Update app.py

app.py

@@ -1,4 +1,5 @@
 import os
+import time
 import gradio as gr
 import requests
 import pandas as pd
@@ -8,24 +9,63 @@ import sys
 
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
+# Track API calls for rate limiting
+_last_call_time = 0
+
+def rate_limited_groq(api_key, prompt, system="", max_tokens=128):
+    """Call Groq with rate limiting — max 25 req/min to stay safe."""
+    global _last_call_time
+    # Ensure at least 2.5 seconds between calls (= 24/min, safely under 30 limit)
+    elapsed = time.time() - _last_call_time
+    if elapsed < 2.5:
+        time.sleep(2.5 - elapsed)
+    _last_call_time = time.time()
+
+    url = "https://api.groq.com/openai/v1/chat/completions"
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    msgs = []
+    if system:
+        msgs.append({"role": "system", "content": system})
+    msgs.append({"role": "user", "content": prompt})
+    body = {"model": "llama-3.3-70b-versatile", "messages": msgs,
+            "temperature": 0.0, "max_tokens": max_tokens}
+    resp = requests.post(url, headers=headers, json=body, timeout=60)
+    if resp.status_code == 429:
+        print("Rate limited! Waiting 60s...")
+        time.sleep(60)
+        resp = requests.post(url, headers=headers, json=body, timeout=60)
+    if resp.status_code != 200:
+        raise Exception(f"Groq {resp.status_code}: {resp.text[:200]}")
+    return resp.json()["choices"][0]["message"]["content"].strip()
+
 def download_task_file(task_id):
     url = f"{DEFAULT_API_URL}/files/{task_id}"
     try:
         resp = requests.get(url, timeout=30)
-        if resp.status_code != 200:
+        print(f"  File request: HTTP {resp.status_code}, size={len(resp.content)}, "
+              f"content-type={resp.headers.get('content-type','?')}")
+        if resp.status_code != 200 or len(resp.content) == 0:
             return None, None
         cd = resp.headers.get("content-disposition", "")
+        ct = resp.headers.get("content-type", "")
         fname = "task_file"
         if "filename=" in cd:
             fname = cd.split("filename=")[-1].strip().strip('"').strip("'")
-        ext = os.path.splitext(fname)[-1] or ".bin"
-        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=ext)
+        ext = os.path.splitext(fname)[-1]
+        if not ext:
+            if "python" in ct: ext = ".py"
+            elif "excel" in ct or "spreadsheet" in ct: ext = ".xlsx"
+            elif "csv" in ct: ext = ".csv"
+            elif "image" in ct: ext = ".png"
+            else: ext = ".bin"
+            fname += ext
+        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=ext, prefix="gaia_")
         tmp.write(resp.content)
         tmp.close()
-        print(f"Downloaded: {fname} ({len(resp.content)} bytes)")
+        print(f"  Saved: {fname} -> {tmp.name}")
         return tmp.name, fname
     except Exception as e:
-        print(f"Download error: {e}")
+        print(f"  Download error: {e}")
         return None, None
 
 def read_file_contents(local_path, fname):
@@ -38,14 +78,15 @@ def read_file_contents(local_path, fname):
             df = pd.read_csv(local_path)
             return f"CSV shape={df.shape}\nColumns={list(df.columns)}\n\n{df.to_string()}"
         elif ext in (".py", ".txt", ".md", ".json"):
-            with open(local_path) as f:
+            with open(local_path, "r", errors="replace") as f:
                 return f.read()
         else:
             try:
-                with open(local_path) as f:
-                    return f.read()
-            except:
-                return f"Binary: {fname}"
+                with open(local_path, "r", errors="replace") as f:
+                    c = f.read()
+                    if c.strip(): return c
+            except: pass
+            return f"Binary: {fname}"
     except Exception as e:
         return f"Error: {e}"
 
@@ -54,8 +95,8 @@ def run_python_file(local_path):
         r = subprocess.run([sys.executable, local_path],
                            capture_output=True, text=True, timeout=15)
         out = (r.stdout + r.stderr).strip()
-        print(f"Python out: '{out[:200]}'")
-        return out or "No output."
+        print(f"  Python output: '{out[:200]}'")
+        return out if out else "No output."
     except Exception as e:
         return f"Error: {e}"
 
@@ -67,20 +108,6 @@ def clean_answer(text):
             text = text[len(p):].strip()
     return text.split("\n")[0].strip().strip('"').strip("'").strip("*").strip()
 
-def call_groq(api_key, prompt, system="", max_tokens=128):
-    url = "https://api.groq.com/openai/v1/chat/completions"
-    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
-    msgs = []
-    if system:
-        msgs.append({"role": "system", "content": system})
-    msgs.append({"role": "user", "content": prompt})
-    body = {"model": "llama-3.3-70b-versatile", "messages": msgs,
-            "temperature": 0.0, "max_tokens": max_tokens}
-    resp = requests.post(url, headers=headers, json=body, timeout=60)
-    if resp.status_code != 200:
-        raise Exception(f"Groq {resp.status_code}: {resp.text[:200]}")
-    return resp.json()["choices"][0]["message"]["content"].strip()
-
 def search_web(query, max_results=6):
     try:
         from duckduckgo_search import DDGS
@@ -99,7 +126,7 @@ def test_api():
     if not key:
         return "❌ GROQ_API_KEY not set!"
     try:
-        ans = call_groq(key, "What is 2+2?", "Reply with only the number.")
+        ans = rate_limited_groq(key, "What is 2+2?", "Reply with only the number.")
         return f"✅ Groq working! Test: '{ans}'"
     except Exception as e:
         return f"❌ {e}"
@@ -112,31 +139,26 @@ class BasicAgent:
     def __init__(self):
         self.key = os.getenv("GROQ_API_KEY", "")
         if not self.key:
-            raise RuntimeError("GROQ_API_KEY not set! Add in Space Settings → Secrets.")
+            raise RuntimeError("GROQ_API_KEY not set!")
         print(f"Agent ready. Key: {self.key[:8]}...")
 
     def ask(self, prompt, max_tokens=128):
-        return clean_answer(call_groq(self.key, prompt, SYSTEM, max_tokens))
-
-    def __call__(self, question: str) -> str:
-        task_id = ""
-        if question.startswith("[TASK_ID:"):
-            end = question.index("]")
-            task_id = question[9:end]
-            question = question[end+1:].strip()
+        return clean_answer(rate_limited_groq(self.key, prompt, SYSTEM, max_tokens))
 
+    def __call__(self, question: str, task_id: str = "") -> str:
         print(f"\n{'='*50}\nTask: {task_id}\nQ: {question[:200]}")
 
         # Handle reversed text
         if "rewsna" in question or "dnatsrednu" in question:
             question = question[::-1]
-            print(f"Reversed: {question}")
+            print(f"  Reversed: {question}")
 
         file_ctx = ""
         is_py = False
 
         # Download file
         if task_id:
+            print(f"  Attempting file download for task_id={task_id}")
             lp, fn = download_task_file(task_id)
             if lp and fn:
                 ext = os.path.splitext(fn)[-1].lower()
@@ -149,67 +171,69 @@ class BasicAgent:
                     contents = read_file_contents(lp, fn)
                     file_ctx = f"\n[File: {fn}]\n{contents[:6000]}\n"
                 elif ext in (".png", ".jpg", ".jpeg"):
-                    file_ctx = f"\n[Image: {fn} attached — answer from your knowledge.]\n"
+                    file_ctx = f"\n[Image: {fn} attached.]\n"
                 else:
                     contents = read_file_contents(lp, fn)
                     file_ctx = f"\n[File: {fn}]\n{contents[:4000]}\n"
+            else:
+                print(f"  No file found for this task.")
 
-        # Web search (skip for python — output is the answer)
+        # Web search
         search_ctx = ""
         if not is_py:
-            r1 = search_web(question[:200])
-            search_ctx = f"\n[Search results]\n{r1[:3500]}\n"
+            results = search_web(question[:200])
+            if results and "error" not in results.lower():
+                search_ctx = f"\n[Search]\n{results[:3500]}\n"
 
-        # Per-question format instructions
-        q_lower = question.lower()
+        # Format hints
+        q = question.lower()
         fmt = ""
-        if "studio album" in q_lower:
-            fmt = "\nIMPORTANT: Count ONLY solo studio albums (not live, compilation, or collaborative albums). Answer with a single integer."
-        elif "first name" in q_lower:
-            fmt = "\nIMPORTANT: Reply with the first name only."
-        elif "surname" in q_lower or "last name" in q_lower:
-            fmt = "\nIMPORTANT: Reply with the surname/last name only."
-        elif "how many" in q_lower and ("at bat" in q_lower or "walk" in q_lower):
-            fmt = "\nIMPORTANT: Reply with a single integer only."
-        elif "country" in q_lower and "olympic" in q_lower:
-            fmt = "\nIMPORTANT: Reply with the country name only. If tied, give alphabetically first country."
-        elif "excel" in q_lower or ("total" in q_lower and "sale" in q_lower):
-            fmt = "\nIMPORTANT: Plain number only. No $ sign, no commas (e.g. 12345.67)."
-        elif "chess" in q_lower:
-            fmt = "\nIMPORTANT: Reply with chess move in standard notation only (e.g. Qd8)."
-        elif "pitcher" in q_lower and "number" in q_lower:
-            fmt = "\nIMPORTANT: Reply with two last names comma-separated, in jersey number order (lower number first)."
-        elif "wikipedia" in q_lower and "nominat" in q_lower:
-            fmt = "\nIMPORTANT: Reply with the Wikipedia username only."
-        elif ("grocery" in q_lower or "shopping" in q_lower) and "list" in q_lower:
-            fmt = "\nIMPORTANT: Reply with comma-separated items in alphabetical order, all lowercase, using the scientific/botanical names the botany professor would use."
-        elif "youtube" in q_lower or "video" in q_lower:
-            fmt = "\nIMPORTANT: Reply with the exact short answer requested — a number, name, or brief quote."
-        elif "grant" in q_lower or "award" in q_lower or "number" in q_lower:
-            fmt = "\nIMPORTANT: Reply with the exact identifier/number only."
+        if "studio album" in q:
+            fmt = "\nCount only SOLO studio albums (exclude collaborative albums). Single integer answer."
+        elif "first name" in q:
+            fmt = "\nFirst name only."
+        elif "surname" in q or "last name" in q:
+            fmt = "\nSurname only."
+        elif "at bat" in q or "at-bat" in q:
+            fmt = "\nSingle integer only."
+        elif "how many" in q:
+            fmt = "\nSingle integer only."
+        elif "ioc" in q or ("country" in q and "olympic" in q):
+            fmt = "\nIOC country code only (3 letters, e.g. USA, GBR). If tied, alphabetically first."
+        elif "excel" in q or ("sale" in q and "food" in q):
+            fmt = "\nUSD with two decimal places (e.g. 89.50). No $ sign."
+        elif "chess" in q:
+            fmt = "\nChess move in algebraic notation only."
+        elif "pitcher" in q and "number" in q:
+            fmt = "\nTwo last names, comma-separated, pitcher with lower jersey number first."
+        elif "wikipedia" in q and "nominat" in q:
+            fmt = "\nWikipedia username only."
+        elif "grocery" in q or ("shopping" in q and "list" in q):
+            fmt = "\nComma-separated list, alphabetical order."
+        elif "youtube" in q or "video" in q:
+            fmt = "\nExact short answer only — quote, number, or name."
+        elif "grant" in q or "award number" in q:
+            fmt = "\nExact identifier only."
 
         prompt = (
             f"Question: {question}"
             f"{file_ctx}"
             f"{search_ctx}"
             f"{fmt}"
-            "\n\nGive ONLY the final answer. Nothing else."
+            "\n\nGive ONLY the final answer."
         )
 
         try:
             answer = self.ask(prompt, max_tokens=64)
-            # If still too long, shorten
             if len(answer.split()) > 20:
-                answer = clean_answer(call_groq(
+                answer = clean_answer(rate_limited_groq(
                     self.key,
-                    f"Extract only the shortest final answer (name/number/phrase) from:\n{answer}",
-                    "Reply with only the bare answer.",
-                    max_tokens=32
-                ))
-            print(f"Final: '{answer}'")
+                    f"Extract only the shortest final answer from:\n{answer}",
+                    "Reply with only the bare answer.", max_tokens=32))
+            print(f"  Final: '{answer}'")
             return answer
         except Exception as e:
-            print(f"Error: {e}")
+            print(f"  Error: {e}")
             return ""
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
@@ -223,7 +247,6 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         return f"❌ {e}", None
 
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
-
     try:
         resp = requests.get(f"{DEFAULT_API_URL}/questions", timeout=15)
         resp.raise_for_status()
@@ -233,15 +256,15 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         return f"Error: {e}", None
 
     results_log, answers_payload = [], []
-
     for i, item in enumerate(questions_data):
-        task_id = item.get("task_id")
+        task_id = item.get("task_id", "")
         question_text = item.get("question")
         if not task_id or question_text is None:
             continue
         print(f"\n[{i+1}/{len(questions_data)}]")
         try:
-            ans = agent(f"[TASK_ID:{task_id}] {question_text}")
+            # Pass task_id directly — no injection needed
+            ans = agent(question_text, task_id=task_id)
         except Exception as e:
             ans = ""
         answers_payload.append({"task_id": task_id, "submitted_answer": ans})