feat(codegen): add --auto-embed support for create/update mutations

Author: pyramation

graphql/codegen/src/core/codegen/cli/table-command-generator.ts

@@ -407,6 +407,79 @@ function buildAutoEmbedBlock(
   );
 }
 
+/**
+ * Build an auto-embed block for mutation handlers (create/update).
+ *
+ * Generates code equivalent to:
+ *   if (argv['auto-embed']) {
+ *     const embedder = resolveEmbedder();
+ *     if (!embedder) {
+ *       console.error('--auto-embed requires an embedder. Set EMBEDDER_PROVIDER=ollama');
+ *       process.exit(1);
+ *     }
+ *     await autoEmbedInput(cleanedData, ['embedding'], embedder);
+ *   }
+ *
+ * @param vectorFieldNames - Names of vector embedding fields detected at codegen time
+ */
+function buildAutoEmbedInputBlock(
+  vectorFieldNames: string[],
+): t.IfStatement {
+  const fieldNamesArray = t.arrayExpression(
+    vectorFieldNames.map((n) => t.stringLiteral(n)),
+  );
+
+  const embedderDecl = t.variableDeclaration('const', [
+    t.variableDeclarator(
+      t.identifier('embedder'),
+      t.callExpression(t.identifier('resolveEmbedder'), []),
+    ),
+  ]);
+
+  const noEmbedderCheck = t.ifStatement(
+    t.unaryExpression('!', t.identifier('embedder')),
+    t.blockStatement([
+      t.expressionStatement(
+        t.callExpression(
+          t.memberExpression(t.identifier('console'), t.identifier('error')),
+          [
+            t.stringLiteral(
+              '--auto-embed requires an embedder. Set EMBEDDER_PROVIDER=ollama (and optionally EMBEDDER_MODEL, EMBEDDER_BASE_URL).',
+            ),
+          ],
+        ),
+      ),
+      t.expressionStatement(
+        t.callExpression(
+          t.memberExpression(t.identifier('process'), t.identifier('exit')),
+          [t.numericLiteral(1)],
+        ),
+      ),
+    ]),
+  );
+
+  const autoEmbedCall = t.awaitExpression(
+    t.callExpression(t.identifier('autoEmbedInput'), [
+      t.identifier('cleanedData'),
+      fieldNamesArray,
+      t.identifier('embedder'),
+    ]),
+  );
+
+  return t.ifStatement(
+    t.memberExpression(
+      t.identifier('argv'),
+      t.stringLiteral('auto-embed'),
+      true,
+    ),
+    t.blockStatement([
+      embedderDecl,
+      noEmbedderCheck,
+      t.expressionStatement(autoEmbedCall),
+    ]),
+  );
+}
+
 function buildListHandler(table: Table, vectorFieldNames: string[], targetName?: string, typeRegistry?: TypeRegistry): t.FunctionDeclaration {
   const { singularName } = getTableNames(table);
   const defaultSelectObj = buildSelectObject(table, typeRegistry);
@@ -904,6 +977,7 @@ export function getFieldsWithDefaults(
 function buildMutationHandler(
   table: Table,
   operation: 'create' | 'update' | 'delete',
+  vectorFieldNames: string[],
   targetName?: string,
   typeRegistry?: TypeRegistry,
   ormTypes?: { createInputTypeName: string; innerFieldName: string; patchTypeName: string },
@@ -1124,6 +1198,12 @@ function buildMutationHandler(
         ),
       ]),
     );
+
+    // Inject --auto-embed block for create/update when table has vector fields.
+    // Converts text strings in vector fields to embeddings before the ORM call.
+    if (vectorFieldNames.length > 0) {
+      tryBody.push(buildAutoEmbedInputBlock(vectorFieldNames));
+    }
   }
 
   tryBody.push(
@@ -1257,7 +1337,7 @@ export function generateTableCommand(table: Table, options?: TableCommandOptions
   if (hasEmbeddings) {
     const embedderPath = options?.targetName ? '../../embedder' : '../embedder';
     statements.push(
-      createImportDeclaration(embedderPath, ['resolveEmbedder', 'autoEmbedWhere']),
+      createImportDeclaration(embedderPath, ['resolveEmbedder', 'autoEmbedWhere', 'autoEmbedInput']),
     );
   }
 
@@ -1280,6 +1360,13 @@ export function generateTableCommand(table: Table, options?: TableCommandOptions
   if (hasGet) usageLines.push(`  get                   Get a ${singularName} by ID`);
   usageLines.push(`  create                Create a new ${singularName}`);
   if (hasUpdate) usageLines.push(`  update                Update an existing ${singularName}`);
+  if (hasEmbeddings) {
+    usageLines.push(
+      '',
+      'Create/Update Options:',
+      '  --auto-embed          Convert text values in vector fields to embeddings before saving',
+    );
+  }
   if (hasDelete) usageLines.push(`  delete                Delete a ${singularName}`);
   usageLines.push(
     '',
@@ -1496,9 +1583,9 @@ export function generateTableCommand(table: Table, options?: TableCommandOptions
   statements.push(buildFindFirstHandler(table, tn, options?.typeRegistry));
   if (hasSearchFields) statements.push(buildSearchHandler(table, specialGroups, vectorFieldNames, tn, options?.typeRegistry));
   if (hasGet) statements.push(buildGetHandler(table, tn, options?.typeRegistry));
-  statements.push(buildMutationHandler(table, 'create', tn, options?.typeRegistry, ormTypes));
-  if (hasUpdate) statements.push(buildMutationHandler(table, 'update', tn, options?.typeRegistry, ormTypes));
-  if (hasDelete) statements.push(buildMutationHandler(table, 'delete', tn, options?.typeRegistry, ormTypes));
+  statements.push(buildMutationHandler(table, 'create', vectorFieldNames, tn, options?.typeRegistry, ormTypes));
+  if (hasUpdate) statements.push(buildMutationHandler(table, 'update', vectorFieldNames, tn, options?.typeRegistry, ormTypes));
+  if (hasDelete) statements.push(buildMutationHandler(table, 'delete', vectorFieldNames, tn, options?.typeRegistry, ormTypes));
 
   const header = getGeneratedFileHeader(`CLI commands for ${table.name}`);
   const code = generateCode(statements);

graphql/codegen/src/core/codegen/docs-utils.ts

@@ -376,6 +376,14 @@ export function buildSearchExamples(
             `EMBEDDER_PROVIDER=ollama ${toolName} ${cmd} list --where.${field.name}.vector "semantic query" --auto-embed --fields title,${field.name}VectorDistance`,
           ],
         });
+        examples.push({
+          description: `Create/update with auto-embedded \`${field.name}\` via --auto-embed`,
+          code: [
+            `# --auto-embed on create/update converts text strings in vector fields to embeddings before saving`,
+            `EMBEDDER_PROVIDER=ollama ${toolName} ${cmd} create --${field.name} "text to embed" --auto-embed`,
+            `EMBEDDER_PROVIDER=ollama ${toolName} ${cmd} update --${field.name} "new text to embed" --auto-embed`,
+          ],
+        });
       }
 
       // searchScore — composite blend field, useful for ordering

graphql/codegen/src/core/codegen/templates/embedder.ts

@@ -1,10 +1,10 @@
 /**
- * CLI Embedder — pluggable text-to-vector embedding for CLI search commands
+ * CLI Embedder — pluggable text-to-vector embedding for CLI commands
  *
  * This is RUNTIME code that gets copied to generated output.
  * Provides a pluggable system for registering embedding functions so that
- * CLI search and list commands can convert text queries into vector arrays
- * for pgvector similarity search.
+ * CLI commands can convert text queries into vector arrays for pgvector
+ * similarity search (list/search) and create/update vector fields inline.
  *
  * Configuration via appstash config or environment variables:
  *   embedder.provider  = 'ollama' | 'custom'
@@ -137,3 +137,39 @@ export async function autoEmbedWhere(
   }
   return where;
 }
+
+/**
+ * Auto-embed text values in mutation input data (create/update).
+ *
+ * When --auto-embed is passed on create or update, any vector field in
+ * the input data that contains a text string will be converted to an
+ * embedding vector using the configured embedder.
+ *
+ * Usage:
+ *   csdk article create --input.embedding "Machine learning concepts" --auto-embed
+ *   csdk article update --id xxx --input.embedding "Updated description" --auto-embed
+ *
+ * This is a CLI-only convenience — in production, database triggers or
+ * a job queue should handle embedding generation.
+ *
+ * @param data - The mutation input data object (mutated in place)
+ * @param vectorFieldNames - Names of vector embedding fields (e.g. ['embedding'])
+ * @param embedder - The resolved embedder function
+ * @returns The modified data object with text values replaced by vectors
+ */
+export async function autoEmbedInput(
+  data: Record<string, unknown>,
+  vectorFieldNames: string[],
+  embedder: EmbedderFunction,
+): Promise<Record<string, unknown>> {
+  for (const fieldName of vectorFieldNames) {
+    const fieldValue = data[fieldName];
+    if (typeof fieldValue === 'string') {
+      // Text string → embed to vector array
+      const embedding = await embedder(fieldValue);
+      data[fieldName] = embedding;
+    }
+    // If it's already an array (pre-computed vector), leave it as-is
+  }
+  return data;
+}