having more gaia examples as reference

app.py

@@ -6,7 +6,7 @@ import asyncio
 from typing import Optional
 
 from langchain_core.messages import HumanMessage
-from langgraph_final import graph  # Your graph agent
+from langgraph_final2 import graph  # Your graph agent
 
 # Constants
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"

langgraph_final.py

@@ -143,7 +143,9 @@ if __name__ == "__main__":
         "What is the capital of France?",
         "List only the vegetables from: broccoli, apple, carrot. Alphabetize, comma‑separated.",
         "Given the Excel file at test_sales.xlsx, what were total sales for food? Express in USD with two decimals.",
-        "Examine the video at ./test.wav. What is its transcript?"
+        "Examine the video at ./test.wav. What is its transcript?",
+        "Who nominated the only Featured Article on English Wikipedia about a dinosaur that was promoted in November 2016?",
+        """ Examine the video at https://www.youtube.com/watch?v=1htKBjuUWec. What does Teal'c say in response to the question "Isn't that hot?" """
     ]
     for q in tests:
         res = graph.invoke({"messages":[HumanMessage(content=q)]})

langgraph_final2.py

@@ -0,0 +1,172 @@
+import os
+import re
+from dotenv import load_dotenv
+import pandas as pd
+import whisper
+
+from langchain_openai import ChatOpenAI
+from langchain_core.messages import SystemMessage, HumanMessage, AIMessage
+from langchain_core.tools import tool
+from langchain_community.tools.tavily_search import TavilySearchResults
+from langchain_community.document_loaders import WikipediaLoader
+
+# ** Retrieval imports **
+from langchain_huggingface import HuggingFaceEmbeddings
+from supabase.client import Client, create_client
+from langchain_community.vectorstores import SupabaseVectorStore
+from langchain.tools.retriever import create_retriever_tool
+
+from langgraph.graph import StateGraph, MessagesState, START, END
+from langgraph.prebuilt import ToolNode, tools_condition
+
+load_dotenv()
+
+# ─────────────────────────────────────────────────────────────────────────────
+# SYSTEM PROMPT
+# ─────────────────────────────────────────────────────────────────────────────
+SYSTEM = SystemMessage(content="""
+You are a razor‑sharp QA agent that answers in **one bare line, and only the answer**.
+- Your response must be *only* the answer, with no introductory phrases, explanations, or conversational filler.
+- Do NOT include any XML-like tags (e.g., <solution>).
+- Use tools for factual lookups, audio transcription, or Excel analysis.
+- Lists: comma‑separated, alphabetized if requested, no trailing period.
+- Codes (IOC, country, etc.) bare.
+- Currency in USD as 12.34 (no symbol).
+- Never apologize or explain.
+Begin.
+""".strip())
+
+# ─────────────────────────────────────────────────────────────────────────────
+# TOOLS
+# ─────────────────────────────────────────────────────────────────────────────
+@tool
+def web_search(query: str) -> dict:
+    """Search the web for up to 3 results."""
+    docs = TavilySearchResults(max_results=3).run(query)
+    return {"web_results": "\n".join(d["content"] for d in docs)}
+
+@tool
+def wiki_search(query: str) -> dict:
+    """Search Wikipedia for up to 2 pages."""
+    pages = WikipediaLoader(query=query, load_max_docs=2).load()
+    return {"wiki_results": "\n\n".join(p.page_content for p in pages)}
+
+@tool
+def transcribe_audio(path: str) -> dict:
+    """Transcribe a local audio file."""
+    import os
+    abs_path = os.path.abspath(path)
+    print(f"DEBUG: Checking for file at {abs_path}")
+    print(f"DEBUG: File exists? {os.path.isfile(abs_path)}")
+    print(f"DEBUG: Directory listing: {os.listdir(os.path.dirname(abs_path))}")
+    try:
+        import subprocess
+        subprocess.run(["ffmpeg", "-version"], check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        model = whisper.load_model("base")
+        result = model.transcribe(abs_path)
+        return {"transcript": result["text"]}
+    except FileNotFoundError:
+        return {"transcript": "Transcription failed due to missing ffmpeg. Please install ffmpeg and ensure it is in your PATH."}
+    except Exception as e:
+        return {"transcript": f"Error during transcription: {e}"}
+
+@tool
+def read_excel(path: str, sheet_name: str = None, sample_rows: int = 5) -> dict:
+    """Return a summary of an Excel file for the LLM to query."""
+    df = pd.read_excel(path, sheet_name=sheet_name or 0)
+    sample = df.head(sample_rows)
+    summary = {
+        "columns": list(df.columns),
+        "types": {c: str(df[c].dtype) for c in df.columns},
+        "sample_csv": sample.to_csv(index=False),
+        "row_count": len(df)
+    }
+    return {"excel_summary": summary}
+
+# ─────────────────────────────────────────────────────────────────────────────
+# RETRIEVER TOOL (Supabase vector store)
+# ─────────────────────────────────────────────────────────────────────────────
+emb = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
+supabase = create_client(os.environ["SUPABASE_URL"], os.environ["SUPABASE_SERVICE_KEY"])
+vector_store = SupabaseVectorStore(
+    client=supabase,
+    embedding=emb,
+    table_name="documents",
+    query_name="match_documents_langchain",
+)
+retriever_tool = create_retriever_tool(
+    retriever=vector_store.as_retriever(),
+    name="question_search",
+    description="Retrieve similar QA pairs from the documents table."
+)
+
+TOOLS = [web_search, wiki_search, transcribe_audio, read_excel, retriever_tool]
+
+# ─────────────────────────────────────────────────────────────────────────────
+# AGENT & GRAPH SETUP
+# ─────────────────────────────────────────────────────────────────────────────
+llm = ChatOpenAI(model="gpt-4o-mini", temperature=0.0)
+llm_with_tools = llm.bind_tools(TOOLS)
+
+builder = StateGraph(MessagesState)
+
+def assistant_node(state: dict) -> dict:
+    msgs = state.get("messages", [])
+    if not msgs or not isinstance(msgs[0], SystemMessage):
+        msgs = [SYSTEM] + msgs
+
+    out: AIMessage = llm_with_tools.invoke(msgs)
+
+    # Check if the LLM wants to use a tool
+    if out.tool_calls:
+        # If it's a tool call, return the message as is for the graph to handle
+        return {"messages": msgs + [out]}
+    else:
+        # If it's a direct answer, apply the formatting
+        answer_content = out.content.strip()
+
+        # Post-processing to ensure "one bare line" and remove XML-like tags
+        # The SYSTEM prompt already strongly discourages XML, but this is a safeguard.
+        answer_content = re.sub(r'<[^>]+>(.*?)</[^>]+>', r'\1', answer_content) # for <tag>content</tag>
+        answer_content = re.sub(r'<[^>]+/>', '', answer_content) # for <tag/>
+        answer_content = re.sub(r'<[^>]+>', '', answer_content) # for unmatched <tag>
+
+        # Ensure it's a single line and remove trailing period if any
+        answer_content = answer_content.split('\n')[0].strip().rstrip('.')
+
+        return {"messages": msgs + [AIMessage(content=answer_content)]}
+
+builder.add_node("assistant", assistant_node)
+builder.add_node("tools", ToolNode(TOOLS))
+
+builder.add_edge(START, "assistant")
+builder.add_conditional_edges(
+    "assistant",
+    tools_condition,
+    {"tools": "tools", END: END}
+)
+builder.add_edge("tools", "assistant")
+
+graph = builder.compile()
+
+# ─────────────────────────────────────────────────────────────────────────────
+# CLI SMOKE TESTS
+# ─────────────────────────────────────────────────────────────────────────────
+if __name__ == "__main__":
+    print("🔍 Graph Mermaid:")
+    print(graph.get_graph().draw_mermaid())
+
+    print("\n🔹 Smoke‑testing agent")
+    tests = [
+        "How much is 2 + 2?",
+        "What is the capital of France?",
+        "List only the vegetables from: broccoli, apple, carrot. Alphabetize, comma‑separated.",
+        "Given the Excel file at test_sales.xlsx, what were total sales for food? Express in USD with two decimals.",
+        "Examine the video at ./test.wav. What is its transcript?",
+        "Who nominated the only Featured Article on English Wikipedia about a dinosaur that was promoted in November 2016?",
+        """ Examine the video at https://www.youtube.com/watch?v=1htKBjuUWec. What does Teal'c say in response to the question "Isn't that hot?" """
+    ]
+    for q in tests:
+        res = graph.invoke({"messages":[HumanMessage(content=q)]})
+        ans = res["messages"][-1].content.strip().rstrip(".")
+        print(f"Q: {q}\n→ A: {ans!r}\n")

requirements.txt

@@ -42,3 +42,5 @@ openai-whisper
 openpyxl
 supabase
 ffmpeg-python
+datasets
+youtube

supabase_fill_table.py

@@ -1,104 +0,0 @@
-import os
-import requests
-import difflib
-from supabase import create_client
-from sentence_transformers import SentenceTransformer
-
-from dotenv import load_dotenv
-
-# Load environment variables from .env file
-load_dotenv()
-
-# Load environment variables
-SUPABASE_URL = os.getenv("SUPABASE_URL")
-SUPABASE_SERVICE_KEY = os.getenv("SUPABASE_SERVICE_KEY")
-
-if not SUPABASE_URL or not SUPABASE_SERVICE_KEY:
-    raise RuntimeError("Please set SUPABASE_URL and SUPABASE_SERVICE_KEY in env")
-
-GAIA_API = "https://agents-course-unit4-scoring.hf.space"
-
-
-
-# Initialize Supabase client and SentenceTransformer model
-supabase = create_client(SUPABASE_URL, SUPABASE_SERVICE_KEY)
-model = SentenceTransformer("all-mpnet-base-v2")
-
-# Local ground-truth mapping
-GROUND_TRUTH = {
-    "What is the capital of Italy?": "Rome",
-    "How much is 2 + 2?": "4",
-    "List only the vegetables from: milk, eggs, broccoli, carrot. Alphabetize.": "broccoli, carrot",
-    "What country had the least number of athletes at the 1928 Summer Olympics? Give IOC code.": "LUX",
-    "Who are the pitchers with the number before and after Taishō Tamai's number as of July 2023? Last names only, comma-separated.": "Lynn, Gilbert",
-    "What is the first name of the only Malko Competition recipient from the 20th Century (after 1977) whose nationality on record is a country that no longer exists?": "Claus",
-    "Given the Excel file at 'test_sales.xlsx', what were total sales for food (not drinks)? Express in USD with two decimal places.": "45.00",
-    "In the video https://www.youtube.com/watch?v=L1vXCYZAYYM, what is the highest number of bird species to be on camera simultaneously?": "270",
-    "What is the surname of the equine veterinarian mentioned in 1.E Exercises from the chemistry materials?": "Louvrier",
-    "Who did the actor who played Ray in the Polish-language version of Everybody Loves Raymond play in Magda M.? First name only.": "Wojciech",
-}
-
-def fetch_gaia_examples():
-    """Fetch GAIA questions from API and pair with ground-truth answers."""
-    try:
-        response = requests.get(f"{GAIA_API}/questions")
-        response.raise_for_status()
-        questions = response.json()  # Assuming the API returns a JSON array of dicts with a 'question' key
-    except requests.RequestException as e:
-        raise RuntimeError(f"Failed to fetch questions from GAIA API: {e}")
-
-    # Show the first 5 questions from the API
-    print("First 5 questions from API:")
-    for q_obj in questions[:5]:
-        question_text = q_obj["question"] if isinstance(q_obj, dict) and "question" in q_obj else q_obj
-        print("-", question_text)
-
-    examples = []
-    for q_obj in questions:
-        # Extract the question string from the dict
-        question_text = q_obj["question"] if isinstance(q_obj, dict) and "question" in q_obj else q_obj
-        # Try exact match first
-        answer = GROUND_TRUTH.get(question_text)
-        # If not found, try fuzzy match
-        if not answer:
-            match = difflib.get_close_matches(question_text, GROUND_TRUTH.keys(), n=1, cutoff=0.8)
-            if match:
-                answer = GROUND_TRUTH[match[0]]
-        if answer:
-            examples.append((question_text, answer))
-        else:
-            print(f"Warning: No ground-truth answer found for question: {question_text}")
-
-    return examples
-
-def main():
-    # Optionally: fetch and print API questions for inspection
-    try:
-        response = requests.get(f"{GAIA_API}/questions")
-        response.raise_for_status()
-        questions = response.json()
-        print("First 5 questions from API:")
-        for q_obj in questions[:5]:
-            question_text = q_obj["question"] if isinstance(q_obj, dict) and "question" in q_obj else q_obj
-            print("-", question_text)
-    except requests.RequestException as e:
-        print(f"Warning: Could not fetch questions from GAIA API: {e}")
-
-    # Insert all ground-truth examples
-    to_insert = []
-    for q, a in GROUND_TRUTH.items():
-        qa = f"Q: {q} A: {a}"
-        emb = model.encode(qa).tolist()
-        to_insert.append({
-            "page_content": qa,
-            "embedding": emb
-        })
-
-    res = supabase.table("documents").insert(to_insert).execute()
-    if res.data:
-        print(f"Inserted {len(to_insert)} GAIA examples from GROUND_TRUTH.")
-    else:
-        print("Error inserting:", res)
-
-if __name__ == "__main__":
-    main()

supabase_fill_table2.py

@@ -0,0 +1,88 @@
+import os
+from supabase import create_client
+from sentence_transformers import SentenceTransformer
+from huggingface_hub import hf_hub_download
+from datasets import load_dataset
+from dotenv import load_dotenv
+
+# -----------------------------------------------------------------------------
+# Load env vars
+# -----------------------------------------------------------------------------
+load_dotenv()
+SUPABASE_URL         = os.getenv("SUPABASE_URL")
+SUPABASE_SERVICE_KEY = os.getenv("SUPABASE_SERVICE_KEY")
+HF_TOKEN             = os.getenv("HUGGINGFACE_API_TOKEN")
+
+if not SUPABASE_URL or not SUPABASE_SERVICE_KEY:
+    raise RuntimeError("Please set SUPABASE_URL and SUPABASE_SERVICE_KEY in your .env")
+
+if not HF_TOKEN:
+    raise RuntimeError(
+        "Please set HUGGINGFACE_API_TOKEN in your .env and ensure you've been granted access to the GAIA dataset."
+    )
+
+# -----------------------------------------------------------------------------
+# Init clients & models
+# -----------------------------------------------------------------------------
+supabase = create_client(SUPABASE_URL, SUPABASE_SERVICE_KEY)
+model     = SentenceTransformer("all-mpnet-base-v2")
+
+# -----------------------------------------------------------------------------
+# GAIA metadata location on HF
+# -----------------------------------------------------------------------------
+GAIA_REPO_ID       = "gaia-benchmark/GAIA"
+GAIA_METADATA_FILE = "2023/validation/metadata.jsonl"
+
+def fetch_gaia_validation_examples():
+    print("🔄 Downloading GAIA metadata.jsonl …")
+    metadata_path = hf_hub_download(
+        repo_id   = GAIA_REPO_ID,
+        filename  = GAIA_METADATA_FILE,
+        token     = HF_TOKEN,
+        repo_type = "dataset",
+    )
+    print(f"✅ Downloaded to {metadata_path!r}")
+
+    print("🔄 Loading JSONL via Datasets …")
+    ds = load_dataset(
+        "json",
+        data_files = metadata_path,
+        split      = "train",
+    )
+    print("Columns in your JSONL:", ds.column_names)
+
+    QUESTION_FIELD = "Question"
+    ANSWER_FIELD   = "Final answer"
+
+    qa = []
+    for row in ds:
+        q = row.get(QUESTION_FIELD)
+        a = row.get(ANSWER_FIELD)
+        if q and a:
+            qa.append((q, a))
+
+    print(f"✅ Found {len(qa)} (Question, Final answer) pairs.")
+    return qa
+
+def main():
+    qa_pairs = fetch_gaia_validation_examples()
+    if not qa_pairs:
+        print("⚠️ No QA pairs—abort.")
+        return
+
+    to_insert = []
+    for q, a in qa_pairs:
+        text = f"Q: {q} A: {a}"
+        emb  = model.encode(text).tolist()
+        to_insert.append({"page_content": text, "embedding": emb})
+
+    print(f"🚀 Inserting {len(to_insert)} records into Supabase…")
+    res = supabase.table("documents").insert(to_insert).execute()
+    if res.data:
+        print(f"🎉 Successfully inserted {len(to_insert)} GAIA examples.")
+    else:
+        print("❌ Insert appeared to fail. Response:")
+        print(res)
+
+if __name__ == "__main__":
+    main()