test(singlestore): add semantic search integration test and docs

.ci/integration.cloudbuild.yaml

@@ -1356,7 +1356,7 @@ steps:
       - "SINGLESTORE_USER=$_SINGLESTORE_USER"
       - "SINGLESTORE_DATABASE=$_SINGLESTORE_DATABASE"
       - "SERVICE_ACCOUNT_EMAIL=$SERVICE_ACCOUNT_EMAIL"
-    secretEnv: ["SINGLESTORE_PASSWORD", "SINGLESTORE_HOST", "CLIENT_ID"]
+    secretEnv: ["SINGLESTORE_PASSWORD", "SINGLESTORE_HOST", "CLIENT_ID", "API_KEY"]
     volumes:
       - name: "go"
         path: "/gopath"
@@ -1370,7 +1370,9 @@ steps:
           .ci/test_with_coverage.sh \
             "SingleStore" \
             singlestore \
-            singlestore
+            singlestore \
+            "" \
+            "API_KEY"
         else
           echo "No relevant changes for SingleStore. Skipping shard."
           exit 0

docs/en/integrations/singlestore/tools/singlestore-sql.md

@@ -91,6 +91,75 @@ templateParameters:
     description: Table to select from
 ```
 
+### Example with Vector Search
+
+SingleStore supports vector operations. When using an `embeddingModel` with a `singlestore-sql` tool, the tool automatically converts text parameters into a JSON string array. You can then use SingleStore's `JSON_ARRAY_PACK()` function in your SQL statement to pack this string into a binary vector format (BLOB) for vector storage and similarity search.
+
+#### Define the Embedding Model
+See [EmbeddingModels](../../../documentation/configuration/embedding-models/_index.md) for more information.
+
+```yaml
+kind: embeddingModel
+name: gemini-model
+type: gemini
+model: gemini-embedding-001
+apiKey: ${GOOGLE_API_KEY}
+dimension: 768
+```
+
+#### Vector Ingestion Tool
+This tool stores both the raw text and its vector representation. It uses `valueFromParam` to hide the vector conversion logic from the LLM, ensuring the Agent only has to provide the content once.
+
+```yaml
+kind: tool
+name: insert_doc_singlestore
+type: singlestore-sql
+source: my-s2-source
+statement: |
+  INSERT INTO vector_table (id, content, embedding)
+  VALUES (1, ?, JSON_ARRAY_PACK(?))
+description: |
+  Index new documents for semantic search in SingleStore.
+parameters:
+  - name: content
+    type: string
+    description: The text content to store.
+  - name: text_to_embed
+    type: string
+    # Automatically copies 'content' and converts it to a vector string array
+    valueFromParam: content
+    embeddedBy: gemini-model
+```
+
+#### Vector Search Tool
+This tool allows the Agent to perform a natural language search. The query string provided by the Agent is converted into a vector string array before the SQL is executed.
+
+```yaml
+kind: tool
+name: search_docs_singlestore
+type: singlestore-sql
+source: my-s2-source
+statement: |
+  SELECT 
+    id, 
+    content, 
+    DOT_PRODUCT(embedding, JSON_ARRAY_PACK(?)) AS score 
+  FROM 
+    vector_table 
+  ORDER BY 
+    score DESC
+  LIMIT 1
+description: |
+  Search for documents in SingleStore using natural language. 
+  Returns the most semantically similar result.
+parameters:
+  - name: query
+    type: string
+    description: The search query to be converted to a vector.
+    embeddedBy: gemini-model
+```
+
+
 ## Reference
 
 | **field**          |                   **type**                   | **required** | **description**                                                                                                                        |

internal/tools/singlestore/singlestoreexecutesql/singlestoreexecutesql.go

@@ -128,7 +128,7 @@ func (t Tool) Invoke(ctx context.Context, resourceMgr tools.SourceProvider, para
 }
 
 func (t Tool) EmbedParams(ctx context.Context, paramValues parameters.ParamValues, embeddingModelsMap map[string]embeddingmodels.EmbeddingModel) (parameters.ParamValues, error) {
-	return parameters.EmbedParams(ctx, t.Parameters, paramValues, embeddingModelsMap, nil)
+	return parameters.EmbedParams(ctx, t.Parameters, paramValues, embeddingModelsMap, embeddingmodels.FormatVectorForPgvector)
 }
 
 func (t Tool) Manifest() tools.Manifest {

internal/tools/singlestore/singlestoresql/singlestoresql.go

@@ -156,7 +156,7 @@ func (t Tool) Invoke(ctx context.Context, resourceMgr tools.SourceProvider, para
 }
 
 func (t Tool) EmbedParams(ctx context.Context, paramValues parameters.ParamValues, embeddingModelsMap map[string]embeddingmodels.EmbeddingModel) (parameters.ParamValues, error) {
-	return parameters.EmbedParams(ctx, t.AllParams, paramValues, embeddingModelsMap, nil)
+	return parameters.EmbedParams(ctx, t.AllParams, paramValues, embeddingModelsMap, embeddingmodels.FormatVectorForPgvector)
 }
 
 func (t Tool) Manifest() tools.Manifest {

tests/singlestore/singlestore_integration_test.go

@@ -263,6 +263,9 @@ func TestSingleStoreToolEndpoints(t *testing.T) {
 	tmplSelectCombined, tmplSelectFilterCombined := getSingleStoreTmplToolStatement()
 	toolsFile = tests.AddTemplateParamConfig(t, toolsFile, SingleStoreToolType, tmplSelectCombined, tmplSelectFilterCombined, "")
 
+	insertStmt := `INSERT INTO senseai_docs (content, embedding) VALUES (?, JSON_ARRAY_PACK(?))`
+	searchStmt := `SELECT content FROM senseai_docs ORDER BY DOT_PRODUCT(embedding, JSON_ARRAY_PACK(?)) DESC LIMIT 1`
+	toolsFile = tests.AddSemanticSearchConfig(t, toolsFile, SingleStoreToolType, insertStmt, searchStmt)
 	cmd, cleanup, err := tests.StartCmd(ctx, toolsFile, args...)
 	if err != nil {
 		t.Fatalf("command initialization returned an error: %s", err)
@@ -286,4 +289,22 @@ func TestSingleStoreToolEndpoints(t *testing.T) {
 	tests.RunMCPToolCallMethod(t, mcpMyFailToolWant, mcpSelect1Want)
 	tests.RunExecuteSqlToolInvokeTest(t, createTableStatement, select1Want)
 	tests.RunToolInvokeWithTemplateParameters(t, tableNameTemplateParam)
+
+	// Create table for semantic search
+	_, err = pool.ExecContext(ctx, "CREATE TABLE IF NOT EXISTS senseai_docs (id INT AUTO_INCREMENT PRIMARY KEY, content TEXT, embedding BLOB);")
+	if err != nil {
+		t.Fatalf("unable to create semantic search table: %s", err)
+	}
+	defer func() {
+		_, err = pool.ExecContext(ctx, "DROP TABLE IF EXISTS senseai_docs;")
+		if err != nil {
+			t.Logf("Teardown failed: %s", err)
+		}
+	}()
+
+	// Semantic search tests
+	httpSemanticInsertWant := `null`
+	mcpSemanticInsertWant := ``
+	semanticSearchWant := `The quick brown fox jumps over the lazy dog`
+	tests.RunSemanticSearchToolInvokeTest(t, httpSemanticInsertWant, mcpSemanticInsertWant, semanticSearchWant)
 }