Update app.py

app.py

@@ -2,17 +2,70 @@ import os
 import gradio as gr
 import requests
 import pandas as pd
-from typing import Optional, Dict, Any
+import time
+import random
+import re
+import math
+from typing import Dict, Any, List, Optional
 
+# (Keep Constants as is)
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
+# --- Provider-agnostic lightweight LLM client (no HF Inference API) ---
 
 class LLMClient:
-    # ... keep your existing __init__ ...
+    """
+    Pluggable LLM client with robust 429/5xx backoff.
+    Supported providers via env LLM_PROVIDER:
+      - openai       (OPENAI_API_KEY, model via OPENAI_MODEL, base via OPENAI_BASE_URL)
+      - together     (TOGETHER_API_KEY, model via TOGETHER_MODEL, base via TOGETHER_BASE_URL)
+      - openrouter   (OPENROUTER_API_KEY, model via OPENROUTER_MODEL, base fixed)
+    Defaults pick small reasoning-capable models.
+    """
+
     def __init__(self):
         self.provider = (os.getenv("LLM_PROVIDER") or "openai").strip().lower()
-        # ... existing provider init ...
+
+        if self.provider == "openai":
+            self.api_key = os.getenv("OPENAI_API_KEY")
+            if not self.api_key:
+                raise ValueError("OPENAI_API_KEY is required when LLM_PROVIDER=openai.")
+            # Reasoning-lite default
+            self.model = os.getenv("OPENAI_MODEL", "gpt-4o-mini")
+            self.base_url = os.getenv("OPENAI_BASE_URL", "https://api.openai.com/v1")
+            self.headers = {
+                "Authorization": f"Bearer {self.api_key}",
+                "Content-Type": "application/json",
+            }
+
+        elif self.provider == "together":
+            self.api_key = os.getenv("TOGETHER_API_KEY")
+            if not self.api_key:
+                raise ValueError("TOGETHER_API_KEY is required when LLM_PROVIDER=together.")
+            self.model = os.getenv("TOGETHER_MODEL", "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo")
+            self.base_url = os.getenv("TOGETHER_BASE_URL", "https://api.together.xyz/v1")
+            self.headers = {
+                "Authorization": f"Bearer {self.api_key}",
+                "Content-Type": "application/json",
+            }
+
+        elif self.provider == "openrouter":
+            self.api_key = os.getenv("OPENROUTER_API_KEY")
+            if not self.api_key:
+                raise ValueError("OPENROUTER_API_KEY is required when LLM_PROVIDER=openrouter.")
+            # Strong, cost-effective default; change as desired
+            self.model = os.getenv("OPENROUTER_MODEL", "meta-llama/llama-3.1-8b-instruct:free")
+            self.base_url = "https://openrouter.ai/api/v1"
+            self.headers = {
+                "Authorization": f"Bearer {self.api_key}",
+                "Content-Type": "application/json",
+            }
+
+        else:
+            raise ValueError("LLM_PROVIDER must be one of: openai, together, openrouter")
+
+        # Backoff settings
         self.max_retries = int(os.getenv("MAX_RETRIES", "8"))
         self.base_backoff = float(os.getenv("BASE_BACKOFF", "1.0"))  # seconds
         self.max_backoff = float(os.getenv("MAX_BACKOFF", "30.0"))   # cap
@@ -23,8 +76,7 @@ class LLMClient:
         for attempt in range(1, self.max_retries + 1):
             try:
                 resp = requests.post(url, headers=headers, json=payload, timeout=timeout)
-                if resp.status_code == 429:
-                    # Respect Retry-After if present
+                if resp.status_code in (429, 500, 502, 503, 504):
                     retry_after = resp.headers.get("Retry-After")
                     if retry_after:
                         try:
@@ -33,14 +85,7 @@ class LLMClient:
                             pass
                     jitter = random.uniform(0, delay * 0.25)
                     wait_s = min(delay + jitter, self.max_backoff)
-                    print(f"⏳ Rate limited (429). Attempt {attempt}/{self.max_retries}. Waiting {wait_s:.2f}s...")
-                    time.sleep(wait_s)
-                    delay = min(delay * 1.8, self.max_backoff)
-                    continue
-                if resp.status_code in (500, 502, 503, 504):
-                    jitter = random.uniform(0, delay * 0.25)
-                    wait_s = min(delay + jitter, self.max_backoff)
-                    print(f"⏳ Server error {resp.status_code}. Attempt {attempt}/{self.max_retries}. Waiting {wait_s:.2f}s...")
+                    print(f"⏳ {self.provider} {resp.status_code}. Attempt {attempt}/{self.max_retries}. Waiting {wait_s:.2f}s...")
                     time.sleep(wait_s)
                     delay = min(delay * 1.8, self.max_backoff)
                     continue
@@ -56,18 +101,19 @@ class LLMClient:
             raise last_exc
         return requests.post(url, headers=headers, json=payload, timeout=timeout)
 
-    def generate(self, prompt: str, max_tokens: int = 512, temperature: float = 0.6, top_p: float = 0.9) -> str:
+    def chat(self, system_prompt: str, user_prompt: str, max_tokens: int = 512, temperature: float = 0.6, top_p: float = 0.9) -> str:
         url = f"{self.base_url}/chat/completions"
         payload: Dict[str, Any] = {
             "model": self.model,
             "messages": [
-                {"role": "system", "content": "You are a helpful, precise assistant."},
-                {"role": "user", "content": prompt},
+                {"role": "system", "content": system_prompt},
+                {"role": "user", "content": user_prompt},
             ],
             "max_tokens": max_tokens,
             "temperature": temperature,
             "top_p": top_p,
         }
+
         try:
             resp = self._post_with_backoff(url, headers=self.headers, payload=payload, timeout=120)
             if resp.status_code == 404:
@@ -84,37 +130,145 @@ class LLMClient:
         except Exception as e:
             return f"Error: Unexpected failure while processing. Details: {e}"
 
+# --- Simple tool-use: safe-ish arithmetic and unit conversions ---
 
-class BasicAgent:
+def maybe_compute_locally(question: str) -> Optional[str]:
     """
-    Agent that delegates to a pluggable non-HF provider (OpenAI or Together).
+    Lightweight math helper for common GAIA-style arithmetic/unit questions.
+    Avoids eval of arbitrary text; only allows digits, ops, dots, spaces, and parentheses.
+    Returns a computed answer string or None to defer to LLM.
     """
+    # Quick patterns: pure arithmetic, percents, simple ratios, sqrt, powers
+    text = question.strip().lower()
+
+    # Extract expression inside "calculate"/"compute"/"what is"
+    match = re.search(r"(?:calculate|compute|what is|evaluate)\s*[:\-]?\s*(.+)", text)
+    expr = match.group(1).strip() if match else text
+
+    # Replace common words to operators
+    expr = expr.replace("×", "*").replace("÷", "/").replace("^", "**")
+    expr = re.sub(r"(\d+)\s+percent", r"(\1/100)", expr)
 
+    # Allow only safe characters
+    if not re.fullmatch(r"[0-9\.\s\+\-\*\/\(\)\%]+", re.sub(r"\*\*", "**", expr)):
+        return None
+
+    # Disallow modulo for now (often not needed)
+    if "%" in expr:
+        return None
+
+    try:
+        # Use Python arithmetic safely
+        result = eval(expr, {"__builtins__": {}}, {"sqrt": math.sqrt, "pow": pow})
+    except Exception:
+        return None
+
+    if isinstance(result, float):
+        # Round to a reasonable precision
+        result = round(result, 6)
+        # normalize -0.0
+        if result == 0:
+            result = 0.0
+    return str(result)
+
+# --- Basic Agent Definition ---
+# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
+
+class BasicAgent:
     def __init__(self):
-        self.client = LLMClient()
-        print(f"✅ BasicAgent initialized with provider: {self.client.provider}, model: {self.client.model}")
-
-    def __call__(self, question: str, max_new_tokens: int = 512, temperature: float = 0.6, top_p: float = 0.9) -> str:
-        print(f"📝 Agent received question (preview): {question[:80]}...")
-        answer = self.client.generate(
-            prompt=question,
-            max_tokens=max_new_tokens,
-            temperature=temperature,
-            top_p=top_p,
+        # Config
+        self.rate_limit_s = float(os.getenv("RATE_LIMIT_SECONDS", "1.0"))
+        self.max_new_tokens = int(os.getenv("MAX_NEW_TOKENS", "256"))
+        self.samples = int(os.getenv("NUM_SAMPLES", "3"))  # self-consistency votes
+        self.temperature = float(os.getenv("TEMPERATURE", "0.6"))
+        self.top_p = float(os.getenv("TOP_P", "0.9"))
+
+        # System prompt optimized for concise, correct answers
+        self.system_prompt = (
+            "You are a precise reasoning assistant. Answer with the final result only, "
+            "without extra explanations, unless the question explicitly asks for steps or justification. "
+            "Prefer exact values, otherwise numeric to a sensible precision. Be factual and avoid speculation."
         )
-        print(f"🤖 Agent returning answer (preview): {answer[:80]}...")
-        return answer
-
 
-def run_and_submit_all(profile: gr.OAuthProfile | None):
+        # LLM client (non-HF endpoints)
+        self.client = LLMClient()
+        print(f"BasicAgent initialized with provider={self.client.provider}, model={self.client.model}, samples={self.samples}")
+
+    def _finalize(self, text: str) -> str:
+        # Trim quotes, code fences, and whitespace
+        t = text.strip()
+        t = re.sub(r"^```.*?\n", "", t, flags=re.DOTALL)  # remove starting fence if any
+        t = t.strip("` \n")
+        # Collapse multiple spaces
+        t = re.sub(r"\s+", " ", t)
+        return t.strip()
+
+    def _sample_answer(self, question: str) -> str:
+        # Try local compute first for arithmetic-like prompts
+        local = maybe_compute_locally(question)
+        if local is not None:
+            return local
+
+        # Otherwise query the LLM
+        reply = self.client.chat(
+            system_prompt=self.system_prompt,
+            user_prompt=question,
+            max_tokens=self.max_new_tokens,
+            temperature=self.temperature,
+            top_p=self.top_p,
+        )
+        return self._finalize(reply)
+
+    def __call__(self, question: str) -> str:
+        print(f"Agent received question (first 50 chars): {question[:50]}...")
+
+        # Self-consistency: multiple samples and vote
+        candidates: List[str] = []
+        for i in range(max(1, self.samples)):
+            ans = self._sample_answer(question)
+            candidates.append(ans)
+            print(f"  sample {i+1}/{self.samples}: {ans[:100]}")
+            if i < self.samples - 1:
+                time.sleep(self.rate_limit_s)
+
+        # Majority vote by normalized string; numeric-aware normalization
+        normalized_counts: Dict[str, int] = {}
+        mapping: Dict[str, str] = {}
+
+        def normalize(s: str) -> str:
+            s_clean = s.strip().lower()
+            # Try extract number
+            num = re.findall(r"-?\d+(?:\.\d+)?", s_clean)
+            if len(num) == 1 and s_clean.replace(num[0], "").strip() in ["", "units", "unit"]:
+                try:
+                    val = float(num[0])
+                    return f"{round(val, 6)}"
+                except Exception:
+                    pass
+            return s_clean
+
+        for c in candidates:
+            key = normalize(c)
+            normalized_counts[key] = normalized_counts.get(key, 0) + 1
+            # Keep first original for formatting
+            if key not in mapping:
+                mapping[key] = c
+
+        best_key = max(normalized_counts, key=normalized_counts.get)
+        final = mapping[best_key]
+        print(f"Agent returning answer: {final}")
+        return final
+
+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")
+    # --- Determine HF Space Runtime URL and Repo URL ---
+    space_id = os.getenv("SPACE_ID") # Get the SPACE_ID for sending link to the code
 
     if profile:
-        username = f"{profile.username}"
+        username= f"{profile.username}"
         print(f"User logged in: {username}")
     else:
         print("User not logged in.")
@@ -124,13 +278,13 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     questions_url = f"{api_url}/questions"
     submit_url = f"{api_url}/submit"
 
-    # 1. Instantiate Agent
+    # 1. Instantiate Agent ( modify this part to create your agent)
     try:
         agent = BasicAgent()
     except Exception as e:
         print(f"Error instantiating agent: {e}")
         return f"Error initializing agent: {e}", None
-
+    # In the case of an app running as a hugging Face space, this link points toward your codebase ( usefull for others so please keep it public)
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
     print(agent_code)
 
@@ -141,16 +295,16 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
         response.raise_for_status()
         questions_data = response.json()
         if not questions_data:
-            print("Fetched questions list is empty.")
-            return "Fetched questions list is empty or invalid format.", None
+             print("Fetched questions list is empty.")
+             return "Fetched questions list is empty or invalid format.", None
         print(f"Fetched {len(questions_data)} questions.")
     except requests.exceptions.RequestException as e:
         print(f"Error fetching questions: {e}")
         return f"Error fetching questions: {e}", None
     except requests.exceptions.JSONDecodeError as e:
-        print(f"Error decoding JSON response from questions endpoint: {e}")
-        print(f"Response text: {response.text[:500]}")
-        return f"Error decoding server response for questions: {e}", None
+         print(f"Error decoding JSON response from questions endpoint: {e}")
+         print(f"Response text: {response.text[:500]}")
+         return f"Error decoding server response for questions: {e}", None
     except Exception as e:
         print(f"An unexpected error occurred fetching questions: {e}")
         return f"An unexpected error occurred fetching questions: {e}", None
@@ -159,7 +313,7 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
-    for item in questions_data:
+    for idx, item in enumerate(questions_data, start=1):
         task_id = item.get("task_id")
         question_text = item.get("question")
         if not task_id or question_text is None:
@@ -170,14 +324,14 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
             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})
         except Exception as e:
-            print(f"Error running agent on task {task_id}: {e}")
-            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
+             print(f"Error running agent on task {task_id}: {e}")
+             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"AGENT ERROR: {e}"})
 
     if not answers_payload:
         print("Agent did not produce any answers to submit.")
         return "Agent did not produce any answers to submit.", pd.DataFrame(results_log)
 
-    # 4. Prepare Submission
+    # 4. Prepare Submission 
     submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
     status_update = f"Agent finished. Submitting {len(answers_payload)} answers for user '{username}'..."
     print(status_update)
@@ -232,11 +386,9 @@ with gr.Blocks() as demo:
     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).
@@ -249,6 +401,7 @@ with gr.Blocks() as demo:
     run_button = gr.Button("Run Evaluation & Submit All Answers")
 
     status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
+    # Removed max_rows=10 from DataFrame constructor
     results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
 
     run_button.click(
@@ -258,8 +411,9 @@ with gr.Blocks() as demo:
 
 if __name__ == "__main__":
     print("\n" + "-"*30 + " App Starting " + "-"*30)
+    # Check for SPACE_HOST and SPACE_ID at startup for information
     space_host_startup = os.getenv("SPACE_HOST")
-    space_id_startup = os.getenv("SPACE_ID")
+    space_id_startup = os.getenv("SPACE_ID") # Get SPACE_ID at startup
 
     if space_host_startup:
         print(f"✅ SPACE_HOST found: {space_host_startup}")
@@ -267,7 +421,7 @@ if __name__ == "__main__":
     else:
         print("ℹ️  SPACE_HOST environment variable not found (running locally?).")
 
-    if space_id_startup:
+    if space_id_startup: # Print repo URLs if SPACE_ID is found
         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")
@@ -275,5 +429,6 @@ if __name__ == "__main__":
         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)