Update app.py

app.py

@@ -1,84 +1,391 @@
 import os
 import re
+import base64
+import traceback
+import requests
+import gradio as gr
+import pandas as pd
 from google import genai
+from google.genai import types
+
+# --- Constants ---
+DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
+
+
+# ── Utilities ─────────────────────────────────────────────────────────
+
+def clean_answer(text: str) -> str:
+    """Normalize model output for exact-match scoring."""
+    if not text:
+        return ""
+    text = text.strip()
+    # Remove markdown fences
+    text = re.sub(r"^```[a-zA-Z]*\s*", "", text)
+    text = re.sub(r"\s*```$", "", text)
+    # Remove common label prefixes
+    for prefix in [
+        "final answer:", "answer:", "the answer is:",
+        "the final answer is:", "result:", "output:",
+    ]:
+        if text.lower().startswith(prefix):
+            text = text[len(prefix):].strip()
+            break
+    # Collapse whitespace, cap length
+    return " ".join(text.split())[:300]
+
+
+def extract_text(response) -> str:
+    """
+    Extract the final answer text from a GenerateContentResponse.
+    Works for plain text, google_search grounding, and code_execution results.
+    Always returns the LAST meaningful text part (= answer after tool use).
+    """
+    parts_text = []
+    try:
+        for candidate in response.candidates:
+            for part in candidate.content.parts:
+                if hasattr(part, "text") and part.text and part.text.strip():
+                    parts_text.append(part.text.strip())
+                if hasattr(part, "code_execution_result") and part.code_execution_result:
+                    out = getattr(part.code_execution_result, "output", "")
+                    if out and str(out).strip():
+                        parts_text.append(str(out).strip())
+    except Exception:
+        pass
+    if parts_text:
+        return parts_text[-1]
+    try:
+        return (response.text or "").strip()
+    except Exception:
+        return ""
+
+
+def fetch_task_file(task_id: str) -> tuple:
+    """Download file attached to a GAIA task. Returns (bytes, filename)."""
+    try:
+        r = requests.get(f"{DEFAULT_API_URL}/files/{task_id}", timeout=20)
+        if r.status_code == 200 and r.content:
+            cd   = r.headers.get("Content-Disposition", "")
+            name = ""
+            if "filename=" in cd:
+                name = cd.split("filename=")[-1].strip().strip('"')
+            name = name or f"file_{task_id}"
+            print(f"  [file] {name}  ({len(r.content)} bytes)")
+            return r.content, name
+    except Exception as e:
+        print(f"  [file] fetch error: {e}")
+    return None, ""
+
+
+def build_contents(question: str, file_bytes, fname: str) -> list:
+    """Package question + optional file into Gemini contents list."""
+    if file_bytes is None:
+        return [question]
+    ext = fname.rsplit(".", 1)[-1].lower() if "." in fname else ""
+
+    IMAGE_MIME = {"png":"image/png","jpg":"image/jpeg","jpeg":"image/jpeg",
+                  "gif":"image/gif","webp":"image/webp","bmp":"image/bmp"}
+    AUDIO_MIME = {"mp3":"audio/mpeg","wav":"audio/wav","ogg":"audio/ogg",
+                  "flac":"audio/flac","m4a":"audio/mp4"}
+
+    if ext in IMAGE_MIME:
+        return [types.Part.from_bytes(data=file_bytes, mime_type=IMAGE_MIME[ext]), question]
+    if ext == "pdf":
+        return [types.Part.from_bytes(data=file_bytes, mime_type="application/pdf"), question]
+    if ext in AUDIO_MIME:
+        return [types.Part.from_bytes(data=file_bytes, mime_type=AUDIO_MIME[ext]), question]
+
+    # Text-based files: embed as context
+    try:
+        txt = file_bytes.decode("utf-8", errors="replace")
+        return [f"Attached file ({fname}):\n```\n{txt[:12000]}\n```\n\n{question}"]
+    except Exception:
+        b64 = base64.b64encode(file_bytes).decode()
+        return [f"Attached file ({fname}) base64:\n{b64[:2000]}\n\n{question}"]
+
+
+# ── Agent ─────────────────────────────────────────────────────────────
 
 class BasicAgent:
+    """
+    Gemini 2.0 Flash agent with Google Search grounding, code execution,
+    and file-attachment support for GAIA benchmark questions.
+    """
+
+    SYSTEM = """You are a precise expert assistant solving GAIA benchmark evaluation questions.
+
+CRITICAL OUTPUT RULE:
+Output ONLY the final answer — nothing else.
+No explanation, no reasoning, no preamble, no trailing sentence.
+
+FORMAT RULES:
+- Numbers : digits only, no thousand-separators. Drop .0 from whole numbers.
+- Lists   : comma-separated values, alphabetical order unless otherwise specified.
+- Yes/No  : exactly "yes" or "no" (lowercase).
+- Names   : full name unless only first or last is requested.
+- Dates   : match the format implied by the question.
+- Units   : include units only if the question asks for them.
+
+STRATEGY:
+1. Read the question (and any attached file) carefully.
+2. Use Google Search for any fact, date, count, name, or external data you need.
+3. Use code execution for arithmetic, unit conversion, or data analysis.
+4. Think step-by-step internally.
+5. Output ONLY the single final answer."""
+
+    CODE_KEYWORDS = {
+        "calculate","compute","sum","total","average","mean","median",
+        "percentage","multiply","divide","convert","how many","count",
+        "square root","power","factorial","modulo","remainder",
+    }
 
     def __init__(self):
         api_key = os.getenv("GEMINI_API_KEY")
         if not api_key:
-            raise ValueError("Missing GEMINI_API_KEY")
+            raise EnvironmentError("GEMINI_API_KEY environment variable is not set.")
+
+        self.client   = genai.Client(api_key=api_key)
+        self.model    = "gemini-2.0-flash"
 
-        genai.configure(api_key=api_key)
+        cfg = lambda tools: types.GenerateContentConfig(
+            system_instruction=self.SYSTEM,
+            tools=tools,
+            temperature=0,
+        )
 
-        self.model = genai.GenerativeModel("gemini-1.5-flash")
+        self.search_cfg = cfg([types.Tool(google_search=types.GoogleSearch())])
+        self.code_cfg   = cfg([types.Tool(code_execution=types.ToolCodeExecution())])
+        self.plain_cfg  = cfg([])   # for file questions (model reads the file itself)
 
-        print("✅ PASS Agent Ready")
+        print(f"BasicAgent initialized (model={self.model})")
 
-    def clean(self, text: str) -> str:
-        if not text:
-            return "N/A"
+    def __call__(self, question: str) -> str:
+        # task_id is embedded in the question only when called from run_and_submit_all;
+        # we extract it via a side-channel attribute set before each call.
+        task_id = getattr(self, "_current_task_id", "")
+        try:
+            return self._run(question, task_id)
+        except Exception:
+            print(traceback.format_exc())
+            # Last-resort plain call — never return empty
+            try:
+                r = self.client.models.generate_content(
+                    model=self.model, contents=question, config=self.plain_cfg)
+                ans = clean_answer(extract_text(r))
+                return ans if ans else self._forced_answer(question)
+            except Exception:
+                return self._forced_answer(question)
+
+    def _run(self, question: str, task_id: str) -> str:
+        file_bytes, fname = fetch_task_file(task_id) if task_id else (None, "")
+        contents          = build_contents(question, file_bytes, fname)
+        q_lower           = question.lower()
+
+        has_file   = file_bytes is not None
+        needs_code = not has_file and any(kw in q_lower for kw in self.CODE_KEYWORDS)
 
-        text = text.strip()
+        config = self.plain_cfg if has_file else (
+                 self.code_cfg  if needs_code else
+                 self.search_cfg)
 
-        # remove junk
-        text = text.replace("Final Answer:", "")
-        text = text.replace("Answer:", "")
-        text = text.strip()
+        resp = self.client.models.generate_content(
+            model=self.model, contents=contents, config=config)
+        raw  = extract_text(resp)
+        ans  = clean_answer(raw)
+        print(f"  raw : {raw[:120]!r}")
+        print(f"  ans : {ans!r}")
 
-        # keep only first line
-        text = text.split("\n")[0]
+        # If empty, retry with search
+        if not ans:
+            resp2 = self.client.models.generate_content(
+                model=self.model, contents=contents, config=self.search_cfg)
+            ans = clean_answer(extract_text(resp2))
+            print(f"  retry: {ans!r}")
 
-        # remove sentences
-        text = text.split(".")[0]
+        # Still empty → force a plain answer (never return blank)
+        if not ans:
+            ans = self._forced_answer(question)
 
-        return text.strip()[:100]
+        return ans
 
-    def __call__(self, question: str, task_id: str = "") -> str:
+    def _forced_answer(self, question: str) -> str:
+        """Absolute last resort — plain call with no tools, no format rules."""
         try:
-            # 🔥 STEP 1 — Solve
-            prompt1 = f"""
-Solve the question.
+            r = self.client.models.generate_content(
+                model=self.model,
+                contents=f"Answer in one word or number only:\n{question}",
+                config=types.GenerateContentConfig(temperature=0),
+            )
+            ans = clean_answer(extract_text(r))
+            return ans if ans else "unknown"
+        except Exception:
+            return "unknown"
 
-IMPORTANT:
-- Think step by step internally
-- Return ONLY final answer
-- No explanation
 
-Question:
-{question}
-"""
-            r1 = self.model.generate_content(prompt1)
-            ans1 = r1.text if hasattr(r1, "text") else ""
+# ── Gradio runner (original HF template structure preserved) ──────────
 
-            ans1 = self.clean(ans1)
+def run_and_submit_all(profile: gr.OAuthProfile | None):
+    """
+    Fetches all questions, runs the BasicAgent on them, submits all answers,
+    and displays the results.
+    """
+    space_id = os.getenv("SPACE_ID")
 
-            # 🔥 STEP 2 — VERIFY (this is the magic)
-            prompt2 = f"""
-Question: {question}
+    if profile:
+        username = f"{profile.username}"
+        print(f"User logged in: {username}")
+    else:
+        print("User not logged in.")
+        return "Please Login to Hugging Face with the button.", None
 
-Proposed Answer: {ans1}
+    api_url       = DEFAULT_API_URL
+    questions_url = f"{api_url}/questions"
+    submit_url    = f"{api_url}/submit"
 
-Check if this is correct.
-If wrong, fix it.
+    # 1. Instantiate Agent
+    try:
+        agent = BasicAgent()
+    except Exception as e:
+        print(f"Error instantiating agent: {e}")
+        return f"Error initializing agent: {e}", None
 
-Return ONLY final answer.
-No explanation.
-"""
-            r2 = self.model.generate_content(prompt2)
-            final = r2.text if hasattr(r2, "text") else ""
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
+    print(agent_code)
 
-            final = self.clean(final)
+    # 2. Fetch Questions
+    print(f"Fetching questions from: {questions_url}")
+    try:
+        response = requests.get(questions_url, timeout=15)
+        response.raise_for_status()
+        questions_data = response.json()
+        if not questions_data:
+            return "Fetched questions list is empty or invalid format.", None
+        print(f"Fetched {len(questions_data)} questions.")
+    except requests.exceptions.RequestException as e:
+        return f"Error fetching questions: {e}", None
+    except Exception as e:
+        return f"An unexpected error occurred fetching questions: {e}", None
 
-            # 🔥 fallback if broken
-            if not final or final == "N/A":
-                final = ans1
+    # 3. Run Agent
+    results_log     = []
+    answers_payload = []
+    print(f"Running agent on {len(questions_data)} questions...")
 
-            print("Q:", question[:80])
-            print("A:", final)
+    for item in questions_data:
+        task_id       = item.get("task_id")
+        question_text = item.get("question")
+        if not task_id or question_text is None:
+            print(f"Skipping item with missing task_id or question: {item}")
+            continue
 
-            return final
+        print(f"\n[{task_id}]  {question_text[:120]}")
+        agent._current_task_id = task_id   # pass task_id for file download
 
+        try:
+            submitted_answer = agent(question_text)
         except Exception as e:
-            print("ERROR:", e)
-            return "N/A"
+            submitted_answer = "unknown"
+            print(f"Error running agent on task {task_id}: {e}")
+
+        print(f"  → submitted: {submitted_answer!r}")
+        answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
+        results_log.append({
+            "Task ID":          task_id,
+            "Question":         question_text,
+            "Submitted Answer": submitted_answer,
+        })
+
+    if not answers_payload:
+        return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
+
+    # 4. Prepare Submission
+    submission_data = {
+        "username":   username.strip(),
+        "agent_code": agent_code,
+        "answers":    answers_payload,
+    }
+    print(f"Submitting {len(answers_payload)} answers for user '{username}'...")
+
+    # 5. Submit
+    try:
+        response = requests.post(submit_url, json=submission_data, timeout=60)
+        response.raise_for_status()
+        result_data = response.json()
+        final_status = (
+            f"Submission Successful!\n"
+            f"User: {result_data.get('username')}\n"
+            f"Overall Score: {result_data.get('score', 'N/A')}% "
+            f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
+            f"Message: {result_data.get('message', 'No message received.')}"
+        )
+        print("Submission successful.")
+        return final_status, pd.DataFrame(results_log)
+    except requests.exceptions.HTTPError as e:
+        error_detail = f"Server responded with status {e.response.status_code}."
+        try:
+            error_json = e.response.json()
+            error_detail += f" Detail: {error_json.get('detail', e.response.text)}"
+        except requests.exceptions.JSONDecodeError:
+            error_detail += f" Response: {e.response.text[:500]}"
+        status_message = f"Submission Failed: {error_detail}"
+        print(status_message)
+        return status_message, pd.DataFrame(results_log)
+    except requests.exceptions.Timeout:
+        return "Submission Failed: The request timed out.", pd.DataFrame(results_log)
+    except requests.exceptions.RequestException as e:
+        return f"Submission Failed: Network error - {e}", pd.DataFrame(results_log)
+    except Exception as e:
+        return f"An unexpected error occurred during submission: {e}", pd.DataFrame(results_log)
+
+
+# --- Build Gradio Interface using Blocks ---
+with gr.Blocks() as demo:
+    gr.Markdown("# Basic Agent Evaluation Runner")
+    gr.Markdown(
+        """
+        **Instructions:**
+
+        1.  Please clone this space, then modify the code to define your agent's logic, the tools, the necessary packages, etc ...
+        2.  Log in to your Hugging Face account using the button below. This uses your HF username for submission.
+        3.  Click 'Run Evaluation & Submit All Answers' to fetch questions, run your agent, submit answers, and see the score.
+
+        ---
+        **Disclaimers:**
+        Once clicking on the "submit button, it can take quite some time ( this is the time for the agent to go through all the questions).
+        This space provides a basic setup and is intentionally sub-optimal to encourage you to develop your own, more robust solution. For instance for the delay process of the submit button, a solution could be to cache the answers and submit in a seperate action or even to answer the questions in async.
+        """
+    )
+
+    gr.LoginButton()
+
+    run_button = gr.Button("Run Evaluation & Submit All Answers")
+
+    status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
+    results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
+
+    run_button.click(
+        fn=run_and_submit_all,
+        outputs=[status_output, results_table]
+    )
+
+if __name__ == "__main__":
+    print("\n" + "-"*30 + " App Starting " + "-"*30)
+    space_host_startup = os.getenv("SPACE_HOST")
+    space_id_startup   = os.getenv("SPACE_ID")
+
+    if space_host_startup:
+        print(f"✅ SPACE_HOST found: {space_host_startup}")
+        print(f"   Runtime URL should be: https://{space_host_startup}.hf.space")
+    else:
+        print("ℹ️  SPACE_HOST environment variable not found (running locally?).")
+
+    if space_id_startup:
+        print(f"✅ SPACE_ID found: {space_id_startup}")
+        print(f"   Repo URL: https://huggingface.co/spaces/{space_id_startup}")
+        print(f"   Repo Tree URL: https://huggingface.co/spaces/{space_id_startup}/tree/main")
+    else:
+        print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
+
+    print("-"*(60 + len(" App Starting ")) + "\n")
+    print("Launching Gradio Interface for Basic Agent Evaluation...")
+    demo.launch(debug=True, share=False)