✨ Add ai correction api endpoint

env.d/development/common.dist

@@ -39,3 +39,6 @@ LOGOUT_REDIRECT_URL=http://localhost:3000
 
 OIDC_REDIRECT_ALLOWED_HOSTS=["http://localhost:8083", "http://localhost:3000"]
 OIDC_AUTH_REQUEST_EXTRA_PARAMS={"acr_values": "eidas1"}
+
+AI_BASE_URL=https://openaiendpoint.com
+AI_API_KEY=password

src/backend/core/api/viewsets.py

@@ -1,5 +1,6 @@
 """API endpoints"""
 
+import json
 import os
 import re
 import uuid
@@ -17,6 +18,9 @@ from django.db.models import (
 from django.http import Http404
 
 from botocore.exceptions import ClientError
+
+from openai import OpenAI
+
 from rest_framework import (
     decorators,
     exceptions,
@@ -371,8 +375,7 @@ class DocumentViewSet(
     @decorators.action(detail=True, methods=["post"], url_path="ai")
     def ai(self, request, *args, **kwargs):
         """
-        Return the document's versions but only those created after the user got access
-        to the document
+        Process text using AI based on the specified action.
         """
         if not request.user.is_authenticated:
             raise exceptions.PermissionDenied("Authentication required.")
@@ -380,7 +383,26 @@ class DocumentViewSet(
         action = request.data.get("action")
         text = request.data.get("text")
 
-        return drf_response.Response(text.upper())
+        client = OpenAI(
+            base_url = settings.AI_BASE_URL,
+            api_key= settings.AI_API_KEY
+        )
+
+        if action == "rephrase":
+            try:
+                response = client.chat.completions.create(
+                    model="meta-llama/Meta-Llama-3.1-70B-Instruct",
+                    messages=[
+                        {"role": "system", "content": 'Tu es un correcteur de texte. Corrige uniquement la grammaire et l\'orthographe de la phrase donnée. Renvoie uniquement un JSON au format suivant: {"phrase_corrigee": "ta phrase corrigée"}. Ne donne aucune autre information.'},
+                        {"role": "user", "content": f'{{"phrase": "{text}"}}'},
+                    ]
+                )
+                corrected_response = json.loads(response.choices[0].message.content)
+                return drf_response.Response(corrected_response['phrase_corrigee'])
+            except (json.JSONDecodeError, KeyError) as e:
+                return drf_response.Response({"error": f"Error processing AI response: {str(e)}"}, status=500)
+        else:
+            return drf_response.Response({"error": "Invalid action"}, status=400)
 
 
     @decorators.action(detail=True, methods=["get"], url_path="versions")

src/backend/impress/settings.py

@@ -387,6 +387,12 @@ class Base(Configuration):
     ALLOW_LOGOUT_GET_METHOD = values.BooleanValue(
         default=True, environ_name="ALLOW_LOGOUT_GET_METHOD", environ_prefix=None
     )
+    AI_BASE_URL = values.Value(
+        None, environ_name="AI_BASE_URL", environ_prefix=None
+    )
+    AI_API_KEY = values.Value(
+        None, environ_name="AI_API_KEY", environ_prefix=None
+    )
 
     # pylint: disable=invalid-name
     @property

src/backend/pyproject.toml

@@ -47,6 +47,7 @@ dependencies = [
     "jsonschema==4.23.0",
     "markdown==3.7",
     "nested-multipart-parser==1.5.0",
+    "openai==1.44.1",
     "psycopg[binary]==3.2.1",
     "PyJWT==2.9.0",
     "pypandoc==1.13",

src/frontend/apps/impress/src/features/docs/doc-editor/components/AIButton.tsx

@@ -36,6 +36,8 @@ export function AIButton({ doc }: AIButtonProps) {
       action,
     });
 
+    console.log('AI response:', newText);
+
     editor.insertInlineContent([
       newText,
       //{ type: 'text', text: 'World', styles: { bold: true } },