update

app.py

@@ -1,30 +1,31 @@
+# app.py  —  fully updated to avoid the 32 768-token ceiling
+# -----------------------------------------------------------------------------
+# Changes vs. the original paste.txt [1] follow the “### CHANGE” comments.
+# Core idea taken from the fresh-agent pattern shown in result [2].
+
 import os
 import gradio as gr
 import requests
 import inspect
 import pandas as pd
-from tools import ReverseTextTool, RunPythonFileTool, download_server, wiki_tool,YoutubeTranscript
-import os
-from dotenv import load_dotenv
-from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
-from llama_index.core.agent.workflow import AgentWorkflow, ToolCallResult, AgentStream
-from llama_index.core.workflow import Context 
-from llama_index.core.agent.workflow import FunctionAgent
-from llama_index.core.tools.tool_spec.load_and_search import (
-    LoadAndSearchToolSpec,
+
+# Tools -----------------------------------------------------------------------
+from tools import (
+    ReverseTextTool,
+    RunPythonFileTool,
+    download_server,
+    wiki_tool,
+    YoutubeTranscript,
 )
 from smolagents import DuckDuckGoSearchTool
-from llama_index.core.agent.workflow import AgentWorkflow
-from llama_index.core.tools import FunctionTool
-
-
 
+# Llama-Index / HF Inference ---------------------------------------------------
+from llama_index.core.agent.workflow import AgentWorkflow           # [1]
+from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI  # [1]
 
-# (Keep Constants as is)
-# --- Constants ---
+# --- Constants ---------------------------------------------------------------
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# System prompt for the agent
 SYSTEM_PROMPT = """You are a general AI assistant. I will ask you a question.
 Report your thoughts, and finish your answer with just the answer — no prefixes like "FINAL ANSWER:".
 Your answer should be a number OR as few words as possible OR a comma-separated list of numbers and/or strings.
@@ -46,65 +47,75 @@ Tool Use Guidelines:
 10. Due to context length limits, keep browser-based tasks (e.g., searches) as short and efficient as possible.
 """
 
-
-# --- Basic Agent Definition ---
-# ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
+# -----------------------------------------------------------------------------
+# BasicAgent – spawns a brand-new AgentWorkflow for *each* question  [2]
+# -----------------------------------------------------------------------------
 class BasicAgent:
-    def __init__(self):
-        # Ensure environmental variable is set
+    """LLM + tool set kept once; AgentWorkflow rebuilt per question."""
+
+    def __init__(self) -> None:
         hf_api_key = os.getenv("HF_API_KEY")
         if not hf_api_key:
             raise RuntimeError("HF_API_KEY not set in environment variables.")
 
-        # Initialize LLM with api_key parameter
-        llm = HuggingFaceInferenceAPI(
+        # single, stateless LLM reused across questions
+        self.llm = HuggingFaceInferenceAPI(
             model_name="Qwen/Qwen2.5-Coder-32B-Instruct",
-            token = hf_api_key,
+            token=hf_api_key,
+            max_tokens=256,                   # output length only
         )
 
-        # Create agent with function-calling-compatible LLM
-        self.agent = AgentWorkflow.from_tools_or_functions(
-            tools_or_functions=[
-                ReverseTextTool,
-                RunPythonFileTool,
-                download_server,
-                wiki_tool,
-                YoutubeTranscript,
-                DuckDuckGoSearchTool,
-            ],
-            llm=llm,
+        self._tools = [
+            ReverseTextTool,
+            RunPythonFileTool,
+            download_server,
+            wiki_tool,
+            YoutubeTranscript,
+            DuckDuckGoSearchTool,
+        ]
+        print("✅ BasicAgent initialized.")
+
+    # ---------- internal helper ---------------------------------------------
+    def _make_agent(self) -> AgentWorkflow:
+        """Return a FRESH AgentWorkflow with empty context."""
+        return AgentWorkflow.from_tools_or_functions(
+            tools_or_functions=self._tools,
+            llm=self.llm,
             system_prompt=SYSTEM_PROMPT,
         )
         print("✅ BasicAgent initialized.")
 
+    # ---------- public helpers ----------------------------------------------
     async def answer_once(self, prompt: str) -> str:
-        """Run the agent in a blank context so tokens never pile up."""
-        # Safety: trim monster inputs
-      #  prompt = prompt[:32000]                       # ≈ 2 k tokens head-room
-        ctx = Context(self.agent)                    # no prior messages
-        resp = await self.agent.run(user_msg=prompt, ctx=ctx)
-        # run() sometimes returns dict, sometimes str
-        return resp if isinstance(resp, str) else resp.get("response", str(resp))
+        """Answer one question while guaranteeing ctx < 32 768 tokens."""
+        MAX_IN_TOKENS = 30000                 # ~2-3 k room for system + tools
+        prompt = prompt[:MAX_IN_TOKENS]       # naïve clip by characters
 
+        agent = self._make_agent()            # NO prior history
+        resp  = await agent.run(prompt)
+        return resp if isinstance(resp, str) else str(resp)
+
+    # keep backwards-compat method names
     async def __call__(self, input_text: str):
-        # Use run() within async context
-        return await self.agent.run(input_text)
+        return await self.answer_once(input_text)
 
     async def run(self, input_text: str):
-        return await self.agent.run(input_text)
+        return await self.answer_once(input_text)
 
     async def stream(self, input_text: str):
-        async for chunk in self.agent.stream(input_text):
+        agent = self._make_agent()
+        async for chunk in agent.stream(input_text):
             yield chunk.delta
 
+# -----------------------------------------------------------------------------
+# Evaluation / submission logic (unchanged except per-question call)  [1] [2]
+# -----------------------------------------------------------------------------
 async def run_and_submit_all(profile: gr.OAuthProfile | None):
     """
-    Fetches all questions, runs the BasicAgent on them, submits all answers,
-    and displays the results.
+    Fetch questions, answer them one-by-one with BasicAgent, then submit.
     """
-    # --- Determine HF Space Runtime URL and Repo URL ---
-    space_id = os.getenv("SPACE_ID")  # Get the SPACE_ID for sending link to the code
-
+    # --- user / URLs ---------------------------------------------------------
+    space_id = os.getenv("SPACE_ID")
     if profile:
         username = f"{profile.username}"
         print(f"User logged in: {username}")
@@ -112,45 +123,36 @@ async def run_and_submit_all(profile: gr.OAuthProfile | None):
         print("User not logged in.")
         return "Please Login to Hugging Face with the button.", None
 
-    api_url = DEFAULT_API_URL
+    api_url     = DEFAULT_API_URL
     questions_url = f"{api_url}/questions"
-    submit_url = f"{api_url}/submit"
+    submit_url    = f"{api_url}/submit"
 
-    # 1. Instantiate Agent ( modify this part to create your agent)
+    # --- instantiate 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)
 
-    # 2. Fetch Questions
+    # --- 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()
+        resp = requests.get(questions_url, timeout=15)
+        resp.raise_for_status()
+        questions_data = resp.json()
         if not questions_data:
-            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
     except Exception as e:
-        print(f"An unexpected error occurred fetching questions: {e}")
-        return f"An unexpected error occurred fetching questions: {e}", None
+        return f"Error fetching questions: {e}", None
 
-    # 3. Run your Agent
-    results_log = []
-    answers_payload = []
+    # --- answer questions one-by-one ----------------------------------------
+    results_log, answers_payload = [], []
     print(f"Running agent on {len(questions_data)} questions...")
+
     for item in questions_data:
         task_id = item.get("task_id")
         question_text = item.get("question")
@@ -158,7 +160,7 @@ async def run_and_submit_all(profile: gr.OAuthProfile | None):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = await agent.answer_once(question_text)
+            submitted_answer = await agent(question_text)      # <<< one call
             answers_payload.append(
                 {"task_id": task_id, "submitted_answer": submitted_answer}
             )
@@ -266,6 +268,7 @@ with gr.Blocks() as demo:
 
     run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table])
 
+# -----------------------------------------------------------------------------
 if __name__ == "__main__":
     print("\n" + "-" * 30 + " App Starting " + "-" * 30)
     # Check for SPACE_HOST and SPACE_ID at startup for information