Miscs

gemini_for_enonce.py

@@ -1,4 +1,6 @@
+import shlex
 import os
+import subprocess
 import sys
 import argparse
 from pathlib import Path
@@ -7,7 +9,9 @@ from typing import List
 from google import genai
 from google.genai import types
 
-MODEL_ID = "gemini-3-flash-preview"
+# Bug : l'output est limité à 8k token…
+# MODEL_ID = "gemini-3-flash-preview"
+MODEL_ID = "gemini-3.1-flash-lite"
 api_key = os.environ.get("GEMINI_API_KEY")
 
 class QuestionItem(BaseModel):
@@ -84,14 +88,25 @@ def process_exam(folder_path: str):
         response_json_schema=ExamExtraction.model_json_schema(),
     )
 
-    print("Sending request to Gemini...")
-    response = client.models.generate_content(
-        model=MODEL_ID,
-        contents=contents,
-        config=config
-    )
+    cache_file = folder / "gemini_response.json"
 
-    extracted_data = ExamExtraction.model_validate_json(response.text)
+    if cache_file.is_file():
+        print("Loading cached response from gemini_response.json...")
+        response_text = cache_file.read_text(encoding="utf-8")
+    else:
+        print("Sending request to Gemini...")
+        response = client.models.generate_content(
+            model=MODEL_ID,
+            contents=contents,
+            config=config
+        )
+        response_text = response.text
+
+        print("Saving response to cache...")
+        cache_file.write_text(response_text, encoding="utf-8")
+
+    # Validate from the text variable (cached or fresh)
+    extracted_data = ExamExtraction.model_validate_json(response_text)
 
     # 2. Setup output directories
     text_dir = folder / "Text"
@@ -101,28 +116,80 @@ def process_exam(folder_path: str):
 
     labels_file = folder / "labels"
 
-    print("Writing files...")
+    # Step 1: Write initial labels
+    print("Writing initial labels file...")
     with open(labels_file, "w", encoding="utf-8") as flabels:
         for q in extracted_data.questions:
-            # Sanitize label for filesystem (prevent directory traversal if label contains '/')
-            safe_label = q.label.replace("/", "_")
+            flabels.write(f"{q.label}\n")
 
-            flabels.write(f"{safe_label}\n")
+    # Step 2: Open labels file for user editing
+    print("Opening labels file for editing...")
+    editor = os.environ.get("EDITOR")
+    try:
+        if editor:
+            subprocess.run(shlex.split(editor) + [str(labels_file)])
+        else:
+        # Fallbacks if $EDITOR is not set
+            if sys.platform.startswith("linux"):
+                subprocess.Popen(["xdg-open", str(labels_file)])
+            elif sys.platform == "darwin":
+                subprocess.Popen(["open", str(labels_file)])
+            else:
+                os.startfile(str(labels_file))
 
-            # Fix double-escaped newlines
-            q_content = q.question_content.replace("\\n", "\n")
-            s_content = q.solution_content.replace("\\n", "\n")
-            
+        # xdg-open/open usually do not block, so we wait for user confirmation
+        input("Press ENTER here once you have saved and closed the labels file...")
+    except Exception:
+        print("Error running editor, using labels as given.")
 
-            # Write Text/label
-            with open(text_dir / safe_label, "w", encoding="utf-8") as f:
-                f.write(f"{q.label}\n{q.question_content}")
+    # Step 3 & 4: Read the edited file back and create a mapping
+    with open(labels_file, "r", encoding="utf-8") as flabels:
+        edited_lines = [line.strip() for line in flabels if line.strip()]
 
-            # Write Sol/label
-            with open(sol_dir / safe_label, "w", encoding="utf-8") as f:
-                f.write(f"{q.label}\n{q.solution_content}")
+    mapping = []
+    final_labels = []
+    orig_idx = 0
 
-    print(f"Success! Processed {len(extracted_data.questions)} questions.")
+    for line in edited_lines:
+        if line.startswith("+"):
+            new_label = line[1:].lstrip()
+            final_labels.append(new_label)
+            # New label, no source content
+            mapping.append((new_label, None))
+        else:
+            new_label = line
+            final_labels.append(new_label)
+            # Map to initial order, advancing index only for non-'+' items
+            q_item = extracted_data.questions[orig_idx] if orig_idx < len(extracted_data.questions) else None
+            mapping.append((new_label, q_item))
+            orig_idx += 1
+
+    # Rewrite the labels file cleanly (removing '+' prefixes)
+    with open(labels_file, "w", encoding="utf-8") as flabels:
+        for lbl in final_labels:
+            flabels.write(f"{lbl}\n")
+
+    # Step 5: Write the final question and solution files
+    print("Writing question and solution files...")
+    for new_label, q_item in mapping:
+        safe_label = new_label.replace("/", "_")
+
+        if q_item:
+            q_content = q_item.question_content.replace("\\n", "\n")
+            s_content = q_item.solution_content.replace("\\n", "\n")
+        else:
+            q_content = ""
+            s_content = ""
+
+        # Write Text/label
+        with open(text_dir / safe_label, "w", encoding="utf-8") as f:
+            f.write(f"{new_label}\n{q_content}")
+
+        # Write Sol/label
+        with open(sol_dir / safe_label, "w", encoding="utf-8") as f:
+            f.write(f"{new_label}\n{s_content}")
+
+    print(f"Success! Processed {len(mapping)} labels.")
 
 if __name__ == "__main__":
     if not api_key: