Migrating from @azure/ai-form-recognizer
Version 4.0.0 to @azure-rest/ai-document-intelligence
Version 1.0.0-beta.1
In this first preview of @azure-rest/ai-document-intelligence
Rest Level Client Library, this package introduces a new design for the Azure AI Document Intelligence service (formerly known as Form Recognizer), targeting service API version "2023-10-31-preview"
.
To leverage features of the newest Azure AI Document Intelligence service API (version "2023-10-31-preview" and newer), the new SDK is required, and application code must be changed to use the new client. Similarly, the new major version 4 of the client library cannot be used to communicate with versions 2.1 of the service API. To summarize:
- Version 3 of the
@azure/ai-form-recognizer
package only supports Form Recognizer service API version 2.1, and will not receive support for newer (date-based) versions of Form Recognizer. - Version 4 of the
@azure/ai-form-recognizer
package supports service API version "2022-08-31" of Form Recognizer.
Note: Form Recognizer has been rebranded to Document Intelligence.
- Version 1 of the
@azure-rest/ai-document-intelligence
package will receive support for newer (date-based) versions of Azure AI Document Intelligence.
This document provides instructions for updating your application code to the new @azure-rest/ai-document-intelligence
SDK client library. In this document, the examples provided use TypeScript to provide type information, but all runtime behavior changes naturally apply to plain JavaScript as well.
To avoid migrating an application all at once, @azure-rest/ai-document-intelligence
Version 1.0.0-beta.1 may be installed alongside @azure/ai-form-recognizer
Version 4.0.0. For instance, add the following to the dependencies
field of package.json
:
{
...,
"dependencies": {
...,
"@azure/ai-form-recognizer": "^4.0.0",
"@azure-rest/ai-document-intelligence": "1.0.0-beta.1"
}
}
Then, the two packages may be used side-by-side, and an application may be migrated partially or over time:
import { DocumentAnalysisClient } from "@azure/ai-form-recognizer";
import DocumentIntelligence from "@azure-rest/ai-document-intelligence";
In the new @azure-rest/ai-document-intelligence
package, several types and method signatures have been introduced which would feel newer owing to the redesign as a Rest Level Client package.
@azure/ai-form-recognizer
package offeredDocumentAnalysisClient
andDocumentModelAdministrationClient
classes.- The new
@azure-rest/ai-document-intelligence
package offers acreateClient
method that allows creating a an instance ofDocumentIntelligenceClient
, which would be the equivalent of the two classes offered through@azure/ai-form-recognizer
package.
The previous DocumentAnalysisClient
class offered
beginAnalyzeDocument(modelId, document: FormRecognizerRequestBody)
beginAnalyzeDocumentFromUrl(modelId, documentUrl)
beginClassifyDocument(classifierId, document: FormRecognizerRequestBody)
beginClassifyDocumentFromUrl(classifierId, documentUrl)
methods to analyze/classify documents.
Equivalently, the new DocumentIntelligence
offers the following instead through the routes:
path("/documentModels/{modelId}:analyze", "<model-id>").post({
contentType: "application/json",
body: { urlSource: "..." }, // or { base64Source: "..." }
});
path("/documentClassifiers/{classifierId}:analyze", "<classifier-id>").post({
contentType: "application/json",
body: { urlSource: "..." }, // or { base64Source: "..." }
});
As in the previous "@azure/ai-form-recognizer" v4.0.0 SDK, to provide a publicly-accessible URL as an input to an analysis operation, use the beginAnalyzeDocumentFromUrl
method, which treats the data passed to it as a file's contents. In the new @azure-rest/ai-document-intelligence
package, the equivalent is to use the path
method to call the analyze
route and pass the URL as body (shown below).
Previous ("@azure/ai-form-recognizer" v4.0.0):
import { AzureKeyCredential, DocumentAnalysisClient } from "@azure/ai-form-recognizer";
const client = new DocumentAnalysisClient(process.env.FORM_RECOGNIZER_ENDPOINT, new AzureKeyCredential(process.env.FORM_RECOGNIZER_API_KEY); );
const url = "https://raw.githubusercontent.com/Azure/azure-sdk-for-js/main/sdk/formrecognizer/ai-form-recognizer/assets/identityDocument/license.jpg";
const poller = await client.beginAnalyzeDocumentFromUrl("<model ID>", url);
Current ("@azure-rest/ai-document-intelligence" v1.0.0-beta.1):
import DocumentIntelligence from "@azure-rest/ai-document-intelligence";
const client = DocumentIntelligence(process.env["DOCUMENT_INTELLIGENCE_ENDPOINT"], {
key: process.env["DOCUMENT_INTELLIGENCE_API_KEY"],
});
const url =
"https://raw.githubusercontent.com/Azure/azure-sdk-for-js/main/sdk/formrecognizer/ai-form-recognizer/assets/identityDocument/license.jpg";
const initialResponse = await client.path("/documentModels/{modelId}:analyze", "<model ID>").post({
contentType: "application/json",
body: { urlSource: url },
});
if (isUnexpected(initialResponse)) {
throw initialResponse.body.error;
}
const poller = await getLongRunningPoller(client, initialResponse);
As in the previous "@azure/ai-form-recognizer" v4.0.0 SDK, to provide a document input to an analysis operation, use the beginAnalyzeDocument
method, which treats the data passed to it as a file's contents.
In the new @azure-rest/ai-document-intelligence
package, the equivalent is to use the path
method to call the analyze
route and pass the base64-encoded contents (string) as body (shown below).
Note: Subject to change in the future previews of this package for better.
Previous ("@azure/ai-form-recognizer" v4.0.0):
const { DocumentAnalysisClient, AzureKeyCredential } = require("@azure/ai-form-recognizer");
const fs = require("fs");
const client = new DocumentAnalysisClient(endpoint, new AzureKeyCredential(apiKey));
const path = "<path to a document>";
const readStream = fs.createReadStream(path);
const poller = await client.beginAnalyzeDocument(modelId, readStream);
Current ("@azure-rest/ai-document-intelligence" v1.0.0-beta.1):
import DocumentIntelligence from "@azure-rest/ai-document-intelligence";
const client = DocumentIntelligence(endpoint, { key: apiKey });
const filePath = "<path to a document>";
const base64Source = fs.readFileSync(filePath, { encoding: "base64" });
const initialResponse = await client.path("/documentModels/{modelId}:analyze", "<model id>").post({
contentType: "application/json",
body: { base64Source },
});
if (isUnexpected(initialResponse)) {
throw initialResponse.body.error;
}
const poller = getLongRunningPoller(client, initialResponse, { ...testPollingOptions });
const endpoint = "<cognitive services endpoint>";
const apiKey = "<api key>";
const path = "<path to a document>"; // pdf/jpeg/png/tiff formats
const readStream = fs.createReadStream(path);
const client = new DocumentAnalysisClient(endpoint, new AzureKeyCredential(apiKey));
const poller = await client.beginAnalyzeDocument("prebuilt-layout", readStream);
const { pages, tables } = await poller.pollUntilDone();
const client = DocumentIntelligence(
process.env["DOCUMENT_INTELLIGENCE_ENDPOINT"] || "<cognitive services endpoint>",
{ key: process.env["DOCUMENT_INTELLIGENCE_API_KEY"] || "<api key>" }
);
const base64Source = fs.readFileSync(filePath, { encoding: "base64" });
const initialResponse = await client
.path("/documentModels/{modelId}:analyze", "prebuilt-layout")
.post({
contentType: "application/json",
body: { base64Source },
});
if (isUnexpected(initialResponse)) {
throw initialResponse.body.error;
}
const poller = await getLongRunningPoller(client, initialResponse);
const analyzeResult = ((await poller.pollUntilDone().body) as AnalyzeResultOperationOutput)
.analyzeResult;
const pages = analyzeResult?.pages;
const tables = analyzeResult?.tables;
The DocumentModelAdministrationClient
class, was used for all model management operations (creating, reading, listing, and deleting models). The new DocumentIntelligence
offers the following instead through the routes:
beginBuildDocumentClassifier(classifierId, [docTypeSources])
beginBuildDocumentModel(modelId, containerUrl, buildMode: "template")
beginBuildDocumentModel(modelId, contentSource, buildMode: "template")
beginComposeDocumentModel(modelId, [componentModelIds])
beginCopyModelTo(sourceModelId, authorization)
deleteDocumentClassifier(classifierId)
deleteDocumentModel(modelId)
getCopyAuthorization(destinationModelId)
getDocumentClassifier(classifierId)
getDocumentModel(modelId)
getOperation(operationId)
getResourceDetails()
listDocumentClassifiers()
listDocumentModels()
listOperations()
Equivalently, the new DocumentIntelligence
offers the following instead through the routes:
path("/documentClassifiers:build").post({
body: {
classifierId: "<classifier-id>",
docTypes: {
foo: {
azureBlobSource: {
containerUrl: "<container-url>",
},
},
},
},
});
path("/documentModels:build").post({
body: {
buildMode: "template",
modelId: modelName,
azureBlobSource: {
containerUrl: "<container-url>",
},
},
});
path("/documentModels:compose").post({
body: {
componentModels: componentModelIds,
modelId,
},
});
path("/documentClassifiers/{classifierId}", classifierId).delete();
path("/documentModels/{modelId}", modelId).delete();
path("/documentModels/{modelId}:copyTo", sourceModel.modelId).post({
body: targetAuth.body,
});
path("/documentClassifiers/{classifierId}", classifierId).get();
path("/documentModels/{modelId}", model.modelId).get();
path("/operations/{operationId}", "<operationId>").get();
path("/info").get();
path("/documentClassifiers").get();
path("/documentModels").get();
path("/operations").get();
All the methods presented above do support options bags to configure the settings
Supports output with Markdown content format along with the default plain text. For now, this is only supported for "prebuilt-layout". Markdown content format is deemed a more friendly format for LLM consumption in a chat or automation use scenario.
Service follows the GFM spec (GitHub Flavored Markdown) for the Markdown format. Also introduces a new contentFormat property with value "text" or "markdown" to indicate the result content format.
import DocumentIntelligence from "@azure-rest/ai-document-intelligence";
const client = DocumentIntelligence(process.env["DOCUMENT_INTELLIGENCE_ENDPOINT"], {
key: process.env["DOCUMENT_INTELLIGENCE_API_KEY"],
});
const initialResponse = await client
.path("/documentModels/{modelId}:analyze", "prebuilt-layout")
.post({
contentType: "application/json",
body: {
urlSource:
"https://raw.githubusercontent.com/Azure/azure-sdk-for-js/6704eff082aaaf2d97c1371a28461f512f8d748a/sdk/formrecognizer/ai-form-recognizer/assets/forms/Invoice_1.pdf",
},
queryParameters: { outputContentFormat: "markdown" }, // <-- new query parameter
});
When this feature flag is specified, the service will further extract the values of the fields specified via the queryFields query parameter to supplement any existing fields defined by the model as fallback.
await client.path("/documentModels/{modelId}:analyze", "prebuilt-layout").post({
contentType: "application/json",
body: { urlSource: "..." },
queryParameters: {
features: ["queryFields"],
queryFields: ["NumberOfGuests", "StoreNumber"],
}, // <-- new query parameter
});
In the previous API versions supported by the older @azure/ai-form-recognizer
library, document splitting and classification operation ("/documentClassifiers/{classifierId}:analyze"
) always tried to split the input file into multiple documents.
To enable a wider set of scenarios, service introduces a "split" query parameter with the new "2023-10-31-preview" service version. The following values are supported:
-
split: "auto"
Let service determine where to split.
-
split: "none"
The entire file is treated as a single document. No splitting is performed.
-
split: "perPage"
Each page is treated as a separate document. Each empty page is kept as its own document.
Checkout our samples for more examples.