initial commit/ pipeline

.gitignore

@@ -0,0 +1,48 @@
+# ======================
+# Python
+# ======================
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+
+# Virtual environments
+enisper/
+venv/
+*.env/
+
+# ======================
+# Audio / Video / Media
+# ======================
+*.mp3
+*.mp4
+*.wav
+
+# ======================
+# Data / Outputs
+# ======================
+input/
+output/
+summary/
+*.json
+*.txt
+
+# ======================
+# Archives
+# ======================
+*.zip
+
+# ======================
+# OS / Editor
+# ======================
+.DS_Store
+Thumbs.db
+.vscode/
+
+# ======================
+# Experimental scripts
+# ======================
+2moodle.py
+app.old_inside_pipeline.py
+feedback_bewertung.py
+

pipeline/osce_pipeline.py

@@ -0,0 +1,229 @@
+# %% s2t iterative
+# ------------------------------------------------------------
+# osce_pipeline.py
+# ------------------------------------------------------------
+import os
+from pathlib import Path
+import re
+from dotenv import load_dotenv
+
+# Load environment variables from .env file
+load_dotenv()
+
+API_KEY   = os.getenv("OPENAI_API_KEY")
+BASE_URL  = os.getenv("OPENAI_BASE_URL")
+
+# Validate that credentials are loaded
+if not API_KEY or not BASE_URL:
+    raise ValueError("Please set OPENAI_API_KEY and OPENAI_BASE_URL in your .env file")
+
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=API_KEY,
+    base_url=BASE_URL,
+    timeout=600,
+)
+
+def sanitize_filename(filename):
+    """Remove or replace invalid characters from filename"""
+    # Replace problematic characters with underscores
+    return re.sub(r'[<>:"/\\|?*]', '_', filename)
+
+def create_output_structure(input_dir, output_admin_dir, output_student_dir):
+    """Create the required output directory structure for both admin and student folders"""
+    # Get all prüfling directories
+    pruefling_dirs = [d for d in input_dir.iterdir() if d.is_dir()]
+
+    for pruefling_dir in pruefling_dirs:
+        # Extract pruefling number from directory name (e.g., prfling12345 -> 12345)
+        pruefling_id = pruefling_dir.name.replace('prfling', '')
+
+        # Create output directories for both admin and student
+        admin_pruefling_dir = Path(output_admin_dir) / f"Prüfling_{pruefling_id}"
+        student_pruefling_dir = Path(output_student_dir) / f"Prüfling_{pruefling_id}"
+
+        # Create directories for admin
+        admin_pruefling_dir.mkdir(parents=True, exist_ok=True)
+        (admin_pruefling_dir / "Transkripte").mkdir(exist_ok=True)
+        (admin_pruefling_dir / "Zusammenfassungen").mkdir(exist_ok=True)
+
+        # Create directories for student
+        student_pruefling_dir.mkdir(parents=True, exist_ok=True)
+        (student_pruefling_dir / "Zusammenfassungen").mkdir(exist_ok=True)
+
+        print(f"Created structure for Prüfling_{pruefling_id}")
+
+def process_pruefling(pruefling_dir, output_admin_dir, output_student_dir):
+    """Process all MP3 files in a pruefling directory and create both admin and student outputs"""
+    # Extract pruefling number from directory name
+    pruefling_id = pruefling_dir.name.replace('prfling', '')
+
+    # Define output directories for this pruefling
+    admin_pruefling_dir = Path(output_admin_dir) / f"Prüfling_{pruefling_id}"
+    student_pruefling_dir = Path(output_student_dir) / f"Prüfling_{pruefling_id}"
+
+    transcribe_admin_dir = admin_pruefling_dir / "Transkripte"
+    summary_admin_dir = admin_pruefling_dir / "Zusammenfassungen"
+    summary_student_dir = student_pruefling_dir / "Zusammenfassungen"
+
+    # Get all MP3 files from input directory
+    mp3_files = list(pruefling_dir.glob("*.mp3"))
+
+    if not mp3_files:
+        print(f"No MP3 files found in {pruefling_dir}")
+        return
+
+    print(f"Found {len(mp3_files)} MP3 files for Prüfling_{pruefling_id}.")
+
+    # Process each MP3 file
+    for demo_mp3_path in mp3_files:
+        try:
+            print(f"\nProcessing: {demo_mp3_path.name}")
+
+            with demo_mp3_path.open("rb") as audio_file:
+                transcription = client.audio.transcriptions.create(
+                    model="faster-whisper-large-v3",
+                    file=audio_file,
+                )
+
+            # Get base filename without extension
+            base_filename = demo_mp3_path.stem
+
+            # Create safe output filenames
+            txt_filename = sanitize_filename(f"{base_filename}.txt")
+            summary_filename = sanitize_filename(f"{base_filename}-summary.txt")
+
+            # Save transcription to admin directory
+            transcript_output_path = transcribe_admin_dir / txt_filename
+            with open(transcript_output_path, 'w', encoding='utf-8') as f:
+                f.write(transcription.text)
+
+            print(f"Transcript saved to {transcript_output_path}")
+
+            # Perform summarization
+            response = client.chat.completions.create(
+                messages=[
+                    {
+                        "role": "system",
+                        "content": "You are a helpful assistant."
+                    },
+                    {
+                        "role": "user",
+                        "content": f'''
+Aufgabe: Fasse das folgende Prüferinnen-Feedback aus einer medizinischen Prüfung präzise, vollständig und übersichtlich zusammen.
+
+Ziel: Eine konsolidierte, neutrale und strukturierte Zusammenfassung, die alle wesentlichen Inhalte widerspiegelt.
+
+Bitte führe Folgendes durch:
+
+Kernthemen des Feedbacks identifizieren.
+
+Stärken der geprüften Person klar benennen.
+
+Verbesserungspotenziale mit konkretem Bezug zu Inhalten oder Verhalten hervorheben.
+
+Fachliche Aspekte getrennt von kommunikativen/sozialen Kompetenzen darstellen.
+
+Eventuelle konkrete Beispiele oder Situationen aus dem Feedback erwähnen.
+
+Widersprüche oder unterschiedliche Sichtweisen zwischen Prüferinnen markieren.
+
+Eine kurze abschließende Meta-Synthese formulieren, die die wichtigsten Prioritäten für zukünftige Prüfungen ableitet.
+
+Format der Antwort:
+
+Thematische Zusammenfassung (Stichpunkte)
+
+Stärken (Stichpunkte)
+
+Verbesserungspotenziale (Stichpunkte)
+
+Auffällige Unterschiede im Feedback (falls vorhanden)
+
+Priorisierte Handlungsempfehlungen (max. 5 Punkte)
+
+Hier ist das Feedback:
+\n\n{transcription.text}'''
+                    }
+                ],
+                model="GPT-OSS-120B",
+                max_tokens=1024,
+                temperature=0.3
+            )
+
+            # Save summary to admin directory
+            summary_output_path = summary_admin_dir / summary_filename
+            with open(summary_output_path, "w", encoding='utf-8') as f:
+                content = response.choices[0].message.content or ""
+                f.write(content)
+
+            print(f"Summary saved to {summary_output_path}")
+
+            # Copy summary to student directory
+            student_summary_path = summary_student_dir / summary_filename
+            with open(student_summary_path, "w", encoding='utf-8') as f:
+                content = response.choices[0].message.content or ""
+                f.write(content)
+
+            print(f"Summary copied to student directory: {student_summary_path}")
+
+        except Exception as e:
+            print(f"Error processing {demo_mp3_path.name}: {str(e)}")
+
+def main():
+    # Ask for input directory
+    input_base_dir = Path(input("Enter the input directory path: "))
+
+    # Validate input directory exists
+    if not input_base_dir.exists():
+        print(f"Error: Input directory {input_base_dir} does not exist!")
+        return
+
+    if not input_base_dir.is_dir():
+        print(f"Error: {input_base_dir} is not a directory!")
+        return
+
+    # Create .env file template if it doesn't exist
+    env_file = Path(".env")
+    if not env_file.exists():
+        with open(env_file, "w") as f:
+            f.write("# OpenAI Configuration\n")
+            f.write("OPENAI_API_KEY=your_api_key_here\n")
+            f.write("OPENAI_BASE_URL=https://api.openai.com/v1\n")
+        print("Created .env file. Please fill in your credentials.")
+        return
+
+    # Get output directory names from environment variables
+    output_admin_name = os.getenv("OUTPUT_ADMIN_DIR", "output_admin")
+    output_student_name = os.getenv("OUTPUT_STUDENT_DIR", "output_student")
+
+    # Create parent directory for outputs
+    parent_output_dir = input_base_dir.parent
+
+    # Define output directories
+    output_admin_dir = parent_output_dir / output_admin_name
+    output_student_dir = parent_output_dir / output_student_name
+
+    print(f"Input directory: {input_base_dir}")
+    print(f"Admin output directory: {output_admin_dir}")
+    print(f"Student output directory: {output_student_dir}")
+
+    # Create output directories
+    output_admin_dir.mkdir(exist_ok=True)
+    output_student_dir.mkdir(exist_ok=True)
+
+    # Create directory structure
+    create_output_structure(input_base_dir, output_admin_dir, output_student_dir)
+
+    # Process all pruefling directories
+    for pruefling_dir in input_base_dir.iterdir():
+        if pruefling_dir.is_dir() and pruefling_dir.name.startswith('prfling'):
+            print(f"\nProcessing {pruefling_dir.name}")
+            process_pruefling(pruefling_dir, output_admin_dir, output_student_dir)
+
+    print("\nProcessing complete!")
+
+if __name__ == "__main__":
+    main()
+