Update agents.py

agents.py

@@ -2,60 +2,76 @@ import os
 import json
 import requests
 
+# GLM-4.5 API configuration
 GLM_API_URL = "https://api.your-glm-provider.com/v1/chat/completions"
-GLM_API_KEY = os.getenv("GLM_API_KEY")
+GLM_API_KEY = os.getenv("ZHIPUAI_API_KEY")  # Hugging Face Secret
 
-def call_glm(prompt, temperature=0.3):
+def call_glm(system_prompt, user_prompt, temperature=0.3):
     headers = {"Authorization": f"Bearer {GLM_API_KEY}"}
     payload = {
         "model": "glm-4.5",
-        "messages": [{"role": "system", "content": "You are an educational assistant."},
-                     {"role": "user", "content": prompt}],
-        "temperature": temperature
+        "messages": [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_prompt},
+        ],
+        "temperature": temperature,
     }
     response = requests.post(GLM_API_URL, headers=headers, json=payload)
     result = response.json()
     return result["choices"][0]["message"]["content"]
 
 class AnalyzerAgent:
-    def analyze(self, answers, question_bank):
-        # Simple deterministic analysis
-        topic_stats = {}
-        for qid, result in answers.items():
-            q = next(item for item in question_bank if str(item["id"]) == qid)
-            for topic in q.get("topics", []):
-                if topic not in topic_stats:
-                    topic_stats[topic] = {"correct": 0, "total": 0}
-                topic_stats[topic]["total"] += 1
-                if result["user"] == result["correct"]:
-                    topic_stats[topic]["correct"] += 1
-        
-        analysis = {t: round(v["correct"]/v["total"], 2) for t,v in topic_stats.items()}
-        return {"topic_accuracy": analysis}
+    def analyze(self, answers, exam_data):
+        system_prompt = "You are an exam analysis assistant."
+        user_prompt = f"""
+        Analyze student answers against correct answers.
+        Answers: {json.dumps(answers)}
+        Exam questions: {json.dumps(exam_data)}
+        Return JSON: {{
+          "topic_accuracy": {{"topic": 0-1}},
+          "weak_topics": ["..."]
+        }}
+        """
+        try:
+            response = call_glm(system_prompt, user_prompt)
+            return json.loads(response)
+        except Exception as e:
+            return {"error": f"Analyzer failed: {str(e)}"}
 
 class ForecastAgent:
     def forecast(self, level, subject):
-        prompt = f"""
-        You are an exam forecast assistant.
-        Predict 3 high-probability topics for {level} {subject} exam.
-        Return JSON: {{"predicted_topics": [{{"topic": "...", "confidence": 0.0}}]}}
+        system_prompt = "You are an exam forecast assistant."
+        user_prompt = f"""
+        Predict 3 high-probability exam topics for {level} {subject}.
+        Return JSON: {{
+          "predicted_topics": [
+            {{"topic": "...", "confidence": 0-1}}
+          ]
+        }}
         """
         try:
-            response = call_glm(prompt)
+            response = call_glm(system_prompt, user_prompt)
             return json.loads(response)
-        except Exception:
-            return {"error": "Forecast unavailable"}
+        except Exception as e:
+            return {"error": f"Forecast failed: {str(e)}"}
 
 class CoachAgent:
     def coach(self, analysis, level, subject):
-        prompt = f"""
-        You are a study coach.
-        The student's weaknesses are: {analysis}.
-        Suggest a study plan and 3 practice questions for {level} {subject}.
-        Return JSON: {{"tips": ["..."], "study_plan": "...", "practice_questions": [{{"text": "...", "answer": "..."}}]}}
+        system_prompt = "You are a study coach for school exams."
+        user_prompt = f"""
+        Based on this analysis: {json.dumps(analysis)},
+        suggest a study plan and 3 practice questions for {level} {subject}.
+        Return JSON: {{
+          "tips": ["..."],
+          "study_plan": "...",
+          "practice_questions": [
+            {{"text": "...", "answer": "..."}}
+          ]
+        }}
         """
         try:
-            response = call_glm(prompt)
+            response = call_glm(system_prompt, user_prompt)
             return json.loads(response)
-        except Exception:
-            return {"error": "Coach unavailable"}
+        except Exception as e:
+            return {"error": f"Coach failed: {str(e)}"}
+