fix: pass client-supplied records/performanceData to Gemini in attendance-analysis and student-insights endpoints

main.py

@@ -324,41 +324,57 @@ def ai_student_insights():
         term_id      = data.get('termId', 'current')
         include_recs = data.get('includeRecommendations', True)
 
+        # Extract any performance data the client supplies
+        performance_data = (
+            data.get('performanceData')
+            or data.get('grades')
+            or data.get('scores')
+            or data.get('results')
+            or []
+        )
+        student_name = data.get('studentName') or data.get('student_name', '')
+
+        has_data = bool(performance_data)
+
         system_instruction = (
             "You are an educational AI analyst. Analyse student performance data and return "
             "a structured JSON report. Be empathetic, data-driven and actionable. "
             "CRITICAL: You do not have access to a database. Never invent, fabricate, or "
             "hallucinate student names, database IDs, scores, or subject data. "
-            "Use only data explicitly provided in this prompt. "
-            "In the summary field, refer to the student by their studentId only — never invent a name. "
-            "In subjectBreakdown, only include subjects explicitly listed in the subjects filter; "
-            "if no filter is given, state that subject-level data was not provided."
+            "Use ONLY data explicitly provided in this prompt. "
+            "In the summary field, refer to the student by their studentId only — never invent a name "
+            "unless studentName is explicitly provided. "
+            "In subjectBreakdown, derive scores and trends strictly from the performanceData records below."
         )
 
         prompt = f"""Generate a student performance insights report as JSON.
 
 Student ID: {student_id}
+Student Name: {student_name if student_name else 'NOT PROVIDED — refer to student by ID only'}
 School ID: {school_id}
-Subjects filter: {subjects if subjects else 'not specified — do not invent subject data'}
+Subjects filter: {subjects if subjects else 'not specified'}
 Term: {term_id}
 Include recommendations: {include_recs}
 
-IMPORTANT RULES:
-- Use the studentId "{student_id}" exactly as provided. Do not invent a student name.
-- Only include subjectBreakdown entries for subjects explicitly listed above.
-- If no real performance data is available, set averageScore to null and explain in the insight field.
-- Do not fabricate trends, scores, or risk levels.
+PERFORMANCE DATA (use these records for all analysis — do NOT invent scores):
+{json.dumps(performance_data, ensure_ascii=False) if has_data else 'NO PERFORMANCE DATA PROVIDED — set all scores to null and explain in summary.'}
 
-Return ONLY valid JSON matching this exact structure:
+STRICT RULES:
+- Use the studentId "{student_id}" exactly as provided.
+- Derive ALL scores, trends, and risk levels from the performance records above only.
+- If a subject appears in the subjects filter but has no records, set averageScore to null.
+- Do not fabricate any numbers, names, or trends.
+
+Return ONLY valid JSON:
 {{
   "studentId": "{student_id}",
-  "summary": "<2-3 sentence overview referring to student by ID, not an invented name>",
+  "summary": "<2-3 sentence overview based only on provided data>",
   "subjectBreakdown": [
     {{
-      "subject": "<subject from the provided filter only>",
+      "subject": "<subject name>",
       "trend": "<improving|stable|declining|unknown>",
-      "averageScore": <number 0-100 or null if not available>,
-      "insight": "<specific insight based only on provided data>",
+      "averageScore": <number 0-100 or null>,
+      "insight": "<specific insight from provided records>",
       "riskLevel": "<low|medium|high|unknown>"
     }}
   ],
@@ -368,8 +384,7 @@ Return ONLY valid JSON matching this exact structure:
 """
         result = gemini_json(prompt, system_instruction=system_instruction)
 
-        # Sanitise: ensure studentId in output matches what was requested,
-        # not a hallucinated value
+        # Sanitise: pin studentId to the request value
         if result:
             result['studentId'] = student_id
 
@@ -885,15 +900,25 @@ def ai_attendance_analysis():
         if scope in ('class', 'student') and not scope_id:
             return jsonify({'error': 'scopeId is required when scope is "class" or "student"'}), 400
 
+        # Extract attendance records the client supplies (several possible key names)
+        records = (
+            data.get('records')
+            or data.get('attendanceData')
+            or data.get('attendanceRecords')
+            or data.get('data')
+            or []
+        )
+        has_records = bool(records)
+
         system_instruction = (
             "You are an educational data analyst specialising in attendance patterns. "
             "Identify at-risk students and actionable patterns. "
-            "CRITICAL: You do not have access to a database. Never invent, fabricate, or "
-            "hallucinate student names, student IDs, percentages, or attendance records. "
-            "Only use data explicitly provided in this prompt. "
-            "If real attendance data is not provided, set overallRate to null, return empty "
-            "arrays for patterns and atRiskStudents, and explain in the summary that no data was supplied. "
-            "Never expose raw database document IDs in summary, detail, or recommendation text fields. "
+            "CRITICAL: You do not have access to a database. "
+            "Use ONLY the attendance records explicitly provided in the prompt below. "
+            "Never invent, fabricate, or hallucinate student names, IDs, percentages, or records. "
+            "If real attendance data is provided, analyse it thoroughly. "
+            "If no records are provided, set overallRate to null and return empty arrays. "
+            "Never expose raw database document IDs in any text field. "
             "Return only JSON."
         )
 
@@ -904,41 +929,45 @@ Scope: {scope}  (school | class | student)
 Scope ID: {scope_id or 'N/A'}
 Date Range: {json.dumps(date_range) if date_range else 'current term'}
 
-IMPORTANT RULES:
-- Do not invent student names or IDs. Only use names/IDs explicitly provided above.
+ATTENDANCE RECORDS (analyse these — do NOT invent data):
+{json.dumps(records, ensure_ascii=False) if has_records else 'NO RECORDS PROVIDED — return null overallRate and empty arrays.'}
+
+STRICT RULES:
+- Derive ALL percentages, patterns, and at-risk flags from the records above only.
+- Do not invent student names or IDs beyond what appears in the records.
 - Do not fabricate attendance percentages or absence counts.
-- If attendance records are not included in this prompt, return empty patterns and atRiskStudents arrays.
 - Never expose raw database document IDs in any text field (summary, detail, recommendation).
+- Calculate overallRate as: (present records / total records) * 100, rounded to 1 decimal.
 
 Return ONLY valid JSON:
 {{
   "scope": "{scope}",
   "scopeId": "{scope_id or ''}",
-  "period": {json.dumps(date_range) if date_range else '{{"from": "term start", "to": "term end"}}'},
-  "overallRate": <percentage or null if no data provided>,
+  "period": {json.dumps(date_range) if date_range else '{"from": "term start", "to": "term end"}'},
+  "overallRate": <calculated percentage or null if no records provided>,
   "patterns": [
     {{
       "type": "<day_of_week|chronic_absence|weather_related|etc>",
-      "detail": "<description based only on provided data>",
+      "detail": "<description based only on provided records>",
       "severity": "<low|medium|high>",
-      "studentIds": ["<id — only if explicitly provided>"]
+      "studentIds": ["<id from records only>"]
     }}
   ],
   "atRiskStudents": [
     {{
-      "studentId": "<id — only if explicitly provided>",
-      "studentName": "<name — only if explicitly provided, never invented>",
-      "absenceRate": <percentage — only if provided>,
+      "studentId": "<id from records only>",
+      "studentName": "<name from records only — never invented>",
+      "absenceRate": <calculated percentage>,
       "trend": "<improving|stable|worsening>",
       "recommendation": "<action>"
     }}
   ],
-  "summary": "<overall summary based only on provided data, no invented names or IDs>"
+  "summary": "<overall summary derived from the provided records only>"
 }}
 """
         result = gemini_json(prompt, system_instruction=system_instruction)
 
-        # Sanitise: pin scope fields to request values so they cannot be hallucinated
+        # Sanitise: pin scope fields to request values
         if result:
             result['scope'] = scope
             result['scopeId'] = scope_id or ''