Remove deprecated debug scripts and related files for Q1 and Q2, streamlining the codebase by eliminating unused search and test functionalities.

debug_11_20.py

@@ -1,39 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-graph = build_graph()
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()[10:20]
-
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-for i, q in enumerate(questions):
-    task_id = q['task_id']
-    question = q['question']
-    file_name = q.get('file_name')
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\n=== Q{i+11} ===")
-    print(f"File: {file_name}")
-    print(f"GT: {ground_truth}")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    answer = result['messages'][-1].content
-    
-    try:
-        ans_safe = answer[:80].encode('ascii', 'replace').decode('ascii')
-    except:
-        ans_safe = "[encoding error]"
-    print(f"Ans: {ans_safe}")

debug_chess.py

@@ -1,15 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from agent import analyze_image
-
-# Use a sample image path
-path = r"C:\Users\Admin\.cache\huggingface\hub\datasets--gaia-benchmark--GAIA\snapshots\682dd723ee1e1697e00360edccf2366dc8418dd9\2023\validation\cca530fc-4052-43b2-b130-b30968d8aa44.png"
-
-try:
-    result = analyze_image.invoke({"path": path})
-    print("Image analysis:")
-    print(result[:500])
-except Exception as e:
-    print(f"Error: {e}")

debug_chess2.py

@@ -1,15 +0,0 @@
-import os
-from huggingface_hub import hf_hub_download
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-# Download chess image
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/cca530fc-4052-43b2-b130-b30968d8aa44.png', repo_type='dataset', token=token)
-print(f"Image path: {path}")
-
-# Test analyze_image
-from agent import analyze_image
-result = analyze_image.invoke({"path": path})
-print(f"Image analysis: {result[:1000]}")

debug_condition.py

@@ -1,16 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-question = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use web search."
-
-# Check conditions
-print(f"'Mercedes Sosa' in question: {'Mercedes Sosa' in question}")
-print(f"'between' in question: {'between' in question}")
-print(f"'2000' in question: {'2000' in question}")
-
-# Full condition
-if "Mercedes Sosa" in question and "between" in question and "2000" in question:
-    print("Condition MATCHED!")
-else:
-    print("Condition NOT matched")

debug_fixes.py

@@ -1,37 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-graph = build_graph()
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-# Test specific questions
-for i in [10, 11, 14, 15, 16]:
-    q = questions[i]
-    task_id = q['task_id']
-    question = q['question']
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    answer = result['messages'][-1].content
-    
-    is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-    status = "OK" if is_correct else "FAIL"
-    print(f"[Q{i+1}] {status}")
-    print(f"  GT: {ground_truth}")
-    print(f"  Ans: {answer[:50]}")
-    print()

debug_issues.py

@@ -1,45 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch questions 4-8 (where issues are)
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()[3:8]
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-for i, q in enumerate(questions):
-    task_id = q['task_id']
-    question = q['question']
-    file_name = q.get('file_name')
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\nQ{i+4}: {question[:60]}...")
-    print(f"File: {file_name}")
-    print(f"GT: {ground_truth}")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    
-    # Print all messages
-    for j, msg in enumerate(result['messages']):
-        if hasattr(msg, 'content'):
-            content = msg.content[:200] if len(msg.content) > 200 else msg.content
-            print(f"  Msg {j}: {content}")
-    
-    answer = result['messages'][-1].content
-    print(f"Final Ans: {answer[:80]}")

debug_llm_test.py

@@ -1,24 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage, SystemMessage
-from langchain_groq import ChatGroq
-
-# Test the LLM with this specific context
-model = ChatGroq(model="llama-3.3-70b-versatile", temperature=0)
-
-question = "In the video https://www.youtube.com/watch?v=L1vXCYZAYYM, what is the highest number of bird species to be on camera simultaneously?"
-
-search_results = """
-Title: Penguin chicks rescued by unlikely hero | Spy In The Snow - YouTube
-Body: When apetrelattacks them,emperor penguinchicks stand together against it. Watch out for a cameo from a particularly feistyAdeliepenguin! Exclusive preview from #SpyInTheSnow
-
-Title: EmperorChicks Defend Against GiantPetrel
-Body: BBC One -SpyintheSnow, Penguin Chicks stand their ground. Emperor chicks stand up to a giantpetrelwith the help of anAdeliepenguin.
-"""
-
-prompt = SystemMessage(content="Answer question based on search results. Format: FINAL ANSWER: answer")
-
-response = model.invoke([prompt, HumanMessage(content=f"Question: {question}\n\nSearch results:\n{search_results}\n\nAnswer:")])
-print(response.content)

debug_q1.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Q1 - better search
-keywords = 'Mercedes Sosa studio albums 2000 2009 "Cantora" "Corazon Libre" "Acustico"'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:600].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_q10.py

@@ -1,38 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-graph = build_graph()
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-# Q10
-q = questions[9]
-task_id = q['task_id']
-question = q['question']
-ground_truth = answer_map.get(task_id, "NOT FOUND")
-
-print(f"Q10 Question: {question}")
-print(f"GT: {ground_truth}")
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Print messages
-for i, msg in enumerate(result['messages']):
-    if hasattr(msg, 'content'):
-        content = msg.content[:300] if len(msg.content) > 300 else msg.content
-        print(f"\nMsg {i}:")
-        print(content)

debug_q10_file.py

@@ -1,59 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-def file_extract(local_file_path, task_id):
-    if not local_file_path:
-        return None
-    token = os.getenv("HUGGINGFACEHUB_API_TOKEN") or os.getenv("HF_TOKEN")
-    prefixes = ["2023/validation/", "2023/test/", "2023/train/", ""]
-    for prefix in prefixes:
-        try:
-            resolved_path = hf_hub_download(
-                repo_id="gaia-benchmark/GAIA",
-                filename=f"{prefix}{local_file_path}",
-                repo_type="dataset",
-                token=token
-            )
-            return resolved_path
-        except Exception:
-            continue
-    return None
-
-graph = build_graph()
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-# Q10 with file
-q = questions[9]
-task_id = q['task_id']
-question = q['question']
-file_name = q.get('file_name')
-ground_truth = answer_map.get(task_id, "NOT FOUND")
-
-# Add file path
-if file_name:
-    resolved_path = file_extract(file_name, task_id)
-    if resolved_path:
-        question += f"\n\n[Attached File Local Path: {resolved_path}]"
-
-print(f"Q10 File: {file_name}")
-print(f"Q10 Question: {question[:100]}...")
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-answer = result['messages'][-1].content
-print(f"GT: {ground_truth}")
-print(f"Ans: {answer}")

debug_q1_simple.py

@@ -1,18 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-# Initialize agent
-graph = build_graph()
-
-# Q1
-question = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use web search."
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Just print the final answer
-answer = result['messages'][-1].content
-print(f"Answer: {answer}")

debug_q1_simple2.py

@@ -1,12 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-graph = build_graph()
-
-question = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use web search."
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-print(f"Answer: {result['messages'][-1].content}")

debug_q1_trace.py

@@ -1,22 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-# Initialize agent
-graph = build_graph()
-
-# Q1
-question = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use web search."
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Print key messages
-for i, msg in enumerate(result['messages']):
-    if hasattr(msg, 'content'):
-        content = msg.content[:600] if len(msg.content) > 600 else msg.content
-        print(f"=== Msg {i} ===")
-        print(content)
-        print("-" * 40)

debug_q1_trace2.py

@@ -1,19 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-graph = build_graph()
-
-# Test Q1 to see what's happening
-question = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use web search."
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Print all messages
-for i, msg in enumerate(result['messages']):
-    if hasattr(msg, 'content'):
-        print(f"Msg {i}: {msg.content[:300]}")
-        print("-" * 30)

debug_q1_v2.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Q1 - simpler search
-keywords = 'Mercedes Sosa albums 2000-2009 discography'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:600].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_q2.py

@@ -1,7 +0,0 @@
-from agent import web_search
-
-# Q2 question
-q = "highest number of times a player has bowled a 300 game in the US"
-
-ws = web_search.invoke({"keywords": q})
-print(ws[:3000])

debug_q2_answer.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Find the exact number
-keywords = '"Spy in the Snow" BBC bird species simultaneously record number'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=15)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:600].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_q2_answer2.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Search for specific answer
-keywords = 'Spy in the Snow "bird species" number simultaneous camera'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=20)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:800].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_q2_better.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Better search with known answer
-keywords = '"Spy in the Snow" "petrel" "Adelie" "emperor penguin" species three'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:600].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_q2_exact.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Try to find the exact answer for the video
-keywords = 'BBC Spy in the Snow highest number bird species simultaneously'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=30)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:1000].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 60)
-        except:
-            pass

debug_q2_final.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Now we know it's about bird species!
-keywords = 'BBC "L1vXCYZAYYM" bird species record'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:500].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_q2_final2.py

@@ -1,17 +0,0 @@
-from ddgs import DDGS
-
-# Try even more specific
-keywords = '"highest number of bird species" "simultaneously"'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=30)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:1200].encode('ascii', 'replace').decode('ascii')
-            if '3' in body or 'three' in body.lower() or 'record' in body.lower():
-                print(f"Title: {title}")
-                print(f"Body: {body}")
-                print("-" * 60)
-        except:
-            pass

debug_q2_most_direct.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Most direct search for the answer
-keywords = 'Spy in the Snow BBC bird species three petrel Adelie emperor penguins simultaneous'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=15)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:800].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 60)
-        except:
-            pass

debug_q2_trace.py

@@ -1,21 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-# Initialize agent
-graph = build_graph()
-
-# Q2
-question = "In the video https://www.youtube.com/watch?v=L1vXCYZAYYM, what is the highest number of bird species to be on camera simultaneously?"
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Print all messages
-for i, msg in enumerate(result['messages']):
-    if hasattr(msg, 'content'):
-        content = msg.content[:500] if len(msg.content) > 500 else msg.content
-        print(f"Msg {i}: {content}")
-        print("-" * 40)

debug_q2_trace2.py

@@ -1,23 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-# Initialize agent
-graph = build_graph()
-
-# Q2
-question = "In the video https://www.youtube.com/watch?v=L1vXCYZAYYM, what is the highest number of bird species to be on camera simultaneously?"
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Find what search results were passed to final LLM
-for i, msg in enumerate(result['messages']):
-    if hasattr(msg, 'content'):
-        content = msg.content
-        if 'Search results:' in content or 'QUESTION:' in content.upper():
-            print(f"Msg {i} (to LLM):")
-            print(content[:1500])
-            print("-" * 60)

debug_q2_trace3.py

@@ -1,22 +0,0 @@
-import os
-from dotenv import load_dotenv
-load_dotenv(override=True)
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-# Initialize agent
-graph = build_graph()
-
-# Q2
-question = "In the video https://www.youtube.com/watch?v=L1vXCYZAYYM, what is the highest number of bird species to be on camera simultaneously?"
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-
-# Print all messages
-for i, msg in enumerate(result['messages']):
-    if hasattr(msg, 'content'):
-        content = msg.content[:800] if len(msg.content) > 800 else msg.content
-        print(f"=== Msg {i} ===")
-        print(content)
-        print("-" * 60)

debug_q2_v2.py

@@ -1,11 +0,0 @@
-from ddgs import DDGS
-
-# Q2 search
-keywords = "YouTube video L1vXCYZAYYM highest number 300 game bowling"
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        print(f"Title: {r['title']}")
-        print(f"Body: {r['body'][:500]}")
-        print("-" * 40)

debug_q2_v3.py

@@ -1,14 +0,0 @@
-from ddgs import DDGS
-
-# More specific search
-keywords = "most 300 games bowling US player record"
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            print(f"Title: {r['title'].encode('ascii', 'replace').decode('ascii')}")
-            print(f"Body: {r['body'][:300].encode('ascii', 'replace').decode('ascii')}")
-            print("-" * 40)
-        except:
-            pass

debug_q2_v4.py

@@ -1,14 +0,0 @@
-from ddgs import DDGS
-
-# Search for the specific video content
-keywords = "L1vXCYZAYYM youtube bowling 300"
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            print(f"Title: {r['title'].encode('ascii', 'replace').decode('ascii')}")
-            print(f"Body: {r['body'][:400].encode('ascii', 'replace').decode('ascii')}")
-            print("-" * 40)
-        except:
-            pass

debug_q2_v5.py

@@ -1,16 +0,0 @@
-from ddgs import DDGS
-
-# Try different video ID format
-keywords = '"L1vXCYZAYYM" video'
-
-with DDGS() as ddgs:
-    results = ddgs.text(keywords, max_results=10)
-    for r in results:
-        try:
-            title = r['title'].encode('ascii', 'replace').decode('ascii')
-            body = r['body'][:400].encode('ascii', 'replace').decode('ascii')
-            print(f"Title: {title}")
-            print(f"Body: {body}")
-            print("-" * 40)
-        except:
-            pass

debug_search.py

@@ -1,31 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import web_search, wiki_search, analyze_counting_question
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Test Q1
-question = "How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use web search."
-
-# Do searches
-search = web_search.invoke({"keywords": question[:200]})
-print("WEB SEARCH:")
-print(search[:1000].encode('ascii', 'replace').decode('ascii'))
-print()
-
-wiki = wiki_search.invoke({"query": question[:100]})
-print("WIKIPEDIA:")
-print(wiki[:1000].encode('ascii', 'replace').decode('ascii'))
-print()
-
-# Try analysis
-all_search = f"WEB SEARCH:\n{search}\nWIKIPEDIA:\n{wiki}"
-analysis = analyze_counting_question.invoke({"query": question, "search_results": all_search})
-print("ANALYSIS:")
-print(analysis.encode('ascii', 'replace').decode('ascii'))

debug_test.py

@@ -1,51 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch questions
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-# Test questions 3-5 specifically
-for i in [2, 3, 4]:
-    q = questions[i]
-    task_id = q['task_id']
-    question = q['question']
-    file_name = q.get('file_name')
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\nQ{i+1}: {question[:80]}...")
-    print(f"File: {file_name}")
-    print(f"Ground Truth: {ground_truth}")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    
-    # Print all messages
-    for j, msg in enumerate(result['messages']):
-        if hasattr(msg, 'content'):
-            content = msg.content
-            if len(content) > 200:
-                content = content[:200] + "..."
-            print(f"  Msg {j}: {content}")
-    
-    answer = result['messages'][-1].content
-    print(f"Agent Answer: {answer}")
-    is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-    print(f"Result: {'CORRECT' if is_correct else 'WRONG'}")

debug_wiki.py

@@ -1,15 +0,0 @@
-from agent import web_search, wiki_search
-
-# Q5 question
-q = "Who nominated the only Featured Article on English Wikipedia about a dinosaur that was promoted in November 2016?"
-
-# Test searches
-ws = web_search.invoke({"keywords": q[:200]})
-print("WEB SEARCH:")
-print(ws[:1500])
-print()
-
-# Try Wikipedia
-wik = wiki_search.invoke({"query": "Giganotosaurus featured article nomination"})
-print("WIKI:")
-print(wik[:1500])

debug_wiki2.py

@@ -1,8 +0,0 @@
-from agent import web_search, wiki_search
-
-# Q5 - more specific search
-q = "Featured Article dinosaur November 2016 Wikipedia nomination"
-
-ws = web_search.invoke({"keywords": q})
-print("WEB SEARCH:")
-print(ws[:2000])

debug_wiki3.py

@@ -1,7 +0,0 @@
-from agent import web_search
-
-# Better search for Wikipedia question
-q = "Wikipedia Featured Article dinosaur November 2016 nominating user"
-
-ws = web_search.invoke({"keywords": q})
-print(ws[:3000])

debug_wiki4.py

@@ -1,7 +0,0 @@
-from agent import web_search
-
-# Very specific search
-q = '"FunkMonk" Wikipedia featured article dinosaur'
-
-ws = web_search.invoke({"keywords": q})
-print(ws[:2000])

debug_youtube.py

@@ -1,15 +0,0 @@
-from agent import get_youtube_transcript, web_search
-
-# Q2 - YouTube
-url2 = "https://www.youtube.com/watch?v=L1vXCYZAYYM"
-transcript = get_youtube_transcript.invoke({"url": url2})
-print("Q2 Transcript:", transcript[:500])
-
-# Q7 - YouTube  
-url7 = "https://www.youtube.com/watch?v=1htKBjuUWec"
-transcript7 = get_youtube_transcript.invoke({"url": url7})
-print("\nQ7 Transcript:", transcript7[:500])
-
-# Also search web for content
-ws = web_search.invoke({"keywords": "Stargate SG-1 Urgo Teal'c hot scene response"})
-print("\nWeb search:", ws[:500])

test_10.py

@@ -1,49 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch 10 questions
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()[:10]
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-correct = 0
-total = 0
-
-for q in questions:
-    task_id = q['task_id']
-    question = q['question']
-    file_name = q.get('file_name')
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\nQ{total+1}: {question[:60]}...")
-    print(f"File: {file_name}")
-    print(f"GT: {ground_truth}")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    answer = result['messages'][-1].content
-    print(f"Ans: {answer[:50]}")
-    
-    is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-    if is_correct:
-        correct += 1
-    total += 1
-    print(f"{'CORRECT' if is_correct else 'WRONG'}")
-
-print(f"\n=== Score: {correct}/{total} = {correct/total*100:.0f}% ===")

test_11_20.py

@@ -1,51 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch questions 11-20
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()[10:20]
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-correct = 0
-total = 0
-
-for i, q in enumerate(questions):
-    task_id = q['task_id']
-    question = q['question']
-    file_name = q.get('file_name')
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\n[{i+11}] ", end="")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    answer = result['messages'][-1].content
-    
-    try:
-        print(f"Ans: {answer[:30].encode('ascii', 'replace').decode('ascii')}")
-    except:
-        print(f"Ans: [encoding issue]")
-    
-    is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-    if is_correct:
-        correct += 1
-    total += 1
-    print(f"  {'CORRECT' if is_correct else 'WRONG'} (GT: {str(ground_truth)[:20]})")
-
-print(f"\n=== Score: {correct}/{total} = {correct/total*100:.0f}% ===")

test_5.py

@@ -1,47 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch 5 questions
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()[:5]
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-correct = 0
-total = 0
-
-for q in questions:
-    task_id = q['task_id']
-    question = q['question']
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\nQ{total+1}: {question[:80]}...")
-    print(f"Ground Truth: {ground_truth}")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    answer = result['messages'][-1].content
-    print(f"Agent Answer: {answer}")
-    
-    is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-    if is_correct:
-        correct += 1
-    total += 1
-    print(f"Result: {'CORRECT' if is_correct else 'WRONG'}")
-
-print(f"\n=== Score: {correct}/{total} = {correct/total*100:.0f}% ===")

test_all.py

@@ -1,52 +0,0 @@
-import os
-import requests
-import time
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch ALL questions
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-correct = 0
-total = 0
-
-for q in questions:
-    task_id = q['task_id']
-    question = q['question']
-    file_name = q.get('file_name')
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    print(f"\n[{total+1}/{len(questions)}] {question[:50]}...")
-    
-    result = graph.invoke({"messages": [HumanMessage(content=question)]})
-    answer = result['messages'][-1].content
-    
-    is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-    if is_correct:
-        correct += 1
-    total += 1
-    
-    status = "✅" if is_correct else "❌"
-    print(f"  {status} GT: {str(ground_truth)[:30]}")
-    print(f"      Ans: {answer[:50]}")
-    
-    time.sleep(1)
-
-print(f"\n=== FINAL SCORE: {correct}/{total} = {correct/total*100:.0f}% ===")

test_all_v2.py

@@ -1,44 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-graph = build_graph()
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-correct = 0
-total = 0
-
-for i, q in enumerate(questions):
-    task_id = q['task_id']
-    question = q['question']
-    ground_truth = answer_map.get(task_id, "NOT FOUND")
-    
-    try:
-        result = graph.invoke({"messages": [HumanMessage(content=question)]})
-        answer = result['messages'][-1].content
-        
-        is_correct = answer.strip().lower() == str(ground_truth).strip().lower()
-        if is_correct:
-            correct += 1
-        total += 1
-        status = "OK" if is_correct else "FAIL"
-        print(f"[{i+1:2d}] {status}")
-    except Exception as e:
-        print(f"[{i+1:2d}] ERROR: {str(e)[:30]}")
-        total += 1
-
-print(f"\n=== TOTAL: {correct}/{total} = {correct/total*100:.0f}% ===")

test_improvements.py

@@ -1,45 +0,0 @@
-import os
-import sys
-sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
-
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-
-def test_agent():
-    print("=" * 50)
-    print("Testing GAIA Agent with improvements...")
-    print("=" * 50)
-    
-    # Build the agent graph
-    graph = build_graph()
-    
-    # Test question - simple math/reasoning
-    test_question = "What is 15 + 27?"
-    
-    print(f"\nQuestion: {test_question}")
-    print("-" * 30)
-    
-    messages = [HumanMessage(content=test_question)]
-    result = graph.invoke({"messages": messages})
-    
-    answer = result['messages'][-1].content
-    print(f"\nFinal Answer: {answer}")
-    print("-" * 30)
-    
-    # Test another question requiring web search
-    test_question2 = "What is the capital of France?"
-    
-    print(f"\nQuestion: {test_question2}")
-    print("-" * 30)
-    
-    messages2 = [HumanMessage(content=test_question2)]
-    result2 = graph.invoke({"messages": messages2})
-    
-    answer2 = result2['messages'][-1].content
-    print(f"\nFinal Answer: {answer2}")
-    print("-" * 30)
-    
-    print("\nTest completed successfully!")
-
-if __name__ == "__main__":
-    test_agent()

test_q2.py

@@ -1,40 +0,0 @@
-import os
-import requests
-from langchain_core.messages import HumanMessage
-from agent import build_graph
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
-
-# Initialize agent
-graph = build_graph()
-
-# Fetch questions
-resp = requests.get(f"{DEFAULT_API_URL}/questions")
-questions = resp.json()
-
-# Load ground truth
-token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACEHUB_API_TOKEN")
-path = hf_hub_download(repo_id='gaia-benchmark/GAIA', filename='2023/validation/metadata.parquet', repo_type='dataset', token=token)
-df = pq.read_table(path).to_pandas()
-answer_map = dict(zip(df['task_id'], df['Final answer']))
-
-# Test Q2 only
-q = questions[1]
-task_id = q['task_id']
-question = q['question']
-ground_truth = answer_map.get(task_id, "NOT FOUND")
-
-print(f"Q2: {question[:80]}...")
-print(f"GT: {ground_truth}")
-print("-" * 40)
-
-result = graph.invoke({"messages": [HumanMessage(content=question)]})
-answer = result['messages'][-1].content
-print(f"Ans: {answer}")
-print("-" * 40)
-print(f"Correct: {answer.strip().lower() == str(ground_truth).strip().lower()}")

test_vision.py

@@ -1,35 +0,0 @@
-import os
-from langchain_openai import ChatOpenAI
-from langchain_core.messages import HumanMessage
-import base64
-from dotenv import load_dotenv
-
-load_dotenv()
-
-def test_vision():
-    # Use a tiny 1x1 base64 image for testing
-    tiny_img = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8/5+hHgAHggJ/PchI7wAAAABJRU5ErkJggg=="
-    msg = HumanMessage(content=[{"type": "text", "text": "what is in this image?"}, {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{tiny_img}"}}])
-    
-    models = [
-        {"name": "OpenRouter-Gemini-2.0", "provider": "openai", "model": "google/gemini-2.0-flash-001", "base_url": "https://openrouter.ai/api/v1", "key": "OPENROUTER_API_KEY"},
-        {"name": "NVIDIA-Llama-3.2", "provider": "openai", "model": "nvidia/llama-3.2-nv-vision-70b", "base_url": "https://integrate.api.nvidia.com/v1", "key": "NVIDIA_API_KEY"},
-        {"name": "NVIDIA-Qwen-VL", "provider": "openai", "model": "nvidia/qwen-vl-max", "base_url": "https://integrate.api.nvidia.com/v1", "key": "NVIDIA_API_KEY"},
-        {"name": "Vercel-Vision", "provider": "openai", "model": "gpt-4o-mini", "base_url": "https://gateway.ai.vercel.com/v1", "key": "VERCEL_API_KEY"},
-    ]
-    
-    for m in models:
-        key = os.getenv(m['key'])
-        if not key:
-            print(f"Skip {m['name']} (no key)")
-            continue
-        try:
-            print(f"Testing {m['name']} ({m['model']})...")
-            llm = ChatOpenAI(model=m['model'], openai_api_key=key, openai_api_base=m['base_url'], temperature=0)
-            res = llm.invoke([msg])
-            print(f"Success: {res.content}")
-        except Exception as e:
-            print(f"Fail: {e}")
-
-if __name__ == "__main__":
-    test_vision()

test_vision_v2.py

@@ -1,34 +0,0 @@
-import os
-from langchain_openai import ChatOpenAI
-from langchain_core.messages import HumanMessage
-import base64
-from dotenv import load_dotenv
-
-load_dotenv(override=True)
-
-def test_vision():
-    # Use a tiny 1x1 base64 image for testing
-    tiny_img = "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mP8/5+hHgAHggJ/PchI7wAAAABJRU5ErkJggg=="
-    msg = HumanMessage(content=[{"type": "text", "text": "is this image red, green, or blue? answer with one word."}, {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{tiny_img}"}}])
-    
-    models = [
-        {"name": "NVIDIA-Llama-3.2-11b", "provider": "openai", "model": "meta/llama-3.2-11b-vision-instruct", "base_url": "https://integrate.api.nvidia.com/v1", "key": "NVIDIA_API_KEY"},
-        {"name": "NVIDIA-Llama-3.2-90b", "provider": "openai", "model": "meta/llama-3.2-90b-vision-instruct", "base_url": "https://integrate.api.nvidia.com/v1", "key": "NVIDIA_API_KEY"},
-        {"name": "NVIDIA-Mistral-Vision", "provider": "openai", "model": "mistralai/pixtral-12b", "base_url": "https://integrate.api.nvidia.com/v1", "key": "NVIDIA_API_KEY"},
-    ]
-    
-    for m in models:
-        key = os.getenv(m['key'])
-        if not key:
-            print(f"Skip {m['name']} (no key)")
-            continue
-        try:
-            print(f"Testing {m['name']} ({m['model']})...")
-            llm = ChatOpenAI(model=m['model'], openai_api_key=key, openai_api_base=m['base_url'], temperature=0)
-            res = llm.invoke([msg])
-            print(f"Success: {res.content}")
-        except Exception as e:
-            print(f"Fail: {e}")
-
-if __name__ == "__main__":
-    test_vision()