Update inference.py

inference.py

@@ -1,116 +1,117 @@
-import os
-import json
-import requests
-from openai import OpenAI
-
-# ==========================================
-# 1. MANDATORY ENVIRONMENT VARIABLES
-# The hackathon automated grader will inject these.
-# For local testing, we provide default fallbacks to Gemini's OpenAI-compatible endpoint.
-# ==========================================
-API_BASE_URL = os.getenv("API_BASE_URL", "https://generativelanguage.googleapis.com/v1beta/openai/")
-MODEL_NAME = os.getenv("MODEL_NAME", "gemini-2.5-flash")
-# Note: For local testing, set your HF_TOKEN environment variable to your Gemini API key!
-HF_TOKEN = os.getenv("HF_TOKEN") 
-
-# ==========================================
-# 2. ENVIRONMENT SETUP
-# ==========================================
-# REPLACE THIS WITH YOUR ACTUAL HUGGING FACE DIRECT URL
-ENV_URL = "https://mark012-logisticsflow-openenv.hf.space"
-TASK_NAME = "hard"
-BENCHMARK = "LogisticsFlow-OpenEnv"
-
-client = OpenAI(api_key=HF_TOKEN, base_url=API_BASE_URL)
-
-# ==========================================
-# 3. STRICT LOGGING FORMATTERS
-# ==========================================
-def log_start(task: str, env: str, model: str) -> None:
-    print(f"[START] task={task} env={env} model={model}", flush=True)
-
-def log_step(step: int, action: str, reward: float, done: bool, error: str) -> None:
-    error_val = error if error else "null"
-    done_val = str(done).lower()
-    print(f"[STEP] step={step} action={action} reward={reward:.2f} done={done_val} error={error_val}", flush=True)
-
-def log_end(success: bool, steps: int, score: float, rewards: list) -> None:
-    rewards_str = ",".join(f"{r:.2f}" for r in rewards)
-    print(f"[END] success={str(success).lower()} steps={steps} score={score:.3f} rewards={rewards_str}", flush=True)
-
-# ==========================================
-# 4. INFERENCE LOOP
-# ==========================================
-def run_inference():
-    log_start(task=TASK_NAME, env=BENCHMARK, model=MODEL_NAME)
-
-    # Reset Environment
-    try:
-        res = requests.post(f"{ENV_URL}/reset", json={"level": TASK_NAME})
-        obs = res.json()
-    except Exception as e:
-        print(f"Failed to connect to environment: {e}")
-        return
-
-    rewards = []
-    steps_taken = 0
-    success = False
-
-    for step in range(1, 21): # Max 20 steps
-        # Extract the observation state
-        current_state = obs.get("observation", obs) if isinstance(obs, dict) else obs
-
-        # Prepare prompt
-        system_prompt = "You are an AI logistics agent. Analyze the state. You must output exactly valid JSON. Actions: {'command': 'ship', 'params': {'order_id': 'ORD-XYZ', 'carrier': 'Standard'}} OR {'command': 'restock', 'params': {'item': 'Electronics'}}."
-        user_prompt = f"Current State: {json.dumps(current_state)}"
-
-        # Get Model Action
-        try:
-            response = client.chat.completions.create(
-                model=MODEL_NAME,
-                messages=[
-                    {"role": "system", "content": system_prompt},
-                    {"role": "user", "content": user_prompt}
-                ],
-                response_format={"type": "json_object"} # Forces valid JSON
-            )
-            action_str = response.choices[0].message.content.strip()
-            action_json = json.loads(action_str)
-            action_log = action_str.replace('\n', '').replace(' ', '') # Flatten for logging
-            error = None
-        except Exception as e:
-            action_json = {"command": "wait", "params": {}}
-            action_log = "error"
-            error = str(e)
-
-        # Step the Environment
-        try:
-            res = requests.post(f"{ENV_URL}/step", json=action_json)
-            obs = res.json()
-            reward = obs.get("reward", 0.0)
-            done = obs.get("done", False)
-        except Exception as e:
-            reward = 0.0
-            done = True
-            error = str(e)
-
-        rewards.append(reward)
-        steps_taken = step
-        
-        # Log strictly
-        log_step(step=step, action=action_log, reward=reward, done=done, error=error)
-
-        if done:
-            break
-
-    # Calculate final score (clamp between 0 and 1 for judges)
-    score = sum(rewards)
-    normalized_score = min(max(score, 0.0), 1.0)
-    success = normalized_score > 0.0
-
-    log_end(success=success, steps=steps_taken, score=normalized_score, rewards=rewards)
-
-if __name__ == "__main__":
-    # Ensure API key is set in terminal before running:
-    # Windows: $env:HF_TOKEN="your_gemini_api_key_here"
+import os
+import json
+import urllib.request
+from openai import OpenAI
+
+# ==========================================
+# 1. MANDATORY ENVIRONMENT VARIABLES
+# ==========================================
+API_BASE_URL = os.getenv("API_BASE_URL", "https://generativelanguage.googleapis.com/v1beta/openai/")
+MODEL_NAME = os.getenv("MODEL_NAME", "gemini-2.5-flash")
+HF_TOKEN = os.getenv("HF_TOKEN") 
+
+# ==========================================
+# 2. ENVIRONMENT SETUP
+# ==========================================
+ENV_URL = "https://mark012-logisticsflow-openenv.hf.space"
+TASK_NAME = "hard"
+BENCHMARK = "LogisticsFlow-OpenEnv"
+
+client = OpenAI(api_key=HF_TOKEN, base_url=API_BASE_URL)
+
+# ==========================================
+# 3. STRICT LOGGING FORMATTERS
+# ==========================================
+def log_start(task: str, env: str, model: str) -> None:
+    print(f"[START] task={task} env={env} model={model}", flush=True)
+
+def log_step(step: int, action: str, reward: float, done: bool, error: str) -> None:
+    error_val = error if error else "null"
+    done_val = str(done).lower()
+    print(f"[STEP] step={step} action={action} reward={reward:.2f} done={done_val} error={error_val}", flush=True)
+
+def log_end(success: bool, steps: int, score: float, rewards: list) -> None:
+    rewards_str = ",".join(f"{r:.2f}" for r in rewards)
+    print(f"[END] success={str(success).lower()} steps={steps} score={score:.3f} rewards={rewards_str}", flush=True)
+
+# ==========================================
+# 4. NETWORK HELPER (No 'requests' library needed)
+# ==========================================
+def send_post_request(url: str, payload: dict) -> dict:
+    data = json.dumps(payload).encode('utf-8')
+    req = urllib.request.Request(url, data=data, headers={'Content-Type': 'application/json'})
+    with urllib.request.urlopen(req) as response:
+        return json.loads(response.read().decode('utf-8'))
+
+# ==========================================
+# 5. INFERENCE LOOP
+# ==========================================
+def run_inference():
+    log_start(task=TASK_NAME, env=BENCHMARK, model=MODEL_NAME)
+
+    # Reset Environment
+    try:
+        obs = send_post_request(f"{ENV_URL}/reset", {"level": TASK_NAME})
+    except Exception as e:
+        print(f"Failed to connect to environment: {e}")
+        # The prompt instructed us to wrap risky operations. If it fails, we gracefully exit.
+        return
+
+    rewards = []
+    steps_taken = 0
+    success = False
+
+    for step in range(1, 21): # Max 20 steps
+        current_state = obs.get("observation", obs) if isinstance(obs, dict) else obs
+
+        # Prepare prompt
+        system_prompt = "You are an AI logistics agent. Analyze the state. You must output exactly valid JSON. Actions: {'command': 'ship', 'params': {'order_id': 'ORD-XYZ', 'carrier': 'Standard'}} OR {'command': 'restock', 'params': {'item': 'Electronics'}}."
+        user_prompt = f"Current State: {json.dumps(current_state)}"
+
+        # Get Model Action
+        try:
+            response = client.chat.completions.create(
+                model=MODEL_NAME,
+                messages=[
+                    {"role": "system", "content": system_prompt},
+                    {"role": "user", "content": user_prompt}
+                ],
+                response_format={"type": "json_object"} # Forces valid JSON
+            )
+            action_str = response.choices[0].message.content.strip()
+            action_json = json.loads(action_str)
+            action_log = action_str.replace('\n', '').replace(' ', '') # Flatten for logging
+            error = None
+        except Exception as e:
+            action_json = {"command": "wait", "params": {}}
+            action_log = "error"
+            error = str(e)
+
+        # Step the Environment
+        try:
+            obs = send_post_request(f"{ENV_URL}/step", action_json)
+            reward = obs.get("reward", 0.0)
+            done = obs.get("done", False)
+        except Exception as e:
+            reward = 0.0
+            done = True
+            error = str(e)
+
+        rewards.append(reward)
+        steps_taken = step
+        
+        # Log strictly
+        log_step(step=step, action=action_log, reward=reward, done=done, error=error)
+
+        if done:
+            break
+
+    # Calculate final score (clamp between 0 and 1 for judges)
+    score = sum(rewards)
+    normalized_score = min(max(score, 0.0), 1.0)
+    success = normalized_score > 0.0
+
+    log_end(success=success, steps=steps_taken, score=normalized_score, rewards=rewards)
+
+if __name__ == "__main__":
     run_inference()