WIP

Set SEARCH_INDEXER_CLASS=None as default configuration for dev.
Rename docker network 'lasuite-net' as 'lasuite' to match with Drive configuration.

Signed-off-by: Fabre Florian <ffabre@hybird.org>

.gitguardian.yaml

@@ -0,0 +1,5 @@
+secret:
+  ignored_matches:
+  - name:
+    match: "na1hhus-OLhq9mb9SO3R-8E4dONuMnqpZSY_SX8xcFk="
+version: 2

.gitignore

@@ -43,6 +43,10 @@ venv.bak/
 env.d/development/*.local
 env.d/terraform
 
+# Docker
+compose.override.yml
+docker/auth/*.local
+
 # npm
 node_modules
 

CHANGELOG.md

@@ -74,6 +74,9 @@ and this project adheres to
   - ♿ update labels and shared document icon accessibility #1442
 - 🍱(frontend) Fonts GDPR compliants #1453
 - ♻️(service-worker) improve SW registration and update handling #1473
+- ✨(backend) add async indexation of documents on save (or access save) #1276
+- ✨(backend) add debounce mechanism to limit indexation jobs #1276
+- ✨(api) add API route to search for indexed documents in Find #1276
 
 ### Fixed
 

Makefile

@@ -247,6 +247,10 @@ demo: ## flush db then create a demo for load testing purpose
 	@$(MANAGE) create_demo
 .PHONY: demo
 
+index: ## index all documents to remote search
+	@$(MANAGE) index
+.PHONY: index
+
 # Nota bene: Black should come after isort just in case they don't agree...
 lint: ## lint back-end python sources
 lint: \

bin/fernetkey

@@ -0,0 +1,6 @@
+#!/usr/bin/env bash
+
+# shellcheck source=bin/_config.sh
+source "$(dirname "${BASH_SOURCE[0]}")/_config.sh"
+
+_dc_run app-dev python -c 'from cryptography.fernet import Fernet;import sys; sys.stdout.write("\n" + Fernet.generate_key().decode() + "\n");'

compose.yml

@@ -72,6 +72,11 @@ services:
       - env.d/development/postgresql.local
     ports:
       - "8071:8000"
+    networks:
+      default: {}
+      lasuite:
+        aliases:
+          - impress
     volumes:
       - ./src/backend:/app
       - ./data/static:/data/static
@@ -92,6 +97,9 @@ services:
     command: ["celery", "-A", "impress.celery_app", "worker", "-l", "DEBUG"]
     environment:
       - DJANGO_CONFIGURATION=Development
+    networks:
+      - default
+      - lasuite
     env_file:
       - env.d/development/common
       - env.d/development/common.local
@@ -107,6 +115,11 @@ services:
     image: nginx:1.25
     ports:
       - "8083:8083"
+    networks:
+      default: {}
+      lasuite:
+        aliases:
+          - nginx
     volumes:
       - ./docker/files/etc/nginx/conf.d:/etc/nginx/conf.d:ro
     depends_on:
@@ -217,3 +230,8 @@ services:
       kc_postgresql:
         condition: service_healthy
         restart: true
+
+networks:
+  lasuite:
+    name: lasuite-network
+    driver: bridge

docs/architecture.md

@@ -12,6 +12,7 @@ flowchart TD
     Back --> DB("Database (PostgreSQL)")
     Back <--> Celery --> DB
     Back ----> S3("Minio (S3)")
+    Back -- REST API --> Find
 ```
 
 ### Architecture decision records

docs/env.md

@@ -93,6 +93,12 @@ These are the environment variables you can set for the `impress-backend` contai
 | OIDC_USERINFO_SHORTNAME_FIELD                   | OIDC token claims to create shortname                                                                                       | first_name                                                              |
 | POSTHOG_KEY                                     | Posthog key for analytics                                                                                                   |                                                                         |
 | REDIS_URL                                       | Cache url                                                                                                                   | redis://redis:6379/1                                                    |
+| SEARCH_INDEXER_CLASS                            | Class of the backend for document indexation & search                                                                       |                                                                         |
+| SEARCH_INDEXER_BATCH_SIZE                       | Size of each batch for indexation of all documents                                                                          | 100000                                                                  |
+| SEARCH_INDEXER_COUNTDOWN                        | Minimum debounce delay of indexation jobs (in seconds)                                                                      | 1                                                                       |
+| SEARCH_INDEXER_URL                              | Find application endpoint for indexation                                                                                    |                                                                         |
+| SEARCH_INDEXER_SECRET                           | Token for indexation queries                                                                                                |                                                                         |
+| SEARCH_INDEXER_QUERY_URL                        | Find application endpoint for search                                                                                        |                                                                         |
 | SENTRY_DSN                                      | Sentry host                                                                                                                 |                                                                         |
 | SESSION_COOKIE_AGE                              | duration of the cookie session                                                                                              | 60*60*12                                                                |
 | SPECTACULAR_SETTINGS_ENABLE_DJANGO_DEPLOY_CHECK |                                                                                                                             | false                                                                   |

docs/search.md

@@ -0,0 +1,217 @@
+# Setup full-text search with Find
+
+This configuration will enable the fulltext search feature for Docs :
+- Each save on **core.Document** or **core.DocumentAccess** will trigger the indexer
+- The `api/v1.0/documents/search/` will work as a proxy with the Find API for fulltext search.
+
+## Create an index service for Docs
+
+In Find application Django admin configure a **Service** with these settings
+
+- **Name**: `docs`<br>_request.auth.name of the Docs application._
+
+- **Client id**: `impress`<br>_Name of the token audience or client_id of the Docs application._
+
+## Configure settings
+
+Add those Django settings to enable the feature.
+
+```shell
+SEARCH_INDEXER_CLASS="core.services.search_indexers.FindDocumentIndexer"
+SEARCH_INDEXER_COUNTDOWN=10  # Debounce delay in seconds for the index calls.
+
+# Indexation endpoint.
+SEARCH_INDEXER_SECRET=my-token-from-the-find-impress-service
+# The token from service "docs" of Find application.
+SEARCH_INDEXER_URL="http://app-find:8000/api/v1.0/documents/index/"
+
+# Search endpoint. Uses the OIDC token for authentication
+SEARCH_INDEXER_QUERY_URL="http://app-find:8000/api/v1.0/documents/search/"
+```
+
+We also need to enable the **OIDC Token** refresh or the authentication will fail quickly.
+
+```shell
+# Store OIDC tokens in the session
+OIDC_STORE_ACCESS_TOKEN = True  # Store the access token in the session
+OIDC_STORE_REFRESH_TOKEN = True  # Store the encrypted refresh token in the session
+OIDC_STORE_REFRESH_TOKEN_KEY = "your-32-byte-encryption-key=="  # Must be a valid Fernet key (32 url-safe base64-encoded bytes)
+```
+
+# Installation with docker compose
+
+First see [compose installation](installation/compose.md) documentation.
+
+## 1. Postgresql
+
+Find uses PostgreSQL as its database.
+
+If you are using the example provided, you need to generate a secure key for `DB_PASSWORD` and set it in `env.d/find_postgresql`.
+
+```shell
+# Postgresql db container configuration
+POSTGRES_DB=find
+POSTGRES_USER=dinum
+POSTGRES_PASSWORD=pass
+
+# App database configuration
+DB_HOST=postgresql-find
+DB_NAME=find
+DB_USER=dinum
+DB_PASSWORD=<password>
+DB_PORT=5432
+```
+
+And the service can be configured in `compose.yml`.
+
+```yaml
+  postgresql-find:
+    image: postgres:15
+    env_file:
+      - env.d/development/find_postgresql
+      - env.d/development/find_postgresql.local
+    ports:
+      - "15433:5432"
+```
+
+## 2. Opensearch
+
+Find uses Opensearch as its indexation database.
+
+If you are using the example provided, you need to generate a secure key for `OPENSEARCH_INITIAL_ADMIN_PASSWORD` and `OPENSEARCH_PASSWORD`
+to set them in `env.d/opensearch`.
+
+```shell
+OPENSEARCH_INITIAL_ADMIN_PASSWORD=find
+OPENSEARCH_PASSWORD=find
+OPENSEARCH_USE_SSL=true  # false in Development mode
+
+# Dashboard
+OPENSEARCH_HOSTS=["http://opensearch:9200"]
+DISABLE_SECURITY_DASHBOARDS_PLUGIN=false  # true in Development mode
+```
+
+And the service & dashboard can be configured in `compose.yml`.
+
+```yaml
+  opensearch:
+    user: ${DOCKER_USER:-1000}
+    image: opensearchproject/opensearch:latest
+    env_file:
+      - env.d/development/opensearch
+      - env.d/development/opensearch.local
+    environment:
+      - discovery.type=single-node
+      - plugins.security.disabled=true
+      - plugins.security.ssl.http.enabled=false
+    ulimits:
+      memlock:
+        soft: -1
+        hard: -1
+    volumes:
+      - ./data/opensearch:/usr/share/opensearch/data
+    ports:
+      - "9200:9200"
+      - "9600:9600"
+
+  opensearch-dashboards:
+    image: opensearchproject/opensearch-dashboards:latest
+    ports:
+      - "5601:5601"
+    env_file:
+      - env.d/development/opensearch
+      - env.d/development/opensearch.local
+    depends_on:
+      - opensearch
+```
+
+## 3. Find (dev mode)
+
+If you are using the example provided, you need to generate a secure key for `OIDC_RS_CLIENT_SECRET` and set it in `env.d/find`.
+
+```shell
+# Django
+DJANGO_ALLOWED_HOSTS=*
+DJANGO_SECRET_KEY=ThisIsAnExampleKeyForDevPurposeOnly
+DJANGO_SETTINGS_MODULE=find.settings
+DJANGO_SUPERUSER_PASSWORD=admin
+
+# Python
+PYTHONPATH=/app
+
+# find settings
+
+# Backend url
+FIND_BASE_URL="http://localhost:9072"
+
+# OIDC
+OIDC_OP_URL=http://localhost:8083/realms/impress
+OIDC_OP_INTROSPECTION_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/token/introspect
+
+OIDC_OP_JWKS_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/certs
+OIDC_OP_AUTHORIZATION_ENDPOINT=http://localhost:8083/realms/impress/protocol/openid-connect/auth
+OIDC_OP_TOKEN_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/token
+OIDC_OP_USER_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/userinfo
+
+OIDC_RP_CLIENT_ID=impress
+OIDC_RP_CLIENT_SECRET=ThisIsAnExampleKeyForDevPurposeOnly
+OIDC_RP_SIGN_ALGO=RS256
+OIDC_RP_SCOPES="openid email"
+
+OIDC_REDIRECT_ALLOWED_HOSTS=["http://localhost:8083", "http://localhost:3000"]
+OIDC_AUTH_REQUEST_EXTRA_PARAMS={"acr_values": "eidas1"}
+
+# OIDC Resource server
+OIDC_RS_SCOPES="openid"
+OIDC_RS_CLIENT_ID=impress
+OIDC_RS_CLIENT_SECRET=ThisIsAnExampleKeyForDevPurposeOnly
+OIDC_RS_SIGN_ALGO=RS256
+
+OIDC_RS_BACKEND_CLASS="core.authentication.FinderResourceServerBackend"
+```
+
+And Find can be configured in `compose.yml`.
+
+```yaml
+  find-dev:
+    user: ${DOCKER_USER:-1000}
+    image: find:backend-development
+    environment:
+      - PYLINTHOME=/app/.pylint.d
+      - DJANGO_CONFIGURATION=Development
+    env_file:
+      - env.d/development/find
+      - env.d/development/find.local
+      - env.d/development/find_postgresql
+      - env.d/development/find_postgresql.local
+      - env.d/development/opensearch
+      - env.d/development/opensearch.local
+    ports:
+      - "9071:8000"
+    volumes:
+      - ../find/src/backend:/app
+      - ../find/data/static:/data/static
+    depends_on:
+        - find_postgresql
+        - opensearch
+        - redis
+```
+
+## Ports (dev defaults)
+
+| Port      | Service               |
+| --------- | --------------------- |
+| 3000      | Next.js               |
+| 8071      | Django (main)         |
+| 9071      | Django (Find)         |
+| 4444      | Y-Provider            |
+| 8080      | Keycloak              |
+| 8083      | Nginx proxy           |
+| 9000/9001 | MinIO                 |
+| 15432     | PostgreSQL (main)     |
+| 15433     | PostgreSQL (Find)     |
+| 5433      | PostgreSQL (Keycloak) |
+| 1081      | MailCatcher           |
+| 9200      | Opensearch            |
+| 9600      | Opensearch admin      |
+| 5601      | Opensearch dashboard  |

env.d/development/common

@@ -36,6 +36,7 @@ OIDC_OP_JWKS_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/c
 OIDC_OP_AUTHORIZATION_ENDPOINT=http://localhost:8083/realms/impress/protocol/openid-connect/auth
 OIDC_OP_TOKEN_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/token
 OIDC_OP_USER_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/userinfo
+OIDC_OP_INTROSPECTION_ENDPOINT=http://nginx:8083/realms/impress/protocol/openid-connect/token/introspect
 
 OIDC_RP_CLIENT_ID=impress
 OIDC_RP_CLIENT_SECRET=ThisIsAnExampleKeyForDevPurposeOnly
@@ -49,6 +50,14 @@ LOGOUT_REDIRECT_URL=http://localhost:3000
 OIDC_REDIRECT_ALLOWED_HOSTS=["http://localhost:8083", "http://localhost:3000"]
 OIDC_AUTH_REQUEST_EXTRA_PARAMS={"acr_values": "eidas1"}
 
+# Store OIDC tokens in the session
+OIDC_STORE_ACCESS_TOKEN = True
+OIDC_STORE_REFRESH_TOKEN = True # Store the encrypted refresh token in the session.
+
+# Must be a valid Fernet key (32 url-safe base64-encoded bytes)
+# To create one, use the bin/fernetkey command.
+OIDC_STORE_REFRESH_TOKEN_KEY = "na1hhus-OLhq9mb9SO3R-8E4dONuMnqpZSY_SX8xcFk="
+
 # AI
 AI_FEATURE_ENABLED=true
 AI_BASE_URL=https://openaiendpoint.com
@@ -68,4 +77,10 @@ Y_PROVIDER_API_BASE_URL=http://y-provider-development:4444/api/
 Y_PROVIDER_API_KEY=yprovider-api-key
 
 # Theme customization
-THEME_CUSTOMIZATION_CACHE_TIMEOUT=15
+THEME_CUSTOMIZATION_CACHE_TIMEOUT=15
+
+# Indexer (disabled)
+# SEARCH_INDEXER_CLASS="core.services.search_indexers.SearchIndexer"
+SEARCH_INDEXER_SECRET=find-api-key-for-docs-with-exactly-50-chars-length  # Key generated by create_demo in Find app.
+SEARCH_INDEXER_URL="http://find:8000/api/v1.0/documents/index/"
+SEARCH_INDEXER_QUERY_URL="http://find:8000/api/v1.0/documents/search/"

src/backend/core/api/serializers.py

@@ -889,3 +889,13 @@ class MoveDocumentSerializer(serializers.Serializer):
         choices=enums.MoveNodePositionChoices.choices,
         default=enums.MoveNodePositionChoices.LAST_CHILD,
     )
+
+
+class SearchDocumentSerializer(serializers.Serializer):
+    """Serializer for fulltext search requests through Find application"""
+
+    q = serializers.CharField(required=True, allow_blank=False, trim_whitespace=True)
+    page_size = serializers.IntegerField(
+        required=False, min_value=1, max_value=50, default=20
+    )
+    page = serializers.IntegerField(required=False, min_value=1, default=1)

src/backend/core/api/viewsets.py

@@ -14,6 +14,7 @@ from django.contrib.postgres.search import TrigramSimilarity
 from django.core.cache import cache
 from django.core.exceptions import ValidationError
 from django.core.files.storage import default_storage
+from django.core.paginator import InvalidPage, Paginator
 from django.core.validators import URLValidator
 from django.db import connection, transaction
 from django.db import models as db
@@ -21,6 +22,7 @@ from django.db.models.expressions import RawSQL
 from django.db.models.functions import Left, Length
 from django.http import Http404, StreamingHttpResponse
 from django.urls import reverse
+from django.utils.decorators import method_decorator
 from django.utils.functional import cached_property
 from django.utils.text import capfirst, slugify
 from django.utils.translation import gettext_lazy as _
@@ -31,9 +33,11 @@ from botocore.exceptions import ClientError
 from csp.constants import NONE
 from csp.decorators import csp_update
 from lasuite.malware_detection import malware_detection
+from lasuite.oidc_login.decorators import refresh_oidc_access_token
 from rest_framework import filters, status, viewsets
 from rest_framework import response as drf_response
 from rest_framework.permissions import AllowAny
+from rest_framework.utils.urls import replace_query_param as drf_replace_query_param
 
 from core import authentication, choices, enums, models
 from core.services.ai_services import AIService
@@ -47,6 +51,10 @@ from core.services.converter_services import (
 from core.services.converter_services import (
     YdocConverter,
 )
+from core.services.search_indexers import (
+    get_document_indexer,
+    get_visited_document_ids_of,
+)
 from core.tasks.mail import send_ask_for_access_mail
 from core.utils import extract_attachments, filter_descendants
 
@@ -373,6 +381,7 @@ class DocumentViewSet(
     list_serializer_class = serializers.ListDocumentSerializer
     trashbin_serializer_class = serializers.ListDocumentSerializer
     tree_serializer_class = serializers.ListDocumentSerializer
+    search_serializer_class = serializers.ListDocumentSerializer
 
     def get_queryset(self):
         """Get queryset performing all annotation and filtering on the document tree structure."""
@@ -1064,6 +1073,114 @@ class DocumentViewSet(
             {"id": str(duplicated_document.id)}, status=status.HTTP_201_CREATED
         )
 
+    def _search_simple(self, request, text):
+        """
+        Returns a queryset filtered by the content of the document title
+        """
+        # As the 'list' view we get a prefiltered queryset (deleted docs are excluded)
+        queryset = self.get_queryset()
+        filterset = DocumentFilter({"title": text}, queryset=queryset)
+
+        if not filterset.is_valid():
+            raise drf.exceptions.ValidationError(filterset.errors)
+
+        queryset = filterset.filter_queryset(queryset)
+
+        return self.get_response_for_queryset(
+            queryset.order_by("-updated_at"),
+            context={
+                "request": request,
+            },
+        )
+
+    def _search_fulltext(self, indexer, request, params):
+        """
+        Returns a queryset from the results the fulltext search of Find
+        """
+        access_token = request.session.get("oidc_access_token")
+        user = request.user
+        text = params.validated_data["q"]
+        page_size = params.validated_data.get("page_size", 20)
+        page_number = params.validated_data.get("page", 1)
+        queryset = models.Document.objects.all()
+
+        # Retrieve the documents ids from Find.
+        results = indexer.search(
+            text=text,
+            token=access_token,
+            visited=get_visited_document_ids_of(queryset, user),
+            page=1,
+            page_size=100,
+        )
+
+        docs_by_uuid = {str(d.pk): d for d in queryset.filter(pk__in=results)}
+        ordered_docs = [docs_by_uuid[id] for id in results]
+
+        paginator = Paginator(
+            ordered_docs, per_page=page_size, allow_empty_first_page=True
+        )
+
+        try:
+            page = paginator.page(page_number)
+        except InvalidPage as e:
+            raise drf.exceptions.NotFound(_("Invalid page.")) from e
+
+        serializer = self.get_serializer(
+            page.object_list,
+            many=True,
+            context={
+                "request": request,
+            },
+        )
+        next_url, prev_url = None, None
+
+        if page.has_next():
+            next_url = request.build_absolute_uri()
+            next_url = drf_replace_query_param(
+                next_url, "page", page.next_page_number()
+            )
+
+        if page.has_previous():
+            prev_url = request.build_absolute_uri()
+            prev_url = drf_replace_query_param(
+                prev_url, "page", page.previous_page_number()
+            )
+
+        output = {
+            "count": paginator.count,
+            "next": next_url,
+            "previous": prev_url,
+            "results": serializer.data,
+        }
+
+        return drf.response.Response(output)
+
+    @drf.decorators.action(detail=False, methods=["get"], url_path="search")
+    @method_decorator(refresh_oidc_access_token)
+    def search(self, request, *args, **kwargs):
+        """
+        Returns a DRF response containing the filtered, annotated and ordered document list.
+
+        Applies filtering based on request parameter 'q' from `SearchDocumentSerializer`.
+        Depending of the configuration it can be:
+         - A fulltext search through the opensearch indexation app "find" if the backend is
+           enabled (see SEARCH_INDEXER_CLASS)
+         - A filtering by the model field 'title'.
+
+        The ordering is always by the most recent first.
+        """
+        params = serializers.SearchDocumentSerializer(data=request.query_params)
+        params.is_valid(raise_exception=True)
+
+        indexer = get_document_indexer()
+
+        if indexer:
+            return self._search_fulltext(indexer, request, params=params)
+
+        # The indexer is not configured, we fallback on a simple icontains filter by the
+        # model field 'title'.
+        return self._search_simple(request, text=params.validated_data["q"])
+
     @drf.decorators.action(detail=True, methods=["get"], url_path="versions")
     def versions_list(self, request, *args, **kwargs):
         """

src/backend/core/apps.py

@@ -1,11 +1,19 @@
 """Impress Core application"""
-# from django.apps import AppConfig
-# from django.utils.translation import gettext_lazy as _
+
+from django.apps import AppConfig
+from django.utils.translation import gettext_lazy as _
 
 
-# class CoreConfig(AppConfig):
-#     """Configuration class for the impress core app."""
+class CoreConfig(AppConfig):
+    """Configuration class for the impress core app."""
 
-#     name = "core"
-#     app_label = "core"
-#     verbose_name = _("impress core application")
+    name = "core"
+    app_label = "core"
+    verbose_name = _("Impress core application")
+
+    def ready(self):
+        """
+        Import signals when the app is ready.
+        """
+        # pylint: disable=import-outside-toplevel, unused-import
+        from . import signals  # noqa: PLC0415

src/backend/core/management/commands/index.py

@@ -0,0 +1,40 @@
+"""
+Handle search setup that needs to be done at bootstrap time.
+"""
+
+import logging
+import time
+
+from django.core.management.base import BaseCommand, CommandError
+
+from core.services.search_indexers import get_document_indexer
+
+logger = logging.getLogger("docs.search.bootstrap_search")
+
+
+class Command(BaseCommand):
+    """Index all documents to remote search service"""
+
+    help = __doc__
+
+    def handle(self, *args, **options):
+        """Launch and log search index generation."""
+        indexer = get_document_indexer()
+
+        if not indexer:
+            raise CommandError("The indexer is not enabled or properly configured.")
+
+        logger.info("Starting to regenerate Find index...")
+        start = time.perf_counter()
+
+        try:
+            count = indexer.index()
+        except Exception as err:
+            raise CommandError("Unable to regenerate index") from err
+
+        duration = time.perf_counter() - start
+        logger.info(
+            "Search index regenerated from %d document(s) in %.2f seconds.",
+            count,
+            duration,
+        )

src/backend/core/models.py

@@ -432,32 +432,35 @@ class Document(MP_Node, BaseModel):
     def save(self, *args, **kwargs):
         """Write content to object storage only if _content has changed."""
         super().save(*args, **kwargs)
-
         if self._content:
-            file_key = self.file_key
-            bytes_content = self._content.encode("utf-8")
+            self.save_content(self._content)
 
-            # Attempt to directly check if the object exists using the storage client.
-            try:
-                response = default_storage.connection.meta.client.head_object(
-                    Bucket=default_storage.bucket_name, Key=file_key
-                )
-            except ClientError as excpt:
-                # If the error is a 404, the object doesn't exist, so we should create it.
-                if excpt.response["Error"]["Code"] == "404":
-                    has_changed = True
-                else:
-                    raise
+    def save_content(self, content):
+        """Save content to object storage."""
+
+        file_key = self.file_key
+        bytes_content = content.encode("utf-8")
+
+        # Attempt to directly check if the object exists using the storage client.
+        try:
+            response = default_storage.connection.meta.client.head_object(
+                Bucket=default_storage.bucket_name, Key=file_key
+            )
+        except ClientError as excpt:
+            # If the error is a 404, the object doesn't exist, so we should create it.
+            if excpt.response["Error"]["Code"] == "404":
+                has_changed = True
             else:
-                # Compare the existing ETag with the MD5 hash of the new content.
-                has_changed = (
-                    response["ETag"].strip('"')
-                    != hashlib.md5(bytes_content).hexdigest()  # noqa: S324
-                )
+                raise
+        else:
+            # Compare the existing ETag with the MD5 hash of the new content.
+            has_changed = (
+                response["ETag"].strip('"') != hashlib.md5(bytes_content).hexdigest()  # noqa: S324
+            )
 
-            if has_changed:
-                content_file = ContentFile(bytes_content)
-                default_storage.save(file_key, content_file)
+        if has_changed:
+            content_file = ContentFile(bytes_content)
+            default_storage.save(file_key, content_file)
 
     def is_leaf(self):
         """

src/backend/core/services/search_indexers.py

@@ -0,0 +1,295 @@
+"""Document search index management utilities and indexers"""
+
+import logging
+from abc import ABC, abstractmethod
+from collections import defaultdict
+from functools import cache
+
+from django.conf import settings
+from django.contrib.auth.models import AnonymousUser
+from django.core.exceptions import ImproperlyConfigured
+from django.db.models import Subquery
+from django.utils.module_loading import import_string
+
+import requests
+
+from core import models, utils
+
+logger = logging.getLogger(__name__)
+
+
+@cache
+def get_document_indexer():
+    """Returns an instance of indexer service if enabled and properly configured."""
+    classpath = settings.SEARCH_INDEXER_CLASS
+
+    # For this usecase an empty indexer class is not an issue but a feature.
+    if not classpath:
+        logger.info("Document indexer is not configured (see SEARCH_INDEXER_CLASS)")
+        return None
+
+    try:
+        indexer_class = import_string(settings.SEARCH_INDEXER_CLASS)
+        return indexer_class()
+    except ImportError as err:
+        logger.error("SEARCH_INDEXER_CLASS setting is not valid : %s", err)
+    except ImproperlyConfigured as err:
+        logger.error("Document indexer is not properly configured : %s", err)
+
+    return None
+
+
+def get_batch_accesses_by_users_and_teams(paths):
+    """
+    Get accesses related to a list of document paths,
+    grouped by users and teams, including all ancestor paths.
+    """
+    ancestor_map = utils.get_ancestor_to_descendants_map(
+        paths, steplen=models.Document.steplen
+    )
+    ancestor_paths = list(ancestor_map.keys())
+
+    access_qs = models.DocumentAccess.objects.filter(
+        document__path__in=ancestor_paths
+    ).values("document__path", "user__sub", "team")
+
+    access_by_document_path = defaultdict(lambda: {"users": set(), "teams": set()})
+
+    for access in access_qs:
+        ancestor_path = access["document__path"]
+        user_sub = access["user__sub"]
+        team = access["team"]
+
+        for descendant_path in ancestor_map.get(ancestor_path, []):
+            if user_sub:
+                access_by_document_path[descendant_path]["users"].add(str(user_sub))
+            if team:
+                access_by_document_path[descendant_path]["teams"].add(team)
+
+    return dict(access_by_document_path)
+
+
+def get_visited_document_ids_of(queryset, user):
+    """
+    Returns the ids of the documents that have a linktrace to the user and NOT owned.
+    It will be use to limit the opensearch responses to the public documents already
+    "visited" by the user.
+    """
+    if isinstance(user, AnonymousUser):
+        return []
+
+    qs = models.LinkTrace.objects.filter(user=user)
+
+    docs = (
+        queryset.exclude(accesses__user=user)
+        .filter(
+            deleted_at__isnull=True,
+            ancestors_deleted_at__isnull=True,
+        )
+        .filter(pk__in=Subquery(qs.values("document_id")))
+        .order_by("pk")
+        .distinct("pk")
+    )
+
+    return [str(id) for id in docs.values_list("pk", flat=True)]
+
+
+class BaseDocumentIndexer(ABC):
+    """
+    Base class for document indexers.
+
+    Handles batching and access resolution. Subclasses must implement both
+    `serialize_document()` and `push()` to define backend-specific behavior.
+    """
+
+    def __init__(self, batch_size=None):
+        """
+        Initialize the indexer.
+
+        Args:
+            batch_size (int, optional): Number of documents per batch.
+                Defaults to settings.SEARCH_INDEXER_BATCH_SIZE.
+        """
+        self.batch_size = batch_size or settings.SEARCH_INDEXER_BATCH_SIZE
+        self.indexer_url = settings.SEARCH_INDEXER_URL
+        self.indexer_secret = settings.SEARCH_INDEXER_SECRET
+        self.search_url = settings.SEARCH_INDEXER_QUERY_URL
+
+        if not self.indexer_url:
+            raise ImproperlyConfigured(
+                "SEARCH_INDEXER_URL must be set in Django settings."
+            )
+
+        if not self.indexer_secret:
+            raise ImproperlyConfigured(
+                "SEARCH_INDEXER_SECRET must be set in Django settings."
+            )
+
+        if not self.search_url:
+            raise ImproperlyConfigured(
+                "SEARCH_INDEXER_QUERY_URL must be set in Django settings."
+            )
+
+    def index(self):
+        """
+        Fetch documents in batches, serialize them, and push to the search backend.
+        """
+        last_id = 0
+        count = 0
+
+        while True:
+            documents_batch = list(
+                models.Document.objects.filter(
+                    id__gt=last_id,
+                ).order_by("id")[: self.batch_size]
+            )
+
+            if not documents_batch:
+                break
+
+            doc_paths = [doc.path for doc in documents_batch]
+            last_id = documents_batch[-1].id
+            accesses_by_document_path = get_batch_accesses_by_users_and_teams(doc_paths)
+
+            serialized_batch = [
+                self.serialize_document(document, accesses_by_document_path)
+                for document in documents_batch
+                if document.content or document.title
+            ]
+
+            self.push(serialized_batch)
+            count += len(serialized_batch)
+
+        return count
+
+    @abstractmethod
+    def serialize_document(self, document, accesses):
+        """
+        Convert a Document instance to a JSON-serializable format for indexing.
+
+        Must be implemented by subclasses.
+        """
+
+    @abstractmethod
+    def push(self, data):
+        """
+        Push a batch of serialized documents to the backend.
+
+        Must be implemented by subclasses.
+        """
+
+    # pylint: disable-next=too-many-arguments,too-many-positional-arguments
+    def search(self, text, token, visited=(), page=1, page_size=50):
+        """
+        Search for documents in Find app.
+        Ensure the same default ordering as "Docs" list : -updated_at
+
+        Returns ids of the documents
+
+        Args:
+            text (str): Text search content.
+            token (str): OIDC Authentication token.
+            visited (list, optional):
+                List of ids of active public documents with LinkTrace
+                Defaults to settings.SEARCH_INDEXER_BATCH_SIZE.
+            page (int, optional):
+                The page number to retrieve.
+                Defaults to 1 if not specified.
+            page_size (int, optional):
+                The number of results to return per page.
+                Defaults to 50 if not specified.
+        """
+        response = self.search_query(
+            data={
+                "q": text,
+                "visited": visited,
+                "services": ["docs"],
+                "page_number": page,
+                "page_size": page_size,
+                "order_by": "updated_at",
+                "order_direction": "desc",
+            },
+            token=token,
+        )
+
+        return [d["_id"] for d in response]
+
+    @abstractmethod
+    def search_query(self, data, token) -> dict:
+        """
+        Retrieve documents from the Find app API.
+
+        Must be implemented by subclasses.
+        """
+
+
+class SearchIndexer(BaseDocumentIndexer):
+    """
+    Document indexer that pushes documents to La Suite Find app.
+    """
+
+    def serialize_document(self, document, accesses):
+        """
+        Convert a Document to the JSON format expected by La Suite Find.
+
+        Args:
+            document (Document): The document instance.
+            accesses (dict): Mapping of document ID to user/team access.
+
+        Returns:
+            dict: A JSON-serializable dictionary.
+        """
+        doc_path = document.path
+        doc_content = document.content
+        text_content = utils.base64_yjs_to_text(doc_content) if doc_content else ""
+
+        return {
+            "id": str(document.id),
+            "title": document.title or "",
+            "content": text_content,
+            "depth": document.depth,
+            "path": document.path,
+            "numchild": document.numchild,
+            "created_at": document.created_at.isoformat(),
+            "updated_at": document.updated_at.isoformat(),
+            "users": list(accesses.get(doc_path, {}).get("users", set())),
+            "groups": list(accesses.get(doc_path, {}).get("teams", set())),
+            "reach": document.computed_link_reach,
+            "size": len(text_content.encode("utf-8")),
+            "is_active": not bool(document.ancestors_deleted_at),
+        }
+
+    def search_query(self, data, token) -> requests.Response:
+        """
+        Retrieve documents from the Find app API.
+
+        Args:
+            data (dict): search data
+            token (str): OICD token
+
+        Returns:
+            dict: A JSON-serializable dictionary.
+        """
+        response = requests.post(
+            self.search_url,
+            json=data,
+            headers={"Authorization": f"Bearer {token}"},
+            timeout=10,
+        )
+        response.raise_for_status()
+        return response.json()
+
+    def push(self, data):
+        """
+        Push a batch of documents to the Find backend.
+
+        Args:
+            data (list): List of document dictionaries.
+        """
+        response = requests.post(
+            self.indexer_url,
+            json=data,
+            headers={"Authorization": f"Bearer {self.indexer_secret}"},
+            timeout=10,
+        )
+        response.raise_for_status()

src/backend/core/signals.py

@@ -0,0 +1,31 @@
+"""
+Declare and configure the signals for the impress core application
+"""
+
+from functools import partial
+
+from django.db import transaction
+from django.db.models import signals
+from django.dispatch import receiver
+
+from . import models
+from .tasks.search import trigger_document_indexer
+
+
+@receiver(signals.post_save, sender=models.Document)
+def document_post_save(sender, instance, **kwargs):  # pylint: disable=unused-argument
+    """
+    Asynchronous call to the document indexer at the end of the transaction.
+    Note : Within the transaction we can have an empty content and a serialization
+    error.
+    """
+    transaction.on_commit(partial(trigger_document_indexer, instance))
+
+
+@receiver(signals.post_save, sender=models.DocumentAccess)
+def document_access_post_save(sender, instance, created, **kwargs):  # pylint: disable=unused-argument
+    """
+    Asynchronous call to the document indexer at the end of the transaction.
+    """
+    if not created:
+        transaction.on_commit(partial(trigger_document_indexer, instance.document))

src/backend/core/tasks/search.py

@@ -0,0 +1,84 @@
+"""Trigger document indexation using celery task."""
+
+from logging import getLogger
+
+from django.conf import settings
+from django.core.cache import cache
+
+from django_redis.cache import RedisCache
+
+from core import models
+from core.services.search_indexers import (
+    get_batch_accesses_by_users_and_teams,
+    get_document_indexer,
+)
+
+from impress.celery_app import app
+
+logger = getLogger(__file__)
+
+
+def indexer_throttle_acquire(document_id, timeout=0, atomic=True):
+    """
+    Enable the task throttle flag for a delay.
+    Uses redis locks if available to ensure atomic changes
+    """
+    key = f"doc-indexer-throttle-{document_id}"
+
+    if isinstance(cache, RedisCache) and atomic:
+        with cache.locks(key):
+            return indexer_throttle_acquire(document_id, timeout, atomic=False)
+
+    # Use add() here :
+    #   - set the flag and returns true if not exist
+    #   - do nothing and return false if exist
+    return cache.add(key, 1, timeout=timeout)
+
+
+@app.task
+def document_indexer_task(document_id):
+    """Celery Task : Sends indexation query for a document."""
+    indexer = get_document_indexer()
+
+    if indexer is None:
+        return
+
+    try:
+        doc = models.Document.objects.get(pk=document_id)
+    except models.Document.DoesNotExist:
+        # Skip the task if the document does not exist.
+        return
+
+    accesses = get_batch_accesses_by_users_and_teams((doc.path,))
+
+    data = indexer.serialize_document(document=doc, accesses=accesses)
+
+    logger.info("Start document %s indexation", document_id)
+    indexer.push(data)
+
+
+def trigger_document_indexer(document):
+    """
+    Trigger indexation task with debounce a delay set by the SEARCH_INDEXER_COUNTDOWN setting.
+
+    Args:
+        document (Document): The document instance.
+    """
+    countdown = settings.SEARCH_INDEXER_COUNTDOWN
+
+    # DO NOT create a task if indexation if disabled
+    if not settings.SEARCH_INDEXER_CLASS:
+        return
+
+    # Each time this method is called during a countdown, we increment the
+    # counter and each task decrease it, so the index be run only once.
+    if indexer_throttle_acquire(document.pk, timeout=countdown):
+        logger.info(
+            "Add task for document %s indexation in %.2f seconds",
+            document.pk,
+            countdown,
+        )
+
+        document_indexer_task.apply_async(args=[document.pk])
+    else:
+        logger.info("Skip task for document %s indexation", document.pk)

src/backend/core/tests/commands/test_index.py

@@ -0,0 +1,65 @@
+"""
+Unit test for `index` command.
+"""
+
+from operator import itemgetter
+from unittest import mock
+
+from django.core.management import CommandError, call_command
+from django.db import transaction
+
+import pytest
+
+from core import factories
+from core.services.search_indexers import SearchIndexer
+
+
+@pytest.mark.django_db
+@pytest.mark.usefixtures("indexer_settings")
+def test_index():
+    """Test the command `index` that run the Find app indexer for all the available documents."""
+    user = factories.UserFactory()
+    indexer = SearchIndexer()
+
+    with transaction.atomic():
+        doc = factories.DocumentFactory()
+        empty_doc = factories.DocumentFactory(title=None, content="")
+        no_title_doc = factories.DocumentFactory(title=None)
+
+        factories.UserDocumentAccessFactory(document=doc, user=user)
+        factories.UserDocumentAccessFactory(document=empty_doc, user=user)
+        factories.UserDocumentAccessFactory(document=no_title_doc, user=user)
+
+    accesses = {
+        str(doc.path): {"users": [user.sub]},
+        str(empty_doc.path): {"users": [user.sub]},
+        str(no_title_doc.path): {"users": [user.sub]},
+    }
+
+    with mock.patch.object(SearchIndexer, "push") as mock_push:
+        call_command("index")
+
+        push_call_args = [call.args[0] for call in mock_push.call_args_list]
+
+        # called once but with a batch of docs
+        mock_push.assert_called_once()
+
+        assert sorted(push_call_args[0], key=itemgetter("id")) == sorted(
+            [
+                indexer.serialize_document(doc, accesses),
+                indexer.serialize_document(no_title_doc, accesses),
+            ],
+            key=itemgetter("id"),
+        )
+
+
+@pytest.mark.django_db
+@pytest.mark.usefixtures("indexer_settings")
+def test_index_improperly_configured(indexer_settings):
+    """The command should raise an exception if the indexer is not configured"""
+    indexer_settings.SEARCH_INDEXER_CLASS = None
+
+    with pytest.raises(CommandError) as err:
+        call_command("index")
+
+    assert str(err.value) == "The indexer is not enabled or properly configured."

src/backend/core/tests/conftest.py

@@ -24,3 +24,29 @@ def mock_user_teams():
         "core.models.User.teams", new_callable=mock.PropertyMock
     ) as mock_teams:
         yield mock_teams
+
+
+@pytest.fixture(name="indexer_settings")
+def indexer_settings_fixture(settings):
+    """
+    Setup valid settings for the document indexer. Clear the indexer cache.
+    """
+
+    # pylint: disable-next=import-outside-toplevel
+    from core.services.search_indexers import (  # noqa: PLC0415
+        get_document_indexer,
+    )
+
+    get_document_indexer.cache_clear()
+
+    settings.SEARCH_INDEXER_CLASS = "core.services.search_indexers.SearchIndexer"
+    settings.SEARCH_INDEXER_SECRET = "ThisIsAKeyForTest"
+    settings.SEARCH_INDEXER_URL = "http://localhost:8081/api/v1.0/documents/index/"
+    settings.SEARCH_INDEXER_QUERY_URL = (
+        "http://localhost:8081/api/v1.0/documents/search/"
+    )
+
+    yield settings
+
+    # clear cache to prevent issues with other tests
+    get_document_indexer.cache_clear()

src/backend/core/tests/documents/test_api_documents_search.py

@@ -0,0 +1,426 @@
+"""
+Tests for Documents API endpoint in impress's core app: list
+"""
+
+import random
+from json import loads as json_loads
+
+from django.test import RequestFactory
+
+import pytest
+import responses
+from faker import Faker
+from rest_framework.test import APIClient
+
+from core import factories, models
+from core.services.search_indexers import get_document_indexer
+
+fake = Faker()
+pytestmark = pytest.mark.django_db
+
+
+def build_search_url(**kwargs):
+    """Build absolute uri for search endpoint with ORDERED query arguments"""
+    return (
+        RequestFactory()
+        .get("/api/v1.0/documents/search/", dict(sorted(kwargs.items())))
+        .build_absolute_uri()
+    )
+
+
+@pytest.mark.parametrize("role", models.LinkRoleChoices.values)
+@pytest.mark.parametrize("reach", models.LinkReachChoices.values)
+@responses.activate
+def test_api_documents_search_anonymous(reach, role, indexer_settings):
+    """
+    Anonymous users should not be allowed to search documents whatever the
+    link reach and link role
+    """
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = "http://find/api/v1.0/search"
+
+    factories.DocumentFactory(link_reach=reach, link_role=role)
+
+    # Find response
+    responses.add(
+        responses.POST,
+        "http://find/api/v1.0/search",
+        json=[],
+        status=200,
+    )
+
+    response = APIClient().get("/api/v1.0/documents/search/", data={"q": "alpha"})
+
+    assert response.status_code == 200
+    assert response.json() == {
+        "count": 0,
+        "next": None,
+        "previous": None,
+        "results": [],
+    }
+
+
+def test_api_documents_search_endpoint_is_none(indexer_settings):
+    """
+    Missing SEARCH_INDEXER_QUERY_URL, so the indexer is not properly configured.
+    Should fallback on title filter
+    """
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = None
+
+    assert get_document_indexer() is None
+
+    user = factories.UserFactory()
+    document = factories.DocumentFactory(title="alpha")
+    access = factories.UserDocumentAccessFactory(document=document, user=user)
+
+    client = APIClient()
+    client.force_login(user)
+
+    response = client.get("/api/v1.0/documents/search/", data={"q": "alpha"})
+
+    assert response.status_code == 200
+    content = response.json()
+    results = content.pop("results")
+    assert content == {
+        "count": 1,
+        "next": None,
+        "previous": None,
+    }
+    assert len(results) == 1
+    assert results[0] == {
+        "id": str(document.id),
+        "abilities": document.get_abilities(user),
+        "ancestors_link_reach": None,
+        "ancestors_link_role": None,
+        "computed_link_reach": document.computed_link_reach,
+        "computed_link_role": document.computed_link_role,
+        "created_at": document.created_at.isoformat().replace("+00:00", "Z"),
+        "creator": str(document.creator.id),
+        "depth": 1,
+        "excerpt": document.excerpt,
+        "link_reach": document.link_reach,
+        "link_role": document.link_role,
+        "nb_accesses_ancestors": 1,
+        "nb_accesses_direct": 1,
+        "numchild": 0,
+        "path": document.path,
+        "title": document.title,
+        "updated_at": document.updated_at.isoformat().replace("+00:00", "Z"),
+        "deleted_at": None,
+        "user_role": access.role,
+    }
+
+
+@responses.activate
+def test_api_documents_search_invalid_params(indexer_settings):
+    """Validate the format of documents as returned by the search view."""
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = "http://find/api/v1.0/search"
+
+    user = factories.UserFactory()
+
+    client = APIClient()
+    client.force_login(user)
+
+    response = client.get("/api/v1.0/documents/search/")
+
+    assert response.status_code == 400
+    assert response.json() == {"q": ["This field is required."]}
+
+    response = client.get("/api/v1.0/documents/search/", data={"q": "    "})
+
+    assert response.status_code == 400
+    assert response.json() == {"q": ["This field may not be blank."]}
+
+    response = client.get(
+        "/api/v1.0/documents/search/", data={"q": "any", "page": "NaN"}
+    )
+
+    assert response.status_code == 400
+    assert response.json() == {"page": ["A valid integer is required."]}
+
+
+@responses.activate
+def test_api_documents_search_format(indexer_settings):
+    """Validate the format of documents as returned by the search view."""
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = "http://find/api/v1.0/search"
+
+    assert get_document_indexer() is not None
+
+    user = factories.UserFactory()
+
+    client = APIClient()
+    client.force_login(user)
+
+    user_a, user_b, user_c = factories.UserFactory.create_batch(3)
+    document = factories.DocumentFactory(
+        title="alpha",
+        users=(user_a, user_c),
+        link_traces=(user, user_b),
+    )
+    access = factories.UserDocumentAccessFactory(document=document, user=user)
+
+    # Find response
+    responses.add(
+        responses.POST,
+        "http://find/api/v1.0/search",
+        json=[
+            {"_id": str(document.pk)},
+        ],
+        status=200,
+    )
+    response = client.get("/api/v1.0/documents/search/", data={"q": "alpha"})
+
+    assert response.status_code == 200
+    content = response.json()
+    results = content.pop("results")
+    assert content == {
+        "count": 1,
+        "next": None,
+        "previous": None,
+    }
+    assert len(results) == 1
+    assert results[0] == {
+        "id": str(document.id),
+        "abilities": document.get_abilities(user),
+        "ancestors_link_reach": None,
+        "ancestors_link_role": None,
+        "computed_link_reach": document.computed_link_reach,
+        "computed_link_role": document.computed_link_role,
+        "created_at": document.created_at.isoformat().replace("+00:00", "Z"),
+        "creator": str(document.creator.id),
+        "depth": 1,
+        "excerpt": document.excerpt,
+        "link_reach": document.link_reach,
+        "link_role": document.link_role,
+        "nb_accesses_ancestors": 3,
+        "nb_accesses_direct": 3,
+        "numchild": 0,
+        "path": document.path,
+        "title": document.title,
+        "updated_at": document.updated_at.isoformat().replace("+00:00", "Z"),
+        "deleted_at": None,
+        "user_role": access.role,
+    }
+
+
+@responses.activate
+@pytest.mark.parametrize(
+    "pagination, status, expected",
+    (
+        (
+            {"page": 1, "page_size": 10},
+            200,
+            {
+                "count": 10,
+                "previous": None,
+                "next": None,
+                "range": (0, None),
+            },
+        ),
+        (
+            {},
+            200,
+            {
+                "count": 10,
+                "previous": None,
+                "next": None,
+                "range": (0, None),
+                "api_page_size": 21,  # default page_size is 20
+            },
+        ),
+        (
+            {"page": 2, "page_size": 10},
+            404,
+            {},
+        ),
+        (
+            {"page": 1, "page_size": 5},
+            200,
+            {
+                "count": 10,
+                "previous": None,
+                "next": {"page": 2, "page_size": 5},
+                "range": (0, 5),
+            },
+        ),
+        (
+            {"page": 2, "page_size": 5},
+            200,
+            {
+                "count": 10,
+                "previous": {"page": 1, "page_size": 5},
+                "next": None,
+                "range": (5, None),
+            },
+        ),
+        ({"page": 3, "page_size": 5}, 404, {}),
+    ),
+)
+def test_api_documents_search_pagination(
+    indexer_settings, pagination, status, expected
+):
+    """Documents should be ordered by descending "score" by default"""
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = "http://find/api/v1.0/search"
+
+    assert get_document_indexer() is not None
+
+    user = factories.UserFactory()
+
+    client = APIClient()
+    client.force_login(user)
+
+    docs = factories.DocumentFactory.create_batch(10, title="alpha", users=[user])
+
+    docs_by_uuid = {str(doc.pk): doc for doc in docs}
+    api_results = [{"_id": id} for id in docs_by_uuid.keys()]
+
+    # reorder randomly to simulate score ordering
+    random.shuffle(api_results)
+
+    # Find response
+    # pylint: disable-next=assignment-from-none
+    api_search = responses.add(
+        responses.POST,
+        "http://find/api/v1.0/search",
+        json=api_results,
+        status=200,
+    )
+
+    response = client.get(
+        "/api/v1.0/documents/search/",
+        data={
+            "q": "alpha",
+            **pagination,
+        },
+    )
+
+    assert response.status_code == status
+
+    if response.status_code < 300:
+        previous_url = (
+            build_search_url(q="alpha", **expected["previous"])
+            if expected["previous"]
+            else None
+        )
+        next_url = (
+            build_search_url(q="alpha", **expected["next"])
+            if expected["next"]
+            else None
+        )
+        start, end = expected["range"]
+
+        content = response.json()
+
+        assert content["count"] == expected["count"]
+        assert content["previous"] == previous_url
+        assert content["next"] == next_url
+
+        results = content.pop("results")
+
+        # The find api results ordering by score is kept
+        assert [r["id"] for r in results] == [r["_id"] for r in api_results[start:end]]
+
+        # Check the query parameters.
+        assert api_search.call_count == 1
+        assert api_search.calls[0].response.status_code == 200
+        assert json_loads(api_search.calls[0].request.body) == {
+            "q": "alpha",
+            "visited": [],
+            "services": ["docs"],
+            "page_number": 1,
+            "page_size": 100,
+            "order_by": "updated_at",
+            "order_direction": "desc",
+        }
+
+
+@responses.activate
+@pytest.mark.parametrize(
+    "pagination, status, expected",
+    (
+        (
+            {"page": 1, "page_size": 10},
+            200,
+            {"count": 10, "previous": None, "next": None, "range": (0, None)},
+        ),
+        (
+            {},
+            200,
+            {"count": 10, "previous": None, "next": None, "range": (0, None)},
+        ),
+        (
+            {"page": 2, "page_size": 10},
+            404,
+            {},
+        ),
+        (
+            {"page": 1, "page_size": 5},
+            200,
+            {
+                "count": 10,
+                "previous": None,
+                "next": {"page": 2, "page_size": 5},
+                "range": (0, 5),
+            },
+        ),
+        (
+            {"page": 2, "page_size": 5},
+            200,
+            {
+                "count": 10,
+                "previous": {"page_size": 5},
+                "next": None,
+                "range": (5, None),
+            },
+        ),
+        ({"page": 3, "page_size": 5}, 404, {}),
+    ),
+)
+def test_api_documents_search_pagination_endpoint_is_none(
+    indexer_settings, pagination, status, expected
+):
+    """Documents should be ordered by descending "-updated_at" by default"""
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = None
+
+    assert get_document_indexer() is None
+
+    user = factories.UserFactory()
+
+    client = APIClient()
+    client.force_login(user)
+
+    factories.DocumentFactory.create_batch(10, title="alpha", users=[user])
+
+    response = client.get(
+        "/api/v1.0/documents/search/",
+        data={
+            "q": "alpha",
+            **pagination,
+        },
+    )
+
+    assert response.status_code == status
+
+    if response.status_code < 300:
+        previous_url = (
+            build_search_url(q="alpha", **expected["previous"])
+            if expected["previous"]
+            else None
+        )
+        next_url = (
+            build_search_url(q="alpha", **expected["next"])
+            if expected["next"]
+            else None
+        )
+        queryset = models.Document.objects.order_by("-updated_at")
+        start, end = expected["range"]
+        expected_results = [str(d.pk) for d in queryset[start:end]]
+
+        content = response.json()
+
+        assert content["count"] == expected["count"]
+        assert content["previous"] == previous_url
+        assert content["next"] == next_url
+
+        results = content.pop("results")
+
+        assert [r["id"] for r in results] == expected_results

src/backend/core/tests/test_models_documents.py

@@ -6,6 +6,7 @@ Unit tests for the Document model
 import random
 import smtplib
 from logging import Logger
+from operator import itemgetter
 from unittest import mock
 
 from django.contrib.auth.models import AnonymousUser
@@ -13,12 +14,15 @@ from django.core import mail
 from django.core.cache import cache
 from django.core.exceptions import ValidationError
 from django.core.files.storage import default_storage
+from django.db import transaction
 from django.test.utils import override_settings
 from django.utils import timezone
 
 import pytest
 
 from core import factories, models
+from core.services.search_indexers import SearchIndexer
+from core.tasks.search import document_indexer_task
 
 pytestmark = pytest.mark.django_db
 
@@ -1445,3 +1449,330 @@ def test_models_documents_compute_ancestors_links_paths_mapping_structure(
                 {"link_reach": sibling.link_reach, "link_role": sibling.link_role},
             ],
         }
+
+
+@mock.patch.object(SearchIndexer, "push")
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer(mock_push, indexer_settings):
+    """Test indexation task on document creation"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    with transaction.atomic():
+        doc1, doc2, doc3 = factories.DocumentFactory.create_batch(3)
+
+    accesses = {}
+    data = [call.args[0] for call in mock_push.call_args_list]
+
+    indexer = SearchIndexer()
+
+    assert sorted(data, key=itemgetter("id")) == sorted(
+        [
+            indexer.serialize_document(doc1, accesses),
+            indexer.serialize_document(doc2, accesses),
+            indexer.serialize_document(doc3, accesses),
+        ],
+        key=itemgetter("id"),
+    )
+
+    # The throttle counters should be reset
+    assert cache.get(f"doc-indexer-throttle-{doc1.pk}") is None
+    assert cache.get(f"doc-indexer-throttle-{doc2.pk}") is None
+    assert cache.get(f"doc-indexer-throttle-{doc3.pk}") is None
+
+
+@mock.patch.object(SearchIndexer, "push")
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer_not_configured(mock_push, indexer_settings):
+    """Task should not start an indexation when disabled"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+    indexer_settings.SEARCH_INDEXER_CLASS = None
+
+    user = factories.UserFactory()
+
+    with transaction.atomic():
+        doc = factories.DocumentFactory()
+        factories.UserDocumentAccessFactory(document=doc, user=user)
+
+    assert mock_push.assert_not_called
+
+
+@mock.patch.object(SearchIndexer, "push")
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer_wrongly_configured(
+    mock_push, indexer_settings
+):
+    """Task should not start an indexation when disabled"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+    indexer_settings.SEARCH_INDEXER_URL = None
+
+    user = factories.UserFactory()
+
+    with transaction.atomic():
+        doc = factories.DocumentFactory()
+        factories.UserDocumentAccessFactory(document=doc, user=user)
+
+    assert mock_push.assert_not_called
+
+
+@mock.patch.object(SearchIndexer, "push")
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer_with_accesses(mock_push, indexer_settings):
+    """Test indexation task on document creation"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    user = factories.UserFactory()
+
+    with transaction.atomic():
+        doc1, doc2, doc3 = factories.DocumentFactory.create_batch(3)
+
+        factories.UserDocumentAccessFactory(document=doc1, user=user)
+        factories.UserDocumentAccessFactory(document=doc2, user=user)
+        factories.UserDocumentAccessFactory(document=doc3, user=user)
+
+    accesses = {
+        str(doc1.path): {"users": [user.sub]},
+        str(doc2.path): {"users": [user.sub]},
+        str(doc3.path): {"users": [user.sub]},
+    }
+
+    data = [call.args[0] for call in mock_push.call_args_list]
+
+    indexer = SearchIndexer()
+
+    assert sorted(data, key=itemgetter("id")) == sorted(
+        [
+            indexer.serialize_document(doc1, accesses),
+            indexer.serialize_document(doc2, accesses),
+            indexer.serialize_document(doc3, accesses),
+        ],
+        key=itemgetter("id"),
+    )
+
+    # The throttle counters should be reset
+    assert cache.get(f"doc-indexer-throttle-{doc1.pk}") is None
+    assert cache.get(f"doc-indexer-throttle-{doc2.pk}") is None
+    assert cache.get(f"doc-indexer-throttle-{doc3.pk}") is None
+
+
+@mock.patch.object(SearchIndexer, "push")
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer_deleted(mock_push, indexer_settings):
+    """Indexation task on deleted or ancestor_deleted documents"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    user = factories.UserFactory()
+
+    with transaction.atomic():
+        doc = factories.DocumentFactory(
+            link_reach=models.LinkReachChoices.AUTHENTICATED
+        )
+        doc_deleted = factories.DocumentFactory(
+            link_reach=models.LinkReachChoices.AUTHENTICATED
+        )
+        doc_ancestor_deleted = factories.DocumentFactory(
+            parent=doc_deleted,
+            link_reach=models.LinkReachChoices.AUTHENTICATED,
+        )
+        doc_deleted.soft_delete()
+        doc_ancestor_deleted.ancestors_deleted_at = doc_deleted.deleted_at
+
+        factories.UserDocumentAccessFactory(document=doc, user=user)
+        factories.UserDocumentAccessFactory(document=doc_deleted, user=user)
+        factories.UserDocumentAccessFactory(document=doc_ancestor_deleted, user=user)
+
+    doc_deleted.refresh_from_db()
+    doc_ancestor_deleted.refresh_from_db()
+
+    assert doc_deleted.deleted_at is not None
+    assert doc_deleted.ancestors_deleted_at is not None
+
+    assert doc_ancestor_deleted.deleted_at is None
+    assert doc_ancestor_deleted.ancestors_deleted_at is not None
+
+    accesses = {
+        str(doc.path): {"users": [user.sub]},
+        str(doc_deleted.path): {"users": [user.sub]},
+        str(doc_ancestor_deleted.path): {"users": [user.sub]},
+    }
+
+    data = [call.args[0] for call in mock_push.call_args_list]
+
+    indexer = SearchIndexer()
+
+    # Even deleted document are re-indexed : only update their status in the future ?
+    assert sorted(data, key=itemgetter("id")) == sorted(
+        [
+            indexer.serialize_document(doc, accesses),
+            indexer.serialize_document(doc_deleted, accesses),
+            indexer.serialize_document(doc_ancestor_deleted, accesses),
+            indexer.serialize_document(doc_deleted, accesses),  # soft_delete()
+        ],
+        key=itemgetter("id"),
+    )
+
+    # The throttle counters should be reset
+    assert cache.get(f"doc-indexer-throttle-{doc.pk}") is None
+    assert cache.get(f"doc-indexer-throttle-{doc_deleted.pk}") is None
+    assert cache.get(f"doc-indexer-throttle-{doc_ancestor_deleted.pk}") is None
+
+
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_indexer_hard_deleted(indexer_settings):
+    """Indexation task on hard deleted document"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    user = factories.UserFactory()
+
+    with transaction.atomic():
+        doc = factories.DocumentFactory(
+            link_reach=models.LinkReachChoices.AUTHENTICATED
+        )
+        factories.UserDocumentAccessFactory(document=doc, user=user)
+
+    doc_id = doc.pk
+    doc.delete()
+
+    # Call task on deleted document.
+    document_indexer_task.apply(args=[doc_id])
+
+    with mock.patch.object(SearchIndexer, "push") as mock_push:
+        # Hard delete document are not re-indexed.
+        assert mock_push.assert_not_called
+
+
+@mock.patch.object(SearchIndexer, "push")
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer_restored(mock_push, indexer_settings):
+    """Restart indexation task on restored documents"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    user = factories.UserFactory()
+
+    with transaction.atomic():
+        doc = factories.DocumentFactory(
+            link_reach=models.LinkReachChoices.AUTHENTICATED
+        )
+        doc_deleted = factories.DocumentFactory(
+            link_reach=models.LinkReachChoices.AUTHENTICATED
+        )
+        doc_ancestor_deleted = factories.DocumentFactory(
+            parent=doc_deleted,
+            link_reach=models.LinkReachChoices.AUTHENTICATED,
+        )
+        doc_deleted.soft_delete()
+        doc_ancestor_deleted.ancestors_deleted_at = doc_deleted.deleted_at
+
+        factories.UserDocumentAccessFactory(document=doc, user=user)
+        factories.UserDocumentAccessFactory(document=doc_deleted, user=user)
+        factories.UserDocumentAccessFactory(document=doc_ancestor_deleted, user=user)
+
+    doc_deleted.refresh_from_db()
+    doc_ancestor_deleted.refresh_from_db()
+
+    assert doc_deleted.deleted_at is not None
+    assert doc_deleted.ancestors_deleted_at is not None
+
+    assert doc_ancestor_deleted.deleted_at is None
+    assert doc_ancestor_deleted.ancestors_deleted_at is not None
+
+    doc_restored = models.Document.objects.get(pk=doc_deleted.pk)
+    doc_restored.restore()
+
+    doc_ancestor_restored = models.Document.objects.get(pk=doc_ancestor_deleted.pk)
+
+    assert doc_restored.deleted_at is None
+    assert doc_restored.ancestors_deleted_at is None
+
+    assert doc_ancestor_restored.deleted_at is None
+    assert doc_ancestor_restored.ancestors_deleted_at is None
+
+    accesses = {
+        str(doc.path): {"users": [user.sub]},
+        str(doc_deleted.path): {"users": [user.sub]},
+        str(doc_ancestor_deleted.path): {"users": [user.sub]},
+    }
+
+    data = [call.args[0] for call in mock_push.call_args_list]
+
+    indexer = SearchIndexer()
+
+    # All docs are re-indexed
+    assert sorted(data, key=itemgetter("id")) == sorted(
+        [
+            indexer.serialize_document(doc, accesses),
+            indexer.serialize_document(doc_deleted, accesses),
+            indexer.serialize_document(doc_deleted, accesses),  # soft_delete()
+            indexer.serialize_document(doc_restored, accesses),  # restore()
+            indexer.serialize_document(doc_ancestor_deleted, accesses),
+        ],
+        key=itemgetter("id"),
+    )
+
+
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_post_save_indexer_throttle(indexer_settings):
+    """Test indexation task skipping on document update"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    indexer = SearchIndexer()
+    user = factories.UserFactory()
+
+    with mock.patch.object(SearchIndexer, "push"):
+        with transaction.atomic():
+            doc = factories.DocumentFactory()
+            factories.UserDocumentAccessFactory(document=doc, user=user)
+
+    accesses = {
+        str(doc.path): {"users": [user.sub]},
+    }
+
+    with mock.patch.object(SearchIndexer, "push") as mock_push:
+        # Simulate 1 running task
+        cache.set(f"doc-indexer-throttle-{doc.pk}", 1)
+
+        # save doc to trigger the indexer, but nothing should be done since
+        # the flag is up
+        with transaction.atomic():
+            doc.save()
+
+        assert [call.args[0] for call in mock_push.call_args_list] == []
+
+    with mock.patch.object(SearchIndexer, "push") as mock_push:
+        # No waiting task
+        cache.delete(f"doc-indexer-throttle-{doc.pk}")
+
+        with transaction.atomic():
+            doc = models.Document.objects.get(pk=doc.pk)
+            doc.save()
+
+        assert [call.args[0] for call in mock_push.call_args_list] == [
+            indexer.serialize_document(doc, accesses),
+        ]
+
+
+@pytest.mark.django_db(transaction=True)
+def test_models_documents_access_post_save_indexer(indexer_settings):
+    """Test indexation task on DocumentAccess update"""
+    indexer_settings.SEARCH_INDEXER_COUNTDOWN = 0
+
+    indexer = SearchIndexer()
+    user = factories.UserFactory()
+
+    with mock.patch.object(SearchIndexer, "push"):
+        with transaction.atomic():
+            doc = factories.DocumentFactory()
+            doc_access = factories.UserDocumentAccessFactory(document=doc, user=user)
+
+    accesses = {
+        str(doc.path): {"users": [user.sub]},
+    }
+
+    indexer = SearchIndexer()
+
+    with mock.patch.object(SearchIndexer, "push") as mock_push:
+        with transaction.atomic():
+            doc_access.save()
+
+        assert [call.args[0] for call in mock_push.call_args_list] == [
+            indexer.serialize_document(doc, accesses),
+        ]

src/backend/core/tests/test_services_search_indexers.py

@@ -0,0 +1,540 @@
+"""Tests for Documents search indexers"""
+
+from functools import partial
+from json import dumps as json_dumps
+from unittest.mock import patch
+
+from django.contrib.auth.models import AnonymousUser
+from django.core.exceptions import ImproperlyConfigured
+from django.utils.module_loading import import_string
+
+import pytest
+import responses
+from requests import HTTPError
+
+from core import factories, models, utils
+from core.services.search_indexers import (
+    BaseDocumentIndexer,
+    SearchIndexer,
+    get_document_indexer,
+    get_visited_document_ids_of,
+)
+
+pytestmark = pytest.mark.django_db
+
+
+class FakeDocumentIndexer(BaseDocumentIndexer):
+    """Fake indexer for test purpose"""
+
+    def serialize_document(self, document, accesses):
+        return {}
+
+    def push(self, data):
+        pass
+
+    def search_query(self, data, token):
+        return {}
+
+
+def test_services_search_indexer_class_invalid(indexer_settings):
+    """
+    Should raise RuntimeError if SEARCH_INDEXER_CLASS cannot be imported.
+    """
+    indexer_settings.SEARCH_INDEXER_CLASS = "unknown.Unknown"
+
+    assert get_document_indexer() is None
+
+
+def test_services_search_indexer_class(indexer_settings):
+    """
+    Import indexer class defined in setting SEARCH_INDEXER_CLASS.
+    """
+    indexer_settings.SEARCH_INDEXER_CLASS = (
+        "core.tests.test_services_search_indexers.FakeDocumentIndexer"
+    )
+
+    assert isinstance(
+        get_document_indexer(),
+        import_string("core.tests.test_services_search_indexers.FakeDocumentIndexer"),
+    )
+
+
+def test_services_search_indexer_is_configured(indexer_settings):
+    """
+    Should return true only when the indexer class and other configuration settings
+    are valid.
+    """
+    indexer_settings.SEARCH_INDEXER_CLASS = None
+
+    # None
+    get_document_indexer.cache_clear()
+    assert not get_document_indexer()
+
+    # Empty
+    indexer_settings.SEARCH_INDEXER_CLASS = ""
+
+    get_document_indexer.cache_clear()
+    assert not get_document_indexer()
+
+    # Valid class
+    indexer_settings.SEARCH_INDEXER_CLASS = (
+        "core.services.search_indexers.SearchIndexer"
+    )
+
+    get_document_indexer.cache_clear()
+    assert get_document_indexer() is not None
+
+    indexer_settings.SEARCH_INDEXER_URL = ""
+
+    # Invalid url
+    get_document_indexer.cache_clear()
+    assert not get_document_indexer()
+
+
+def test_services_search_indexer_url_is_none(indexer_settings):
+    """
+    Indexer should raise RuntimeError if SEARCH_INDEXER_URL is None or empty.
+    """
+    indexer_settings.SEARCH_INDEXER_URL = None
+
+    with pytest.raises(ImproperlyConfigured) as exc_info:
+        SearchIndexer()
+
+    assert "SEARCH_INDEXER_URL must be set in Django settings." in str(exc_info.value)
+
+
+def test_services_search_indexer_url_is_empty(indexer_settings):
+    """
+    Indexer should raise RuntimeError if SEARCH_INDEXER_URL is empty string.
+    """
+    indexer_settings.SEARCH_INDEXER_URL = ""
+
+    with pytest.raises(ImproperlyConfigured) as exc_info:
+        SearchIndexer()
+
+    assert "SEARCH_INDEXER_URL must be set in Django settings." in str(exc_info.value)
+
+
+def test_services_search_indexer_secret_is_none(indexer_settings):
+    """
+    Indexer should raise RuntimeError if SEARCH_INDEXER_SECRET is None.
+    """
+    indexer_settings.SEARCH_INDEXER_SECRET = None
+
+    with pytest.raises(ImproperlyConfigured) as exc_info:
+        SearchIndexer()
+
+    assert "SEARCH_INDEXER_SECRET must be set in Django settings." in str(
+        exc_info.value
+    )
+
+
+def test_services_search_indexer_secret_is_empty(indexer_settings):
+    """
+    Indexer should raise RuntimeError if SEARCH_INDEXER_SECRET is empty string.
+    """
+    indexer_settings.SEARCH_INDEXER_SECRET = ""
+
+    with pytest.raises(ImproperlyConfigured) as exc_info:
+        SearchIndexer()
+
+    assert "SEARCH_INDEXER_SECRET must be set in Django settings." in str(
+        exc_info.value
+    )
+
+
+def test_services_search_endpoint_is_none(indexer_settings):
+    """
+    Indexer should raise RuntimeError if SEARCH_INDEXER_QUERY_URL is None.
+    """
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = None
+
+    with pytest.raises(ImproperlyConfigured) as exc_info:
+        SearchIndexer()
+
+    assert "SEARCH_INDEXER_QUERY_URL must be set in Django settings." in str(
+        exc_info.value
+    )
+
+
+def test_services_search_endpoint_is_empty(indexer_settings):
+    """
+    Indexer should raise RuntimeError if SEARCH_INDEXER_QUERY_URL is empty.
+    """
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = ""
+
+    with pytest.raises(ImproperlyConfigured) as exc_info:
+        SearchIndexer()
+
+    assert "SEARCH_INDEXER_QUERY_URL must be set in Django settings." in str(
+        exc_info.value
+    )
+
+
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_serialize_document_returns_expected_json():
+    """
+    It should serialize documents with correct metadata and access control.
+    """
+    user_a, user_b = factories.UserFactory.create_batch(2)
+    document = factories.DocumentFactory()
+    factories.DocumentFactory(parent=document)
+
+    factories.UserDocumentAccessFactory(document=document, user=user_a)
+    factories.UserDocumentAccessFactory(document=document, user=user_b)
+    factories.TeamDocumentAccessFactory(document=document, team="team1")
+    factories.TeamDocumentAccessFactory(document=document, team="team2")
+
+    accesses = {
+        document.path: {
+            "users": {str(user_a.sub), str(user_b.sub)},
+            "teams": {"team1", "team2"},
+        }
+    }
+
+    indexer = SearchIndexer()
+    result = indexer.serialize_document(document, accesses)
+
+    assert set(result.pop("users")) == {str(user_a.sub), str(user_b.sub)}
+    assert set(result.pop("groups")) == {"team1", "team2"}
+    assert result == {
+        "id": str(document.id),
+        "title": document.title,
+        "depth": 1,
+        "path": document.path,
+        "numchild": 1,
+        "content": utils.base64_yjs_to_text(document.content),
+        "created_at": document.created_at.isoformat(),
+        "updated_at": document.updated_at.isoformat(),
+        "reach": document.link_reach,
+        "size": 13,
+        "is_active": True,
+    }
+
+
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_serialize_document_deleted():
+    """Deleted documents are marked as just in the serialized json."""
+    parent = factories.DocumentFactory()
+    document = factories.DocumentFactory(parent=parent)
+
+    parent.soft_delete()
+    document.refresh_from_db()
+
+    indexer = SearchIndexer()
+    result = indexer.serialize_document(document, {})
+
+    assert result["is_active"] is False
+
+
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_serialize_document_empty():
+    """Empty documents returns empty content in the serialized json."""
+    document = factories.DocumentFactory(content="", title=None)
+
+    indexer = SearchIndexer()
+    result = indexer.serialize_document(document, {})
+
+    assert result["content"] == ""
+    assert result["title"] == ""
+
+
+@responses.activate
+def test_services_search_indexers_index_errors(indexer_settings):
+    """
+    Documents indexing response handling on Find API HTTP errors.
+    """
+    factories.DocumentFactory()
+
+    indexer_settings.SEARCH_INDEXER_URL = "http://app-find/api/v1.0/documents/index/"
+
+    responses.add(
+        responses.POST,
+        "http://app-find/api/v1.0/documents/index/",
+        status=401,
+        body=json_dumps({"message": "Authentication failed."}),
+    )
+
+    with pytest.raises(HTTPError):
+        SearchIndexer().index()
+
+
+@patch.object(SearchIndexer, "push")
+def test_services_search_indexers_batches_pass_only_batch_accesses(
+    mock_push, indexer_settings
+):
+    """
+    Documents indexing should be processed in batches,
+    and only the access data relevant to each batch should be used.
+    """
+    indexer_settings.SEARCH_INDEXER_BATCH_SIZE = 2
+    documents = factories.DocumentFactory.create_batch(5)
+
+    # Attach a single user access to each document
+    expected_user_subs = {}
+    for document in documents:
+        access = factories.UserDocumentAccessFactory(document=document)
+        expected_user_subs[str(document.id)] = str(access.user.sub)
+
+    assert SearchIndexer().index() == 5
+
+    # Should be 3 batches: 2 + 2 + 1
+    assert mock_push.call_count == 3
+
+    seen_doc_ids = set()
+
+    for call in mock_push.call_args_list:
+        batch = call.args[0]
+        assert isinstance(batch, list)
+
+        for doc_json in batch:
+            doc_id = doc_json["id"]
+            seen_doc_ids.add(doc_id)
+
+            # Only one user expected per document
+            assert doc_json["users"] == [expected_user_subs[doc_id]]
+            assert doc_json["groups"] == []
+
+    # Make sure all 5 documents were indexed
+    assert seen_doc_ids == {str(d.id) for d in documents}
+
+
+@patch.object(SearchIndexer, "push")
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_ignore_empty_documents(mock_push):
+    """
+    Documents indexing should be processed in batches,
+    and only the access data relevant to each batch should be used.
+    """
+    document = factories.DocumentFactory()
+    factories.DocumentFactory(content="", title="")
+    empty_title = factories.DocumentFactory(title="")
+    empty_content = factories.DocumentFactory(content="")
+
+    assert SearchIndexer().index() == 3
+
+    assert mock_push.call_count == 1
+
+    # Make sure only not eempty documents are indexed
+    results = {doc["id"] for doc in mock_push.call_args[0][0]}
+    assert results == {
+        str(d.id)
+        for d in (
+            document,
+            empty_content,
+            empty_title,
+        )
+    }
+
+
+@patch.object(SearchIndexer, "push")
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_ancestors_link_reach(mock_push):
+    """Document accesses and reach should take into account ancestors link reaches."""
+    great_grand_parent = factories.DocumentFactory(link_reach="restricted")
+    grand_parent = factories.DocumentFactory(
+        parent=great_grand_parent, link_reach="authenticated"
+    )
+    parent = factories.DocumentFactory(parent=grand_parent, link_reach="public")
+    document = factories.DocumentFactory(parent=parent, link_reach="restricted")
+
+    assert SearchIndexer().index() == 4
+
+    results = {doc["id"]: doc for doc in mock_push.call_args[0][0]}
+    assert len(results) == 4
+    assert results[str(great_grand_parent.id)]["reach"] == "restricted"
+    assert results[str(grand_parent.id)]["reach"] == "authenticated"
+    assert results[str(parent.id)]["reach"] == "public"
+    assert results[str(document.id)]["reach"] == "public"
+
+
+@patch.object(SearchIndexer, "push")
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_ancestors_users(mock_push):
+    """Document accesses and reach should include users from ancestors."""
+    user_gp, user_p, user_d = factories.UserFactory.create_batch(3)
+
+    grand_parent = factories.DocumentFactory(users=[user_gp])
+    parent = factories.DocumentFactory(parent=grand_parent, users=[user_p])
+    document = factories.DocumentFactory(parent=parent, users=[user_d])
+
+    assert SearchIndexer().index() == 3
+
+    results = {doc["id"]: doc for doc in mock_push.call_args[0][0]}
+    assert len(results) == 3
+    assert results[str(grand_parent.id)]["users"] == [str(user_gp.sub)]
+    assert set(results[str(parent.id)]["users"]) == {str(user_gp.sub), str(user_p.sub)}
+    assert set(results[str(document.id)]["users"]) == {
+        str(user_gp.sub),
+        str(user_p.sub),
+        str(user_d.sub),
+    }
+
+
+@patch.object(SearchIndexer, "push")
+@pytest.mark.usefixtures("indexer_settings")
+def test_services_search_indexers_ancestors_teams(mock_push):
+    """Document accesses and reach should include teams from ancestors."""
+    grand_parent = factories.DocumentFactory(teams=["team_gp"])
+    parent = factories.DocumentFactory(parent=grand_parent, teams=["team_p"])
+    document = factories.DocumentFactory(parent=parent, teams=["team_d"])
+
+    assert SearchIndexer().index() == 3
+
+    results = {doc["id"]: doc for doc in mock_push.call_args[0][0]}
+    assert len(results) == 3
+    assert results[str(grand_parent.id)]["groups"] == ["team_gp"]
+    assert set(results[str(parent.id)]["groups"]) == {"team_gp", "team_p"}
+    assert set(results[str(document.id)]["groups"]) == {"team_gp", "team_p", "team_d"}
+
+
+@patch("requests.post")
+def test_push_uses_correct_url_and_data(mock_post, indexer_settings):
+    """
+    push() should call requests.post with the correct URL from settings
+    the timeout set to 10 seconds and the data as JSON.
+    """
+    indexer_settings.SEARCH_INDEXER_URL = "http://example.com/index"
+
+    indexer = SearchIndexer()
+    sample_data = [{"id": "123", "title": "Test"}]
+
+    mock_response = mock_post.return_value
+    mock_response.raise_for_status.return_value = None  # No error
+
+    indexer.push(sample_data)
+
+    mock_post.assert_called_once()
+    args, kwargs = mock_post.call_args
+
+    assert args[0] == indexer_settings.SEARCH_INDEXER_URL
+    assert kwargs.get("json") == sample_data
+    assert kwargs.get("timeout") == 10
+
+
+def test_get_visited_document_ids_of():
+    """
+    get_visited_document_ids_of() returns the ids of the documents viewed
+    by the user BUT without specific access configuration (like public ones)
+    """
+    user = factories.UserFactory()
+    other = factories.UserFactory()
+    anonymous = AnonymousUser()
+    queryset = models.Document.objects.all()
+
+    assert not get_visited_document_ids_of(queryset, anonymous)
+    assert not get_visited_document_ids_of(queryset, user)
+
+    doc1, doc2, _ = factories.DocumentFactory.create_batch(3)
+
+    create_link = partial(models.LinkTrace.objects.create, user=user, is_masked=False)
+
+    create_link(document=doc1)
+    create_link(document=doc2)
+
+    # The third document is not visited
+    assert sorted(get_visited_document_ids_of(queryset, user)) == sorted(
+        [str(doc1.pk), str(doc2.pk)]
+    )
+
+    factories.UserDocumentAccessFactory(user=other, document=doc1)
+    factories.UserDocumentAccessFactory(user=user, document=doc2)
+
+    # The second document have an access for the user
+    assert get_visited_document_ids_of(queryset, user) == [str(doc1.pk)]
+
+
+@pytest.mark.usefixtures("indexer_settings")
+def test_get_visited_document_ids_of_deleted():
+    """
+    get_visited_document_ids_of() returns the ids of the documents viewed
+    by the user if they are not deleted.
+    """
+    user = factories.UserFactory()
+    anonymous = AnonymousUser()
+    queryset = models.Document.objects.all()
+
+    assert not get_visited_document_ids_of(queryset, anonymous)
+    assert not get_visited_document_ids_of(queryset, user)
+
+    doc = factories.DocumentFactory()
+    doc_deleted = factories.DocumentFactory()
+    doc_ancestor_deleted = factories.DocumentFactory(parent=doc_deleted)
+
+    create_link = partial(models.LinkTrace.objects.create, user=user, is_masked=False)
+
+    create_link(document=doc)
+    create_link(document=doc_deleted)
+    create_link(document=doc_ancestor_deleted)
+
+    # The all documents are visited
+    assert sorted(get_visited_document_ids_of(queryset, user)) == sorted(
+        [str(doc.pk), str(doc_deleted.pk), str(doc_ancestor_deleted.pk)]
+    )
+
+    doc_deleted.soft_delete()
+
+    # Only the first document is not deleted
+    assert get_visited_document_ids_of(queryset, user) == [str(doc.pk)]
+
+
+@responses.activate
+def test_services_search_indexers_search_errors(indexer_settings):
+    """
+    Documents indexing response handling on Find API HTTP errors.
+    """
+    factories.DocumentFactory()
+
+    indexer_settings.SEARCH_INDEXER_QUERY_URL = (
+        "http://app-find/api/v1.0/documents/search/"
+    )
+
+    responses.add(
+        responses.POST,
+        "http://app-find/api/v1.0/documents/search/",
+        status=401,
+        body=json_dumps({"message": "Authentication failed."}),
+    )
+
+    with pytest.raises(HTTPError):
+        SearchIndexer().search("alpha", token="mytoken")
+
+
+@patch("requests.post")
+def test_services_search_indexers_search(mock_post, indexer_settings):
+    """
+    search() should call requests.post to SEARCH_INDEXER_QUERY_URL with the
+    document ids from linktraces.
+    """
+    user = factories.UserFactory()
+    indexer = SearchIndexer()
+
+    mock_response = mock_post.return_value
+    mock_response.raise_for_status.return_value = None  # No error
+
+    doc1, doc2, _ = factories.DocumentFactory.create_batch(3)
+
+    create_link = partial(models.LinkTrace.objects.create, user=user, is_masked=False)
+
+    create_link(document=doc1)
+    create_link(document=doc2)
+
+    visited = get_visited_document_ids_of(models.Document.objects.all(), user)
+
+    indexer.search("alpha", visited=visited, token="mytoken")
+
+    args, kwargs = mock_post.call_args
+
+    assert args[0] == indexer_settings.SEARCH_INDEXER_QUERY_URL
+
+    query_data = kwargs.get("json")
+    assert query_data["q"] == "alpha"
+    assert sorted(query_data["visited"]) == sorted([str(doc1.pk), str(doc2.pk)])
+    assert query_data["services"] == ["docs"]
+    assert query_data["page_number"] == 1
+    assert query_data["page_size"] == 50
+    assert query_data["order_by"] == "updated_at"
+    assert query_data["order_direction"] == "desc"
+
+    assert kwargs.get("headers") == {"Authorization": "Bearer mytoken"}
+    assert kwargs.get("timeout") == 10

src/backend/core/tests/test_utils.py

@@ -75,3 +75,28 @@ def test_utils_extract_attachments():
     base64_string = base64.b64encode(update).decode("utf-8")
     # image_key2 is missing the "/media/" part and shouldn't get extracted
     assert utils.extract_attachments(base64_string) == [image_key1, image_key3]
+
+
+def test_utils_get_ancestor_to_descendants_map_single_path():
+    """Test ancestor mapping of a single path."""
+    paths = ["000100020005"]
+    result = utils.get_ancestor_to_descendants_map(paths, steplen=4)
+
+    assert result == {
+        "0001": {"000100020005"},
+        "00010002": {"000100020005"},
+        "000100020005": {"000100020005"},
+    }
+
+
+def test_utils_get_ancestor_to_descendants_map_multiple_paths():
+    """Test ancestor mapping of multiple paths with shared prefixes."""
+    paths = ["000100020005", "00010003"]
+    result = utils.get_ancestor_to_descendants_map(paths, steplen=4)
+
+    assert result == {
+        "0001": {"000100020005", "00010003"},
+        "00010002": {"000100020005"},
+        "000100020005": {"000100020005"},
+        "00010003": {"00010003"},
+    }

src/backend/core/utils.py

@@ -2,6 +2,7 @@
 
 import base64
 import re
+from collections import defaultdict
 
 import pycrdt
 from bs4 import BeautifulSoup
@@ -9,6 +10,27 @@ from bs4 import BeautifulSoup
 from core import enums
 
 
+def get_ancestor_to_descendants_map(paths, steplen):
+    """
+    Given a list of document paths, return a mapping of ancestor_path -> set of descendant_paths.
+
+    Each path is assumed to use materialized path format with fixed-length segments.
+
+    Args:
+        paths (list of str): List of full document paths.
+        steplen (int): Length of each path segment.
+
+    Returns:
+        dict[str, set[str]]: Mapping from ancestor path to its descendant paths (including itself).
+    """
+    ancestor_map = defaultdict(set)
+    for path in paths:
+        for i in range(steplen, len(path) + 1, steplen):
+            ancestor = path[:i]
+            ancestor_map[ancestor].add(path)
+    return ancestor_map
+
+
 def filter_descendants(paths, root_paths, skip_sorting=False):
     """
     Filters paths to keep only those that are descendants of any path in root_paths.

src/backend/demo/management/commands/create_demo.py

@@ -1,16 +1,19 @@
 # ruff: noqa: S311, S106
 """create_demo management command"""
 
+import base64
 import logging
 import math
 import random
 import time
 from collections import defaultdict
+from uuid import uuid4
 
 from django import db
 from django.conf import settings
 from django.core.management.base import BaseCommand, CommandError
 
+import pycrdt
 from faker import Faker
 
 from core import models
@@ -27,6 +30,16 @@ def random_true_with_probability(probability):
     return random.random() < probability
 
 
+def get_ydoc_for_text(text):
+    """Return a ydoc from plain text for demo purposes."""
+    ydoc = pycrdt.Doc()
+    paragraph = pycrdt.XmlElement("p", {}, [pycrdt.XmlText(text)])
+    fragment = pycrdt.XmlFragment([paragraph])
+    ydoc["document-store"] = fragment
+    update = ydoc.get_update()
+    return base64.b64encode(update).decode("utf-8")
+
+
 class BulkQueue:
     """A utility class to create Django model instances in bulk by just pushing to a queue."""
 
@@ -48,7 +61,7 @@ class BulkQueue:
         self.queue[objects[0]._meta.model.__name__] = []  # noqa: SLF001
 
     def push(self, obj):
-        """Add a model instance to queue to that it gets created in bulk."""
+        """Add a model instance to queue so that it gets created in bulk."""
         objects = self.queue[obj._meta.model.__name__]  # noqa: SLF001
         objects.append(obj)
         if len(objects) > self.BATCH_SIZE:
@@ -139,17 +152,19 @@ def create_demo(stdout):
             # pylint: disable=protected-access
             key = models.Document._int2str(i)  # noqa: SLF001
             padding = models.Document.alphabet[0] * (models.Document.steplen - len(key))
-            queue.push(
-                models.Document(
-                    depth=1,
-                    path=f"{padding}{key}",
-                    creator_id=random.choice(users_ids),
-                    title=fake.sentence(nb_words=4),
-                    link_reach=models.LinkReachChoices.AUTHENTICATED
-                    if random_true_with_probability(0.5)
-                    else random.choice(models.LinkReachChoices.values),
-                )
+            title = fake.sentence(nb_words=4)
+            document = models.Document(
+                id=uuid4(),
+                depth=1,
+                path=f"{padding}{key}",
+                creator_id=random.choice(users_ids),
+                title=title,
+                link_reach=models.LinkReachChoices.AUTHENTICATED
+                if random_true_with_probability(0.5)
+                else random.choice(models.LinkReachChoices.values),
             )
+            document.save_content(get_ydoc_for_text(f"Content for {title:s}"))
+            queue.push(document)
 
         queue.flush()
 

src/backend/impress/settings.py

@@ -99,6 +99,28 @@ class Base(Configuration):
     }
     DEFAULT_AUTO_FIELD = "django.db.models.AutoField"
 
+    # Search
+    SEARCH_INDEXER_CLASS = values.Value(
+        default=None,
+        environ_name="SEARCH_INDEXER_CLASS",
+        environ_prefix=None,
+    )
+    SEARCH_INDEXER_BATCH_SIZE = values.IntegerValue(
+        default=100_000, environ_name="SEARCH_INDEXER_BATCH_SIZE", environ_prefix=None
+    )
+    SEARCH_INDEXER_URL = values.Value(
+        default=None, environ_name="SEARCH_INDEXER_URL", environ_prefix=None
+    )
+    SEARCH_INDEXER_COUNTDOWN = values.IntegerValue(
+        default=1, environ_name="SEARCH_INDEXER_COUNTDOWN", environ_prefix=None
+    )
+    SEARCH_INDEXER_SECRET = values.Value(
+        default=None, environ_name="SEARCH_INDEXER_SECRET", environ_prefix=None
+    )
+    SEARCH_INDEXER_QUERY_URL = values.Value(
+        default=None, environ_name="SEARCH_INDEXER_QUERY_URL", environ_prefix=None
+    )
+
     # Static files (CSS, JavaScript, Images)
     STATIC_URL = "/static/"
     STATIC_ROOT = os.path.join(DATA_DIR, "static")

src/frontend/apps/impress/src/features/docs/doc-management/api/useDocs.tsx

@@ -44,7 +44,16 @@ export const constructParams = (params: DocsParams): URLSearchParams => {
 export type DocsResponse = APIList<Doc>;
 export const getDocs = async (params: DocsParams): Promise<DocsResponse> => {
   const searchParams = constructParams(params);
-  const response = await fetchAPI(`documents/?${searchParams.toString()}`);
+  let response
+
+  // HACK for fulltext search feature
+  if (searchParams.has('title')) {
+    searchParams.set('q', searchParams.get('title') || '');
+    searchParams.delete('title');
+    response = await fetchAPI(`documents/search?${searchParams.toString()}`);
+  } else {
+    response = await fetchAPI(`documents/?${searchParams.toString()}`);
+  }
 
   if (!response.ok) {
     throw new APIError('Failed to get the docs', await errorCauses(response));