Update agent.py

agent.py

@@ -1,173 +1,65 @@
-"""LangGraph Agent – retry 5s, 30s, 60s • no Supabase"""
-
-import os
-import time
-from dotenv import load_dotenv
-from langgraph.graph import START, StateGraph, MessagesState
-from langgraph.prebuilt import ToolNode, tools_condition
-
-# LLM providers
-from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain_groq import ChatGroq
-from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
-
-# Tools & loaders
-from langchain_community.tools.tavily_search import TavilySearchResults
-from langchain_community.document_loaders import WikipediaLoader, ArxivLoader
-from langchain_core.messages import SystemMessage, HumanMessage
-from langchain_core.tools import tool
-
-load_dotenv()
-
-# --------------------------------------------------------------------------- #
-#  TOOL DEFINITIONS                                                           #
-# --------------------------------------------------------------------------- #
-@tool
-def multiply(a: int, b: int) -> int:
-    """Multiply two integers and return the product."""
-    return a * b
-
-@tool
-def add(a: int, b: int) -> int:
-    """Add two integers and return the sum."""
-    return a + b
-
-@tool
-def subtract(a: int, b: int) -> int:
-    """Subtract the second integer from the first and return the difference."""
-    return a - b
-
-@tool
-def divide(a: int, b: int) -> float:
-    """Divide a by b and return the quotient (error if b == 0)."""
-    if b == 0:
-        raise ValueError("Cannot divide by zero.")
-    return a / b
-
-@tool
-def modulus(a: int, b: int) -> int:
-    """Return the remainder of the division of a by b."""
-    return a % b
-
-@tool
-def wiki_search(query: str) -> str:
-    """Search Wikipedia (max 2 docs) and return formatted content."""
-    docs = WikipediaLoader(query=query, load_max_docs=2).load()
-    return "\n\n---\n\n".join(
-        f'<Document source="{d.metadata["source"]}" page="{d.metadata.get("page","")}"/>\n'
-        f"{d.page_content}\n</Document>"
-        for d in docs
-    )
-
-@tool
-def web_search(query: str) -> str:
-    """Perform a web search with Tavily (max 3 docs) and return formatted content."""
-    docs = TavilySearchResults(max_results=3).invoke(query=query)
-    return "\n\n---\n\n".join(
-        f'<Document source="{d.metadata["source"]}" page="{d.metadata.get("page","")}"/>\n'
-        f"{d.page_content}\n</Document>"
-        for d in docs
-    )
-
-@tool
-def arxiv_search(query: str) -> str:
-    """Search ArXiv (max 3 docs) and return first 1000 characters per paper."""
-    docs = ArxivLoader(query=query, load_max_docs=3).load()
-    return "\n\n---\n\n".join(
-        f'<Document source="{d.metadata["source"]}" page="{d.metadata.get("page","")}"/>\n'
-        f"{d.page_content[:1000]}\n</Document>"
-        for d in docs
+"""
+GAIA Level-1 agent powered by smolagents.
+
+* Planner/esecutore: CodeAgent (smolagents)
+* LLM backend      : GPT-4.1 via OpenAI
+* Tools            : DuckDuckGo (builtin), WikipediaTool, ArxivTool
+* Output           : UNA sola riga (exact-match)
+"""
+from __future__ import annotations
+import os, textwrap
+from smolagents import CodeAgent, DuckDuckGoSearchTool, OpenAIModel
+from tools import WikipediaTool, ArxivTool
+import openai
+
+# ─── API key check ──────────────────────────────────────────────
+openai.api_key = os.getenv("OPENAI_API_KEY") or ""
+if not openai.api_key:
+    raise EnvironmentError(
+        "OPENAI_API_KEY non impostata: aggiungila nei Secrets dello Space "
+        "o in un file .env locale."
     )
 
-# --------------------------------------------------------------------------- #
-#  System prompt                                                              #
-# --------------------------------------------------------------------------- #
-with open("system_prompt.txt", "r", encoding="utf-8") as f:
-    system_prompt = f.read()
-sys_msg = SystemMessage(content=system_prompt)
+# ─── Prompt di sistema rigido (exact-match) ─────────────────────
+SYSTEM_PROMPT = textwrap.dedent("""
+You are a helpful assistant tasked with answering questions using a set of tools.
+Your final answer should be a number OR as few words as possible OR a comma separated list of numbers and/or strings.
+If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise.
+If you are asked for a string, don't use articles, neither abbreviations, and write digits in plain text unless specified otherwise.
+Return ONLY the final answer line.
+""").strip()
+
+# ─── Costruzione del “core” CodeAgent ───────────────────────────
+model = OpenAIModel(
+    model_id="gpt-4.1",
+    temperature=0,
+    system_prompt=SYSTEM_PROMPT
+)
 
 tools = [
-    multiply, add, subtract, divide, modulus,
-    wiki_search, web_search, arxiv_search,
+    DuckDuckGoSearchTool(),  # incorporato in smolagents
+    WikipediaTool(),
+    ArxivTool(),
 ]
 
-# --------------------------------------------------------------------------- #
-#  Retry parameters                                                           #
-# --------------------------------------------------------------------------- #
-RETRY_DELAYS = [0, 5, 30, 60]   # seconds for attempts 0-3
-MAX_ATTEMPTS = len(RETRY_DELAYS)
-
-# --------------------------------------------------------------------------- #
-#  Build LangGraph                                                            #
-# --------------------------------------------------------------------------- #
-def build_graph(provider: str = "groq"):
-    """Return a LangGraph graph with explicit retry logic."""
-
-    # ----------- LLM selection -------------------------------------------- #
-    if provider == "google":
-        llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0)
-
-    elif provider == "groq":
-        llm = ChatGroq(
-            model="qwen-qwq-32b",
-            temperature=0,
-            max_retries=0,  # we handle retries manually
-        )
-
-    elif provider == "huggingface":
-        llm = ChatHuggingFace(
-            llm=HuggingFaceEndpoint(
-                url="https://api-inference.huggingface.co/models/"
-                    "Meta-DeepLearning/llama-2-7b-chat-hf",
-                temperature=0,
-            )
-        )
-    else:
-        raise ValueError("Invalid provider. Choose 'google', 'groq' or 'huggingface'.")
-
-    llm_with_tools = llm.bind_tools(tools)
-
-    # ---------------- Retry wrapper -------------------------------------- #
-    def invoke_with_retry(messages):
-        last_err = None
-        for attempt, delay in enumerate(RETRY_DELAYS):
-            if delay:
-                print(f"[Retry {attempt}/{MAX_ATTEMPTS-1}] waiting {delay}s")
-                time.sleep(delay)
-            try:
-                return llm_with_tools.invoke(messages)
-            except Exception as e:
-                if ("503" in str(e) or "Service Unavailable" in str(e) or "429" in str(e)) and attempt < MAX_ATTEMPTS - 1:
-                    last_err = e
-                    continue
-                raise
-        raise last_err or RuntimeError("Unknown error during LLM invocation")
-
-    # ---------------- Nodes ---------------------------------------------- #
-    def assistant(state: MessagesState):
-        messages = [sys_msg] + state["messages"]
-        return {"messages": [invoke_with_retry(messages)]}
-
-    # ---------------- Graph ---------------------------------------------- #
-    builder = StateGraph(MessagesState)
-    builder.add_node("assistant", assistant)
-    builder.add_node("tools", ToolNode(tools))
-
-    builder.add_edge(START, "assistant")
-    builder.add_conditional_edges("assistant", tools_condition)
-    builder.add_edge("tools", "assistant")
-
-    return builder.compile()
-
-
-# --------------------------------------------------------------------------- #
-#  Stand-alone test                                                           #
-# --------------------------------------------------------------------------- #
-if __name__ == "__main__":
-    g = build_graph(provider="groq")
-    q = ("When was a picture of St. Thomas Aquinas first added to the Wikipedia "
-         "page on the Principle of double effect?")
-    msgs = [HumanMessage(content=q)]
-    res = g.invoke({"messages": msgs})
-    for m in res["messages"]:
-        m.pretty_print()
+core_agent = CodeAgent(
+    model=model,
+    tools=tools,
+    max_steps=6,          # previene loop infiniti
+    scratchpad="minimal"  # log conciso
+)
+
+# ─── Thin wrapper usato da app.py ───────────────────────────────
+class BasicAgent:  # (mantiene lo stesso nome già importato in app.py)
+    def __init__(self):
+        print("✅ smolagents BasicAgent inizializzato")
+
+    def __call__(self, question: str) -> str:
+        """
+        Esegue CodeAgent e restituisce SOLO la prima riga,
+        così il grader riceve una stringa exact-match.
+        """
+        raw_answer: str = core_agent.run(question)
+        answer = raw_answer.strip().split("\n", 1)[0]
+        print(f"[ANSWER] {answer}")
+        return answer