Applying evaluation framework

config/config.yaml

@@ -4,8 +4,9 @@ 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_name: BAAI/bge-m3
+embeddings_model_size: 1024
 
 vector_store: 'qdrant'  # {'qdrant', 'pinecone', 'supabase'}
 top_k_results: 10
@@ -27,11 +28,18 @@ prompt_system_context: |
 collections:
   - justicio
   - boe
+  - boe-multilingual-e5-large
+  - boe-bge-m3
   - bocm
+  - bocm-multilingual-e5-large
   - bopz
+  - bopz-multilingual-e5-large
   - bopv
+  - bopv-multilingual-e5-large
   - boja
+  - boja-multilingual-e5-large
   - boa
+  - boa-multilingual-e5-large
 
 # Openai
 llm_model_name: 'gpt-3.5-turbo-0125'  # 'gpt-3.5-turbo-1106', 'gpt-4-1106-preview'

doc/qdrant/queries.json

@@ -4,6 +4,20 @@ GET collections
 // Get collection info
 GET collections/justicio
 
+// Create a new collection
+// https://qdrant.tech/documentation/tutorials/bulk-upload/
+PUT /collections/{collection_name}
+{
+    "vectors": {
+      "size": 1024,
+      "distance": "Cosine"
+    },
+    "shard_number": 2,
+    "optimizers_config": {
+        "indexing_threshold": 0
+    }
+}
+
 // List points in a collection, using filter by metadata
 POST collections/justicio/points/scroll
 {
@@ -77,5 +91,21 @@ POST collections/justicio/points/delete
   }
 }
 
+// No funciona porque tenemos las fechas como strings en lugar de como números.
+POST /collections/boe-bge-m3/points/count
+{
+    "filter": {
+        "should": [
+            {
+                "key": "metadata.anio",
+                "range": {
+                    "gte": 2000
+                }
+            }
+        ]
+    }
+}
+
+
 // https://qdrant.tech/documentation/concepts/filtering/
 // https://qdrant.tech/documentation/concepts/points/