Update app.py

app.py

@@ -12,40 +12,70 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 from smolagents import (
     CodeAgent,
     DuckDuckGoSearchTool,
-    InferenceClientModel,
     TransformersModel,
-    LogLevel,
+    #InferenceClientModel,
+    #LogLevel,
 )
+
+import re
+
+def answer_only(text: str) -> str:
+    """Keep the minimal exact answer; strip markdown/explanations/punctuation."""
+    if text is None:
+        return ""
+    s = str(text).strip()
+
+    # If there is a single integer in the output, return just that integer.
+    m = re.findall(r"-?\d+", s)
+    if len(m) == 1:
+        return m[0]
+
+    # If output is wrapped in markdown or quotes, strip them.
+    s = s.strip("`").strip('"').strip("'")
+
+    # Remove markdown like **bold** or *italics*
+    s = re.sub(r"\*\*(.*?)\*\*", r"\1", s)
+    s = re.sub(r"\*(.*?)\*", r"\1", s)
+
+    # Trim trailing period if answer looks like a number or a simple token.
+    if s.endswith(".") and re.fullmatch(r"[A-Za-z0-9\- ]+\.", s):
+        s = s[:-1]
+
+    return s.strip()
+
+
+
 def build_model():
-    hf_token = os.getenv("HF_TOKEN")
-    if not hf_token:
-        raise RuntimeError("HF_TOKEN is not set (Space → Settings → Variables and secrets).")
-
-    # Try a prioritized list of chat-capable models on the serverless chat router
-    serverless_candidates = [
-        "Qwen/Qwen2.5-7B-Instruct-1M",
-        "mistralai/Mistral-7B-Instruct-v0.3",
-        "HuggingFaceH4/zephyr-7b-beta",
-    ]
-
-    last_err = None
-    for mid in serverless_candidates:
-        try:
-            return InferenceClientModel(
-                model_id=mid,
-                token=hf_token,     # let provider auto-select, don't force provider=
-                timeout=120,
-            )
-        except Exception as e:
-            print(f"[warn] Serverless model '{mid}' failed: {e!r}")
-            last_err = e
+    # hf_token = os.getenv("HF_TOKEN")
+    # if not hf_token:
+    #     raise RuntimeError("HF_TOKEN is not set (Space → Settings → Variables and secrets).")
+
+    # # Try a prioritized list of chat-capable models on the serverless chat router
+    # serverless_candidates = [
+    #     "Qwen/Qwen2.5-7B-Instruct-1M",
+    #     "mistralai/Mistral-7B-Instruct-v0.3",
+    #     "HuggingFaceH4/zephyr-7b-beta",
+    # ]
+
+    # last_err = None
+    # for mid in serverless_candidates:
+    #     try:
+    #         return InferenceClientModel(
+    #             model_id=mid,
+    #             token=hf_token,     # let provider auto-select, don't force provider=
+    #             timeout=120,
+    #         )
+    #     except Exception as e:
+    #         print(f"[warn] Serverless model '{mid}' failed: {e!r}")
+    #         last_err = e
 
     # Final fallback: local transformers (CPU; slower, but never hits the chat router)
     try:
         return TransformersModel(
-            model_id="HuggingFaceTB/SmolLM2-1.7B-Instruct",
-            max_new_tokens=256,
-        )
+        model_id="HuggingFaceTB/SmolLM2-1.7B-Instruct",  # or "Qwen/Qwen2.5-0.5B-Instruct"
+        max_new_tokens=220,
+        temperature=0.2,
+    )
     except Exception as e:
         raise RuntimeError(f"All model inits failed. Last serverless error: {last_err!r}; "
                            f"Transformers fallback error: {e!r}")
@@ -56,9 +86,9 @@ class BasicAgent:
     def __init__(self):
         self.search = DuckDuckGoSearchTool()
 
-        hf_token = os.getenv("HF_TOKEN")
-        if not hf_token:
-            raise RuntimeError("HF_TOKEN is not set (Space → Settings → Variables and secrets).")
+        # hf_token = os.getenv("HF_TOKEN")
+        # if not hf_token:
+        #     raise RuntimeError("HF_TOKEN is not set (Space → Settings → Variables and secrets).")
 
         # Use a chat-capable model; let provider auto-route
         self.model = build_model()
@@ -67,16 +97,22 @@ class BasicAgent:
         self.agent = CodeAgent(
             tools=[self.search],
             model=self.model,
-            max_steps=5,
+            max_steps=3,
             verbosity_level=1
         )
 
         # Health check WITHOUT verbosity_level in run()
         _ = self.agent.run("Reply with the single word: OK")
-
-    def __call__(self, question: str) -> str:
-        # No verbosity_level here either
-        return self.agent.run(question)
+    
+    def __call__(self, item: dict) -> str:
+            q = (item.get("question") or "").strip()
+            prompt = (
+                "Answer with the shortest exact phrase/number only. "
+                "No explanations.\n\n"
+                f"Question: {q}\nAnswer:"
+            )
+            raw = self.agent.run(prompt)
+            return answer_only(raw)  # exact-match friendly
 
 
 
@@ -136,17 +172,34 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     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")
-        if not task_id or question_text is None:
+        question_text = (item.get("question") or "").strip()
+        if not task_id or not question_text:
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
-            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})
+            # ✅ pass the whole item so the agent can use task_id (e.g., GET /files/{task_id})
+            submitted_answer_raw = agent(item)
+    
+            # ✅ keep answers exact-match friendly (only if you added answer_only())
+            submitted_answer = answer_only(submitted_answer_raw) if callable(globals().get("answer_only")) else submitted_answer_raw
+    
+            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.")
@@ -219,18 +272,16 @@ with gr.Blocks() as demo:
         """
     )
 
-    gr.LoginButton()
+    login_btn = gr.LoginButton()  # capture the profile object
     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(
         fn=run_and_submit_all,
+        inputs=[login_btn],              # ✅ pass the OAuth profile
         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