ArangoDB: Vector Store

comps/vectorstores/arango/README.md

@@ -0,0 +1,57 @@
+# Start ArangoDB server
+
+**Additional Links**:
+- https://arangodb.com/2024/11/vector-search-in-arangodb-practical-insights-and-hands-on-examples/
+
+## 1. Start ArangoDB via Docker
+
+```bash
+docker run -d --name arango -p 8529:8529 -e ARANGO_ROOT_PASSWORD=test arangodb/arangodb:3.12
+```
+
+## 2. Create a Vector Index
+
+**Using `arangosh`**:
+
+```bash
+127.0.0.1:8529@_system > db.myCollection.ensureIndex(
+{
+        name: "my-vector-index",
+        type: "vector",
+        fields: ["embeddings"]
+        params: { metric: "cosine", dimension: 128, nLists: 100 }
+}
+```
+
+**Using the `python-arango` driver**:
+
+```python
+from arango import ArangoClient
+
+db = ArangoClient().db('_system', username='root', password='test')
+
+db.collection("myCollection").add_index(
+    {
+        "name": "my-vector-index",
+        "type": "vector",
+        "fields": ["embeddings"],
+        "params": {
+            "metric": "cosine",
+            "dimensions": 128,
+            "nLists": 100,
+        },
+    }
+)
+```
+
+## 3. Use the Vector Index via the Arango Query Language (AQL)
+
+```bash
+LET query_embedding = [0.1, 0.3, 0.5, …]
+
+FOR doc IN myCollection
+    LET score = APPROX_NEAR_COSINE(doc.embeddings, query_embedding)
+    SORT score DESC
+    LIMIT 5
+    RETURN {doc, score}   
+```

comps/vectorstores/arango/__init__.py

@@ -0,0 +1,2 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0

comps/vectorstores/arango/docker-compose.yml

@@ -0,0 +1,11 @@
+# Copyright (C) 2024 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+
+services:
+  arango:
+    image: arangodb/arangodb:3.12
+    container_name: arangodb
+    ports:
+      - 8529:8529
+    environment:
+      ARANGO_ROOT_PASSWORD: ${ARANGO_ROOT_PASSWORD}