TypedDict):
     messages: Annotated[List[BaseMessage], operator.add]
 
 # 2. Define the tools
-tools = [TavilySearch(max_results=1)]
+@tool
+def get_youtube_transcript(url: str) -> str:
+    """Retrieves the transcript of a YouTube video given its URL."""
+    try:
+        loader = YoutubeLoader.from_youtube_url(url, add_video_info=False)
+        docs = loader.load()
+        return "\n".join([doc.page_content for doc in docs])
+    except Exception as e:
+        return f"Error getting transcript: {e}"
+
+tools = [TavilySearch(max_results=1), get_youtube_transcript]
 tool_node = ToolNode(tools)
 
 # 3. Define the model
-LLM = "gemini-2.0-flash-001"
+LLM = "gemini-3-pro-preview"
 model = ChatGoogleGenerativeAI(model=LLM, temperature=0)
 model = model.bind_tools(tools)
 
@@ -27,55 +41,61 @@ model = model.bind_tools(tools)
 def should_continue(state):
     messages = state['messages']
     last_message = messages[-1]
-    # If there are no tool calls, then we finish
     if not last_message.tool_calls:
         return "end"
-    # Otherwise if there are tool calls, we continue
     else:
         return "continue"
 
 def call_model(state):
     messages = state['messages']
     response = model.invoke(messages)
-    # We return a list, because this will get added to the existing list
     return {"messages": [response]}
 
 # 5. Create the graph
 workflow = StateGraph(AgentState)
-
-# Define the two nodes we will cycle between
 workflow.add_node("agent", call_model)
 workflow.add_node("action", tool_node)
-
-# Set the entrypoint as `agent`
-# This means that this node is the first one called
 workflow.add_edge(START, "agent")
-
-# We now add a conditional edge
-workflow.add_conditional_edges(
-    "agent",
-    should_continue,
-    {
-        "continue": "action",
-        "end": END,
-    },
-)
-
-# We now add a normal edge from `tools` to `agent`.
-# This means that after `tools` is called, `agent` node is called next.
+workflow.add_conditional_edges("agent", should_continue, {"continue": "action", "end": END})
 workflow.add_edge("action", "agent")
-
-# Finally, we compile it!
-# This compiles it into a LangChain Runnable,
-# meaning you can use it as you would any other runnable
 app = workflow.compile()
 
-
 class LangGraphAgent:
     def __init__(self):
         self.app = app
 
-    def __call__(self, question: str) -> str:
-        inputs = {"messages": [HumanMessage(content=question)]}
+    def __call__(self, question: str, task_id: str = None) -> str:
+        messages = [
+            SystemMessage(content="You are a helpful assistant. Answer the user's question directly and concisely. Do not include any introductory text or 'Final Answer:'. Just output the answer. If the question involves an image or video provided in the context, analyze it to answer."),
+        ]
+        
+        content = []
+        content.append({"type": "text", "text": question})
+
+        if task_id:
+            image_url = f"https://agents-course-unit4-scoring.hf.space/files/{task_id}"
+            try:
+                # Check headers first
+                response = requests.head(image_url, timeout=5)
+                if response.status_code == 200 and "image" in response.headers.get("Content-Type", ""):
+                    # Fetch the image
+                    img_response = requests.get(image_url, timeout=10)
+                    if img_response.status_code == 200:
+                        image_data = base64.b64encode(img_response.content).decode("utf-8")
+                        # Determine MIME type from header or default to jpeg
+                        mime_type = response.headers.get("Content-Type", "image/jpeg")
+                        content.append({
+                            "type": "image_url",
+                            "image_url": {"url": f"data:{mime_type};base64,{image_data}"}
+                        })
+            except Exception as e:
+                print(f"Error checking/fetching image: {e}")
+
+        messages.append(HumanMessage(content=content))
+        
+        inputs = {"messages": messages}
         final_state = self.app.invoke(inputs)
-        return final_state['messages'][-1].content
+        result = final_state['messages'][-1].content
+        if isinstance(result, list):
+            return " ".join([str(c) for c in result])
+        return str(result)

app.py

@@ -81,7 +81,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None, *args):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
+            submitted_answer = agent(question_text, task_id=task_id)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer.strip()})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e:
@@ -200,6 +200,29 @@ with gr.Blocks() as demo:
         outputs=[answer_textbox]
     )
 
+    def export_results(df):
+        if df is None or df.empty:
+            return None
+        file_path = "results.txt"
+        with open(file_path, "w", encoding="utf-8") as f:
+            for _, row in df.iterrows():
+                f.write(f"Task ID: {row.get('Task ID', 'N/A')}\n")
+                f.write(f"Question: {row.get('Question', 'N/A')}\n")
+                f.write(f"Answer: {row.get('Submitted Answer', 'N/A')}\n")
+                f.write("-" * 40 + "\n")
+        return file_path
+
+    gr.Markdown("---")
+    gr.Markdown("## Tools")
+    export_button = gr.Button("Export Results to Text")
+    file_output = gr.File(label="Download Results")
+
+    export_button.click(
+        fn=export_results,
+        inputs=[results_table],
+        outputs=[file_output]
+    )
+
 if __name__ == "__main__":
     print("\n" + "-"*30 + " App Starting " + "-"*30)
     # Check for SPACE_HOST and SPACE_ID at startup for information

requirements.txt

@@ -8,4 +8,5 @@ tavily-python
 langchain-google-genai
 google-auth
 langchain-tavily
-google-cloud-aiplatform
+google-cloud-aiplatform
+youtube-transcript-api