Sign in to your Google Cloud account. If you're new to
Google Cloud,
create an account to evaluate how our products perform in
real-world scenarios. New customers also get $300 in free credits to
run, test, and deploy workloads.
In the Google Cloud console, on the project selector page,
select or create a Google Cloud project.
[[["わかりやすい","easyToUnderstand","thumb-up"],["問題の解決に役立った","solvedMyProblem","thumb-up"],["その他","otherUp","thumb-up"]],[["わかりにくい","hardToUnderstand","thumb-down"],["情報またはサンプルコードが不正確","incorrectInformationOrSampleCode","thumb-down"],["必要な情報 / サンプルがない","missingTheInformationSamplesINeed","thumb-down"],["翻訳に関する問題","translationIssue","thumb-down"],["その他","otherDown","thumb-down"]],["最終更新日 2025-09-04 UTC。"],[[["\u003cp\u003eRemote functions allow for the analysis of unstructured data within object tables by interacting with services like pre-trained AI models, open-source libraries, and custom models.\u003c/p\u003e\n"],["\u003cp\u003eAnalyzing object table data with remote functions requires passing signed URLs, which are generated for the objects and provide the necessary access authorization to the remote function.\u003c/p\u003e\n"],["\u003cp\u003eSpecific permissions are needed to create connections, create, and invoke remote functions, as well as to access the data in the object table itself.\u003c/p\u003e\n"],["\u003cp\u003eWhen creating a remote function for object table analysis, it's recommended to set the \u003ccode\u003emax_batching_rows\u003c/code\u003e option to 1 to prevent Cloud Run function timeouts and enhance processing parallelism.\u003c/p\u003e\n"],["\u003cp\u003eTo use a remote function on object table data, it must be referenced in the \u003ccode\u003eSELECT\u003c/code\u003e list and use \u003ccode\u003eEXTERNAL_OBJECT_TRANSFORM\u003c/code\u003e in the \u003ccode\u003eFROM\u003c/code\u003e clause to generate signed URLs, with the option to filter using \u003ccode\u003eWHERE\u003c/code\u003e.\u003c/p\u003e\n"]]],[],null,["Analyze object tables by using remote functions\n\nThis document describes how to analyze unstructured data in\n[object tables](/bigquery/docs/object-table-introduction) by using\n[remote functions](/bigquery/docs/remote-functions).\n\nOverview\n\nYou can analyze the unstructured data represented by an object table by using\na remote function. A remote function lets you call a function running on\nCloud Run functions or Cloud Run, which you can program to access\nresources such as:\n\n- Google's pre-trained AI models, including Cloud Vision API and Document AI.\n- Open source libraries such as [Apache Tika](https://tika.apache.org/).\n- Your own custom models.\n\nTo analyze object table data by using a remote function, you must\ngenerate and pass in\n[signed URLs](/bigquery/docs/object-table-introduction#signed_urls) for the\nobjects in the object table when you call the remote function. These signed\nURLs are what grant the remote function access to the objects.\n\nRequired permissions\n\n- To create the connection resource used by the remote function, you need the following permissions:\n\n - `bigquery.connections.create`\n - `bigquery.connections.get`\n - `bigquery.connections.list`\n - `bigquery.connections.update`\n - `bigquery.connections.use`\n - `bigquery.connections.delete`\n- To create a remote function, you need the permissions associated with the\n [Cloud Functions Developer](/functions/docs/reference/iam/roles#cloudfunctions.developer)\n or [Cloud Run Developer](/iam/docs/understanding-roles#run.developer) roles.\n\n- To invoke a remote function, you need the permissions described in\n [Remote functions](/bigquery/docs/remote-functions#grant_permission_on_function).\n\n- To analyze an object table with a remote function, you need the\n `bigquery.tables.getData` permission on the object table.\n\nBefore you begin\n\n- Sign in to your Google Cloud account. If you're new to Google Cloud, [create an account](https://console.cloud.google.com/freetrial) to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.\n- In the Google Cloud console, on the project selector page,\n select or create a Google Cloud project.\n\n | **Note**: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.\n\n [Go to project selector](https://console.cloud.google.com/projectselector2/home/dashboard)\n-\n [Verify that billing is enabled for your Google Cloud project](/billing/docs/how-to/verify-billing-enabled#confirm_billing_is_enabled_on_a_project).\n\n-\n\n\n Enable the BigQuery, BigQuery Connection API, Cloud Run functions APIs.\n\n\n [Enable the APIs](https://console.cloud.google.com/flows/enableapi?apiid=bigquery.googleapis.com,bigqueryconnection.googleapis.com,cloudfunctions.googleapis.com)\n\n- In the Google Cloud console, on the project selector page,\n select or create a Google Cloud project.\n\n | **Note**: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.\n\n [Go to project selector](https://console.cloud.google.com/projectselector2/home/dashboard)\n-\n [Verify that billing is enabled for your Google Cloud project](/billing/docs/how-to/verify-billing-enabled#confirm_billing_is_enabled_on_a_project).\n\n-\n\n\n Enable the BigQuery, BigQuery Connection API, Cloud Run functions APIs.\n\n\n [Enable the APIs](https://console.cloud.google.com/flows/enableapi?apiid=bigquery.googleapis.com,bigqueryconnection.googleapis.com,cloudfunctions.googleapis.com)\n\n1. Ensure that your BigQuery administrator has [created a connection](/bigquery/docs/create-cloud-resource-connection#create-cloud-resource-connection) and [set up\n access to Cloud Storage](/bigquery/docs/create-cloud-resource-connection#access-storage).\n\n\u003cbr /\u003e\n\nCreate a remote function\n\nFor general instructions on creating a remote function, see\n[Working with remote functions](/bigquery/docs/remote-functions).\n\nWhen you create a remote function to analyze object table data, you must\npass in [signed URLS](/bigquery/docs/object-table-introduction#signed_urls)\nthat have been generated for the objects in the object table. You can do this\nby using an input parameter with a `STRING` data type. The signed URLS are\nmade available to the remote function as input data in the\n[`calls` field of the HTTP `POST` request](/bigquery/docs/remote-functions#input_format).\nAn example of a request is: \n\n {\n // Other fields omitted.\n \"calls\": [\n [\"https://storage.googleapis.com/mybucket/1.pdf?X-Goog-SignedHeaders=abcd\"],\n [\"https://storage.googleapis.com/mybucket/2.pdf?X-Goog-SignedHeaders=wxyz\"]\n ]\n }\n\nYou can read an object in your remote function by using a method that makes\nan HTTP `GET` request to the signed URL. The remote function can access the\nobject because the signed URL contains authentication information in its\nquery string.\n\nWhen you specify the\n[`CREATE FUNCTION` statement](/bigquery/docs/reference/standard-sql/data-definition-language#create_function_statement)\nfor the remote function, we recommend that you set the `max_batching_rows`\noption to 1 in order to\n[avoid Cloud Run functions timeout](/functions/docs/concepts/exec#timeout)\nand increase processing parallelism.\n\nExample\n\nThe following Cloud Run functions Python code example reads storage\nobjects and returns their content length to BigQuery: \n\n import functions_framework\n import json\n import urllib.request\n\n @functions_framework.http\n def object_length(request):\n calls = request.get_json()['calls']\n replies = []\n for call in calls:\n object_content = urllib.request.urlopen(call[0]).read()\n replies.append(len(object_content))\n return json.dumps({'replies': replies})\n\nDeployed, this function would have an endpoint similar to\n`https://us-central1-myproject.cloudfunctions.net/object_length`.\n\nThe following example shows how to create a BigQuery remote\nfunction based on this Cloud Run functions function: \n\n```googlesql\nCREATE FUNCTION mydataset.object_length(signed_url STRING) RETURNS INT64\nREMOTE WITH CONNECTION `us.myconnection`\nOPTIONS(\n endpoint = \"https://us-central1-myproject.cloudfunctions.net/object_length\",\n max_batching_rows = 1\n);\n```\n\nFor step-by-step guidance, see\n[Tutorial: Analyze an object table with a remote function](/bigquery/docs/remote-function-tutorial).\n\nCall a remote function\n\nTo call a remote function on object table data, reference the remote\nfunction in the\n[`select_list`](/bigquery/docs/reference/standard-sql/query-syntax#select_list)\nof the query, and then call the\n[`EXTERNAL_OBJECT_TRANSFORM` function](/bigquery/docs/reference/standard-sql/table-functions-built-in#external_object_transform)\nin the\n[`FROM` clause](/bigquery/docs/reference/standard-sql/query-syntax#from_clause)\nto generate the signed URLs for the objects.\n| **Note:** When using one of the [AI APIs](/products/ai), be aware of the relevant quotas for the API you are targeting. Use a `LIMIT` clause to limit the results returned if necessary to stay within quota.\n\nThe following example shows typical statement syntax: \n\n```googlesql\nSELECT uri, function_name(signed_url) AS function_output\nFROM EXTERNAL_OBJECT_TRANSFORM(TABLE my_dataset.object_table, [\"SIGNED_URL\"])\nLIMIT 10000;\n```\n\nThe following example shows how to process only a subset of the object table\ncontents with a remote function: \n\n```googlesql\nSELECT uri, function_name(signed_url) AS function_output\nFROM EXTERNAL_OBJECT_TRANSFORM(TABLE my_dataset.object_table, [\"SIGNED_URL\"])\nWHERE content_type = \"application/pdf\";\n```\n\nWhat's next\n\nLearn how to [run inference on image object tables](/bigquery/docs/object-table-inference)."]]