Ouverture de la recherche par mot clé (#1073)

* add geo_range in meta

* add city field

* tests for semantic search

* factor autocomplete

lemarche/api/perimeters/tests.py

@@ -93,6 +93,12 @@ def test_should_filter_perimeters_autocomplete_by_q_code(self):
         self.assertEqual(len(response.data), 1 + 1)
         self.assertEqual(response.data[0]["name"], "Isère")
 
+    def test_should_filter_perimeters_autocomplete_by_q_code_only_cities(self):
+        url = reverse("api:perimeters-autocomplete-list") + "?q=38&kind=CITY"  # anonymous user
+        response = self.client.get(url)
+        self.assertEqual(len(response.data), 1)
+        self.assertEqual(response.data[0]["name"], "Grenoble")
+
     def test_should_filter_perimeters_autocomplete_by_q_post_code(self):
         url = reverse("api:perimeters-autocomplete-list") + "?q=38100"  # anonymous user
         response = self.client.get(url)

lemarche/api/perimeters/views.py

@@ -41,6 +41,12 @@ def list(self, request, *args, **kwargs):
         """
         return super().list(request, args, kwargs)
 
+    def get_queryset(self):
+        kind = self.request.query_params.get("kind", None)
+        if kind and kind in [id for (id, name) in Perimeter.KIND_CHOICES]:
+            return Perimeter.objects.filter(kind=kind)
+        return self.queryset
+
 
 class PerimeterKindViewSet(mixins.ListModelMixin, viewsets.GenericViewSet):
     serializer_class = PerimeterChoiceSerializer

lemarche/perimeters/models.py

@@ -12,6 +12,9 @@
 
 
 class PerimeterQuerySet(models.QuerySet):
+    def cities(self):
+        return self.filter(kind="CITY")
+
     def regions(self):
         return self.filter(kind="REGION")
 

lemarche/siaes/models.py

@@ -1180,6 +1180,16 @@ def elasticsearch_index_metadata(self):
                 "lat": self.latitude,
                 "lon": self.longitude,
             }
+
+        if self.geo_range == siae_constants.GEO_RANGE_COUNTRY:
+            metadata["geo_country"] = True
+        elif self.geo_range == siae_constants.GEO_RANGE_REGION:
+            metadata["geo_reg"] = self.region
+        elif self.geo_range == siae_constants.GEO_RANGE_DEPARTMENT:
+            metadata["geo_dep"] = self.department
+        elif self.geo_range == siae_constants.GEO_RANGE_CUSTOM:
+            metadata["geo_dist"] = self.geo_range_custom_distance
+
         return metadata
 
     def sectors_list_string(self, display_max=3):

lemarche/static/js/perimeter_autocomplete_field.js

@@ -29,16 +29,18 @@ const debounce = (callback, wait) => {
   };
 }
 
-async function fetchSource(query) {
-  const res = await fetch(`${API_ENDPOINT}?q=${query}&results=10`);
+async function fetchSource(query, kind = undefined) {
+  const res = await fetch(`${API_ENDPOINT}?q=${query}&results=10${kind ? `&kind=${kind}` : ''}`);
   const data = await res.json();
   return data;  // data.results
 }
 
 class PerimeterAutocomplete {
-  constructor(perimeter_container_name, perimeter_input_id) {
+  constructor(perimeter_container_name, perimeter_input_id, perimeter_placeholder='Région, ville…', perimeter_kind = undefined) {
     this.perimeter_container_name= perimeter_container_name;
     this.perimeter_input_id= perimeter_input_id;
+    this.perimeter_kind= perimeter_kind;
+    this.perimeter_placeholder= perimeter_placeholder;
     this.perimeter_name_input_id= `${this.perimeter_input_id}_name`;
     this.perimeterAutocompleteContainer = document.getElementById(perimeter_container_name);
     this.perimeterInput = document.getElementById(perimeter_input_id);  // hidden
@@ -53,10 +55,10 @@ class PerimeterAutocomplete {
         element: this.perimeterAutocompleteContainer,
         id: this.perimeter_name_input_id,
         name: this.perimeter_name_input_id,  // url GET param name
-        placeholder: 'Région, ville…',  // 'Autour de (Arras, Bobigny, Strasbourg…)', 'Région, département, ville'
+        placeholder: this.perimeter_placeholder,
         minLength: 2,
         defaultValue: this.initial_value_name,
-        source:  this.getSource,
+        source: this.perimeter_kind === 'CITY' ? this.getSourceCity : this.getSource,
         displayMenu: 'overlay',
         templates: {
           inputValue: this.inputValue,  // returns the string value to be inserted into the input
@@ -89,6 +91,11 @@ class PerimeterAutocomplete {
     populateResults(res);
   }
 
+  async getSourceCity(query, populateResults) {
+    const res = await fetchSource(query, "CITY");
+    populateResults(res);
+  }
+
   inputValue(result) {
     // strip html from suggestion
     if(!result) {
@@ -220,7 +227,7 @@ class PerimetersMultiAutocomplete {
   inputValue(result) {
     return "";
   }
-  
+
   inputValueHiddenField(result) {
     // we want to avoid clicks outside that return 'undefined'
     if (result) {
@@ -239,20 +246,20 @@ class PerimetersMultiAutocomplete {
   suggestion(result) {
     // display suggestions as `name (kind)`
     let resultName, resultKind = '';
-  
+
     // build resultName & resultKind from the result object
     if (typeof result === 'object') {
       resultName = result.name;
       resultKind = (result.kind === 'CITY') ? result.department_code : KIND_MAPPING[result.kind];
     }
-  
+
     // Edge case: if there is an initial value
     // reconstruct resultName & resultKind from the result string
     if (typeof result === 'string') {
       resultName = result.substring(0, result.lastIndexOf(' '));
       resultKind = result.includes('(') ? result.substring(result.lastIndexOf(' ') + 2, result.length - 1) : '';
     }
-  
+
     let nameWithKind = '<strong>' + resultName + '</strong>';
     if (resultKind) {
       nameWithKind += ' <small>(' + resultKind + ')</small>';
@@ -266,7 +273,7 @@ class PerimetersMultiAutocomplete {
     $(`#${idRefInput}`).remove();
     $(this).remove();
   }
-  
+
   createHiddenInputPerimeter(resultId, resultName) {
     let removeIcon = $('<i>', { class: "ri-close-line font-weight-bold mr-0", "aria-hidden": true });
     let resultIdString = `${this.perimeter_hidden_input_selector_prefix}-${resultId}`;

lemarche/templates/siaes/search_results.html

@@ -53,7 +53,6 @@
                                 Recherche par SIRET / nom
                             </a>
                         </li>
-                        {% if is_admin %}
                         <li class="nav-item" role="presentation">
                             <a class="nav-link"
                                 id="search-semantic-tab"
@@ -62,11 +61,11 @@
                                 role="tab"
                                 aria-controls="search-semantic"
                                 aria-selected="false">
-                                Recherche sémantique
-                                <span class="badge badge-sm badge-pill badge-important ml-2">Nouveauté</span>
+                                Recherche par mot clé
+                                <span class="badge badge-sm badge-pill badge-marche ml-2">Beta</span>
+                                <span class="badge badge-sm badge-pill badge-important">Nouveauté</span>
                             </a>
                         </li>
-                        {% endif %}
                         <li class="nav-item-dropdown dropdown">
                             <a class="nav-link dropdown-toggle"
                                href="#"
@@ -205,7 +204,6 @@
                                 </div>
                             </form>
                         </div>
-                        {% if is_admin %}
                         <div class="tab-pane fade"
                             id="search-semantic"
                             role="tabpanel"
@@ -218,12 +216,17 @@
                                 id="text-semantic-search-form">
                                 {% bootstrap_form_errors form_semantic type="all" %}
                                 <div class="row">
-                                    <div class="col-12 col-lg-8">
-                                        <div class="row">
-                                            <div class="col-12">{% bootstrap_field form_semantic.search_query %}</div>
+                                    <div class="col-12 col-lg-5">
+                                        {% bootstrap_field form_semantic.search_query %}
+                                    </div>
+                                    <div class="col-12 col-lg-5">
+                                        <div class="form-group">
+                                            <label for="id_city">{{ form_semantic.city.label }}</label>
+                                            <div id="dir_form_city" data-input-name="{{ form_semantic.city.name }}"></div>
+                                            {{ form_semantic.city }}
                                         </div>
                                     </div>
-                                    <div class="col-12 col-lg-4">
+                                    <div class="col-12 col-lg-2">
                                         <span class="mb-2 d-none d-md-inline-block">&nbsp;</span>
                                         <button id="text-search-submit"
                                                 class="btn btn-primary btn-block btn-ico"
@@ -235,7 +238,6 @@
                                 </div>
                             </form>
                         </div>
-                        {% endif %}
                     </div>
                 </div>
             </div>
@@ -363,6 +365,10 @@ <h1 class="h4 mb-0">
         const locationsAutoComplete = new PerimetersMultiAutocomplete(LOCATION_AUTOCOMPLETE_ID, LOCATION_AUTOCOMPLETE_CONTAINER_SELECTOR, LOCATION_SELECTED_CONTAINER_SELECTOR, LOCATION_HIDDEN_INPUT_SELECTOR_PREFIX, LOCATION_CURRENT_ID);
         locationsAutoComplete.init();
     {% endif %}
+
+    // init city form fields
+    const cityAutoComplete = new PerimeterAutocomplete("dir_form_city", "id_city", "Ville", "CITY");
+    cityAutoComplete.init();
 });
     </script>
     <script type="text/javascript">

lemarche/templates/siaes/semantic_search_results.html

@@ -2,6 +2,17 @@
 {% block content %}
     <div class="col-12 col-lg-8">
         <div class="c-box mb-3">
+            <div class="d-flex align-items-center mb-3">
+                <div class="flex-grow-1">
+                    <h1 class="h4 mb-0">
+                        {% if siaes %}
+                                {{ siaes.count }} prestataire{{ siaes.count|pluralize }} correspond{{ siaes.count|pluralize:"ent" }} à vos critères
+                        {% else %}
+                            Oups, aucun prestataire trouvé !
+                        {% endif %}
+                    </h1>
+                </div>
+            </div>
             {% if siaes %}
                 {% for siae in siaes %}
                     {% include "siaes/_card_search_result.html" with siae=siae %}

lemarche/tenders/models.py

@@ -565,7 +565,7 @@ def set_siae_found_list(self):
                 and self.distance_location > 0
             ):
                 # with geo distance
-                siae_ids = api_elasticsearch.siaes_similarity_search(
+                siae_ids = api_elasticsearch.siaes_similarity_search_with_geo_distance(
                     self.description,
                     geo_distance=self.distance_location,
                     geo_lat=self.location.coords.y,

lemarche/utils/apis/api_elasticsearch.py

@@ -2,6 +2,8 @@
 from langchain.embeddings.openai import OpenAIEmbeddings
 from langchain.vectorstores.elasticsearch import ElasticsearchStore
 
+from lemarche.perimeters.models import Perimeter
+
 
 BASE_URL = f"{settings.ELASTICSEARCH_HOST}:{settings.ELASTICSEARCH_PORT}"
 URL = f"{settings.ELASTICSEARCH_SCHEME}://{BASE_URL}"
@@ -10,7 +12,7 @@
 )
 
 
-def siaes_similarity_search(search_text: str, geo_distance: int = None, geo_lat: float = None, geo_lon: float = None):
+def siaes_similarity_search(search_text: str, search_filter: dict = {}):
     """Performs semantic search with Elasticsearch as a vector db
 
     Args:
@@ -26,7 +28,20 @@ def siaes_similarity_search(search_text: str, geo_distance: int = None, geo_lat:
         es_url=URL,
         index_name=settings.ELASTICSEARCH_INDEX_SIAES,
     )
+    similar_docs = db.similarity_search_with_score(search_text, k=50, filter=search_filter)
+    siaes_id = []
+    for similar_doc, similar_score in similar_docs:
+        # Discussion to understand score :
+        # https://github.com/langchain-ai/langchain/discussions/9984#discussioncomment-6860841
+        if similar_score > settings.ELASTICSEARCH_MIN_SCORE:
+            siaes_id.append(similar_doc.metadata["id"])
+
+    return siaes_id
+
 
+def siaes_similarity_search_with_geo_distance(
+    search_text: str, geo_distance: int = None, geo_lat: float = None, geo_lon: float = None
+):
     search_filter = []
     if geo_distance and geo_lat and geo_lon:
         search_filter = [
@@ -41,12 +56,36 @@ def siaes_similarity_search(search_text: str, geo_distance: int = None, geo_lat:
             }
         ]
 
-    similar_docs = db.similarity_search_with_score(search_text, k=50, filter=search_filter)
-    siaes_id = []
-    for similar_doc, similar_score in similar_docs:
-        # Discussion to understand score :
-        # https://github.com/langchain-ai/langchain/discussions/9984#discussioncomment-6860841
-        if similar_score > settings.ELASTICSEARCH_MIN_SCORE:
-            siaes_id.append(similar_doc.metadata["id"])
+    return siaes_similarity_search(search_text, search_filter)
 
-    return siaes_id
+
+def siaes_similarity_search_with_city(search_text: str, city: Perimeter):
+    search_filter = [
+        {
+            "bool": {
+                "should": [
+                    {"bool": {"must": [], "filter": [{"match_phrase": {"metadata.geo_country": True}}]}},
+                    {"bool": {"must": [], "filter": [{"match_phrase": {"metadata.geo_reg": city.region_code}}]}},
+                    {"bool": {"must": [], "filter": [{"match_phrase": {"metadata.geo_dep": city.department_code}}]}},
+                    {
+                        "bool": {
+                            "must": [],
+                            "filter": [
+                                {
+                                    "geo_distance": {
+                                        "distance": "50km",
+                                        "metadata.geo_location": {
+                                            "lat": city.latitude,
+                                            "lon": city.longitude,
+                                        },
+                                    }
+                                }
+                            ],
+                        }
+                    },
+                ],
+                "minimum_should_match": 1,
+            }
+        }
+    ]
+    return siaes_similarity_search(search_text, search_filter)

lemarche/www/siaes/forms.py

@@ -497,13 +497,20 @@ def filter_queryset(self, qs=None):
 
 class SiaeSemanticForm(forms.Form):
     search_query = forms.CharField(
-        label="Recherche sémantique",
+        label="Prestation recherchée",
         required=False,
-        widget=forms.TextInput(attrs={"placeholder": "Je cherche…"}),
-        help_text=" ".join(
-            [
-                "Soyez le plus précis possible (Exemple: nettoyage des locaux d'entreprise",
-                "/ entretien des espaces verts)",
-            ]
-        ),
+        widget=forms.TextInput(attrs={"placeholder": "Nettoyage de locaux"}),
+        min_length=5,
+    )
+    city = forms.ModelChoiceField(
+        label="Localisation de votre besoin",
+        queryset=Perimeter.objects.cities(),
+        to_field_name="slug",
+        required=False,
+        widget=forms.HiddenInput(),
     )
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        # this field are autocompletes
+        self.fields["city"].choices = []