Merge pull request #68 from bukosabino/develop-new-embedding-model

New embedding model and collection

.gitignore

@@ -169,6 +169,7 @@ cython_debug/
 
 .secrets
 .continous_deployment.md
+.qdrant.yaml
 cron_etl_daily.sh
 cron_etl_initial.sh
 

config/config.yaml

@@ -4,8 +4,8 @@ chunk_overlap: 100
 
 admin_email: [email protected]
 
-embeddings_model_name: dariolopez/roberta-base-bne-finetuned-msmarco-qa-es-mnrl-mn
-embeddings_model_size: 768
+embeddings_model_name: intfloat/multilingual-e5-large
+embeddings_model_size: 1024
 
 vector_store: 'qdrant'  # {'qdrant', 'pinecone', 'supabase'}
 top_k_results: 10
@@ -26,6 +26,7 @@ prompt_system_context: |
 # Qdrant
 collections:
   - justicio
+  - boe
   - bocm
   - bopz
 

doc/qdrant/queries.json

@@ -35,6 +35,32 @@ POST collections/justicio/points/count
   }
 }
 
+// Count points in a collection, using filter by multiple metadata
+POST collections/justicio/points/count
+{
+  "filter": {
+    "must": [
+      {
+        "key": "metadata.anio",
+        "match": {
+          "value": "2024"
+        }
+      },
+      {
+        "key": "metadata.mes",
+        "match": {
+          "value": "02"
+        }
+      },
+      {
+        "key": "metadata.dia",
+        "match": {
+          "value": "20"
+        }
+      }
+    ]
+  }
+}
 
 // Delete points in a collection, using filter by metadata
 POST collections/justicio/points/delete