assess-replies: media fixes, faster grading, debugging knobs

canvigator.py

@@ -47,7 +47,7 @@ def print_help():
     ch.print_global_help(task_groups)
 
 
-def _run_assignment_task(task, course, canvas, canv_config, dry_run, auto_grade_flag):
+def _run_assignment_task(task, course, canvas, canv_config, dry_run, auto_grade_flag, save_transcript=False):
     """Dispatch the media-recording-assignment tasks."""
     import canvigator_assignment as ca
     if task == 'create-media-recording-assignment':
@@ -59,7 +59,7 @@ def _run_assignment_task(task, course, canvas, canv_config, dry_run, auto_grade_
     if task == 'analyze-media-recordings':
         cassign.analyzeRecordings()
     else:
-        cassign.getMediaRecordings(auto_grade=auto_grade_flag, dry_run=dry_run)
+        cassign.getMediaRecordings(auto_grade=auto_grade_flag, dry_run=dry_run, save_transcript=save_transcript)
 
 
 # Course-level tasks that need only `course` + a few primitives. Kept in a
@@ -87,7 +87,7 @@ def _run_course_task(task, course, canv_config, n_days, cloud_questions):
         cd.prepClassDigest(course, days=n_days, cloud_questions=cloud_questions)
 
 
-def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False):
+def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False, choose_model=False, save_transcript=False):
     """Dispatch a quiz-level task to the appropriate method."""
     if task == 'get-quiz-questions':
         quiz.getQuizQuestions(tag=tag)
@@ -96,7 +96,14 @@ def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False):
         quiz.getAllSubmissionsAndEvents()
         quiz.generateFirstAttemptHistograms()
     elif task == 'assess-replies':
-        quiz.assessFollowUpReplies(reply_window_days=reply_window_days)
+        chosen_model = None
+        if choose_model:
+            import canvigator_llm
+            chosen_model = canvigator_llm.pickLocalGemmaModel()
+        quiz.assessFollowUpReplies(
+            reply_window_days=reply_window_days, chosen_model=chosen_model,
+            save_transcript=save_transcript,
+        )
     elif task == 'send-quiz-reminder':
         quiz.sendQuizReminders(dry_run=dry_run, send_all=send_all)
     elif task == 'send-follow-up-question':
@@ -133,6 +140,8 @@ def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False):
     '-n': '--days',
     '-q': '--cloud-questions',
     '-s': '--send-all',
+    '-M': '--choose-model',
+    '-S': '--save-transcript',
     '-h': '--help',
 }
 args = [_SHORT_TO_LONG.get(a, a) for a in sys.argv[1:]]
@@ -173,6 +182,14 @@ def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False):
 if cloud_questions_flag:
     args.remove('--cloud-questions')
 
+choose_model_flag = '--choose-model' in args
+if choose_model_flag:
+    args.remove('--choose-model')
+
+save_transcript_flag = '--save-transcript' in args
+if save_transcript_flag:
+    args.remove('--save-transcript')
+
 crn = None
 if '--crn' in args:
     crn_idx = args.index('--crn')
@@ -358,7 +375,7 @@ def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False):
     course.sendAllQuizReminders(dry_run=dry_run, send_all=send_all)
 
 elif task in ('create-media-recording-assignment', 'get-media-recordings', 'analyze-media-recordings'):
-    _run_assignment_task(task, course, canvas, canv_config, dry_run, auto_grade_flag)
+    _run_assignment_task(task, course, canvas, canv_config, dry_run, auto_grade_flag, save_transcript=save_transcript_flag)
 
 elif task == 'generate-follow-up-questions':
     # Driven by a pre-selected tagged-questions CSV, not a Canvas quiz
@@ -392,6 +409,9 @@ def _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=False):
     print(f"\nSelected quiz: {quiz_choice.title}")
     skip = task in ('get-quiz-questions', 'assess-replies', 'send-follow-up-assessments')
     quiz = cq.CanvigatorQuiz(canvas, course, quiz_choice, canv_config, verbose=False, skip_student_data=skip)
-    _run_quiz_task(task, quiz, dry_run, tag, reply_window_days, send_all=send_all)
+    _run_quiz_task(
+        task, quiz, dry_run, tag, reply_window_days,
+        send_all=send_all, choose_model=choose_model_flag, save_transcript=save_transcript_flag,
+    )
 
 print("\n*** Done ***\n")

canvigator_assignment.py

@@ -355,7 +355,7 @@ def _writeRecordingsCsv(self, rows):
         df.to_csv(csv_path, index=False)
         return csv_path, df
 
-    def getMediaRecordings(self, auto_grade=False, dry_run=False):
+    def getMediaRecordings(self, auto_grade=False, dry_run=False, save_transcript=False):
         """Fetch every submission, transcribe locally, optionally review and grade interactively.
 
         For each submitter, ffmpeg pulls the media URL directly (handling DASH
@@ -410,7 +410,7 @@ def getMediaRecordings(self, auto_grade=False, dry_run=False):
                     rows.append(self._buildRecordingRow(sub, student_name, audio_rel, ""))
                     continue
                 print(" transcribing...", end="", flush=True)
-                transcript = transcribe_audio(str(local_path), client, audio_model)
+                transcript = transcribe_audio(str(local_path), client, audio_model, save_transcript=save_transcript)
                 chars = len(transcript)
                 print(f" done ({chars} chars)")
 

canvigator_course.py

@@ -1154,6 +1154,12 @@ def exportAnonymizedData(data_path):
     n_total = len(csv_files)
     print(f"Exported {n_total} file{'s' if n_total != 1 else ''} ({n_anonymized} anonymized) to {anon_dir.name}/")
     logger.info(f"Exported {n_total} anonymized files to {zip_path}.zip")
+    print(
+        "\nNOTE: This export is NOT automatically privacy-safe. Before sharing, "
+        "manually verify the CSVs contain no student IDs or names, and no "
+        "open-ended text (transcripts, replies, feedback, conversation subjects), "
+        "audio, or images."
+    )
 
 
 def _renderQuestionPreview(question_dict):

canvigator_help.py

@@ -68,6 +68,18 @@
         "Requires a matching --dry-run run on the same task within the last "
         "10 minutes (otherwise refuses). Mutually exclusive with --dry-run.",
     ),
+    '--choose-model': (
+        '-M', None,
+        "Interactively pick which local gemma4 model to use for grading; "
+        "defaults to OLLAMA_MODEL. Useful for trading some assessment quality "
+        "for throughput (e.g. swap gemma4:31b for gemma4:e4b).",
+    ),
+    '--save-transcript': (
+        '-S', None,
+        "Save each audio transcription as a sibling .txt file alongside the "
+        "audio (debug aid for noisy or mistranscribed recordings — listen + "
+        "read side by side).",
+    ),
 }
 
 
@@ -571,10 +583,12 @@
             "data/<course>/quiz<id>_followup_assessments.csv (persistent; "
             "instructor edits 'feedback' before send-follow-up-assessments).",
         ],
-        'flags': ['--reply-window-days'],
+        'flags': ['--reply-window-days', '--choose-model', '--save-transcript'],
         'examples': [
             "python canvigator.py --crn 12345 assess-replies",
             "python canvigator.py --crn 12345 -w 10 assess-replies",
+            "python canvigator.py --crn 12345 -M assess-replies",
+            "python canvigator.py --crn 12345 -S assess-replies",
         ],
         'run_before': ['send-follow-up-question'],
         'run_after': ['send-follow-up-assessments', 'prep-class-digest'],
@@ -675,11 +689,12 @@
             "audio_path, transcript, transcribed_at, grade, graded_at).",
             "Canvas: posted_grade on each submission (skipped under --dry-run).",
         ],
-        'flags': ['--auto-grade', '--dry-run'],
+        'flags': ['--auto-grade', '--dry-run', '--save-transcript'],
         'examples': [
             "python canvigator.py --crn 12345 get-media-recordings",
             "python canvigator.py --crn 12345 --auto-grade --dry-run "
             "get-media-recordings",
+            "python canvigator.py --crn 12345 -S get-media-recordings",
         ],
         'run_before': ['create-media-recording-assignment'],
         'run_after': ['analyze-media-recordings', 'prep-class-digest'],