Deploy the Search Service
Go to Deploy a Custom Template in the Azure Portal
Click "Build your own template in the editor”
Click "Load file" and upload the json file here using these parameters: Customer Name: the name used to deploy your portal (can be found in the update page of your AI Portal) Storage Name: the name you’d like to give your default storage for rag
Click Review + Create and then Create
Get key for Embeddings model
Head to Microsoft Foundry in the Azure Portal
In Use with Foundry → Foundry click on the foundry project created with the name embeddings (named {name}-uf-aiportal-embeddings)
Click on the button that says Go To Foundry Portal
Once in Foundry portal, Under My Assets → Models + endpoints find the text-embedding-3-large deployment and copy the key. You’ll use it in one of the following steps. Also copy the endpoint field (should be something like:
https://{NAME}-aiportal-embeddings.cognitiveservices.azure.com/
Add Data Source
Back in Microsoft Foundry in the Azure Portal, go to Use with Foundry → AI Search. Click on the search named {name}-rag-aiportal
In Search Management → Data Sources click Add a Data Source
Your data source should say: Name: Any identifying name (here I chose rag-datasource) Subscription: Subscription of your AI Portal Storage Account: {name}aiportal Blob Container: The name you gave the storage account in step 1 Click Create
Once created, click into the datasource, click edit, and turn on Deletion Tracking:
Add Index
In Search Management → Indexes click Add Index (JSON)
Paste this JSON into the index and click create (Feel free to update the name docsindex to better match your storage)
{
"name": "docsindex",
"purviewEnabled": false,
"fields": [
{
"name": "id",
"type": "Edm.String",
"searchable": true,
"filterable": true,
"retrievable": true,
"stored": true,
"sortable": true,
"facetable": true,
"key": true,
"analyzer": "keyword",
"synonymMaps": []
},
{
"name": "content",
"type": "Edm.String",
"searchable": true,
"filterable": false,
"retrievable": true,
"stored": true,
"sortable": false,
"facetable": false,
"key": false,
"synonymMaps": []
},
{
"name": "allowedUsers",
"type": "Collection(Edm.String)",
"searchable": true,
"filterable": true,
"retrievable": true,
"stored": true,
"sortable": false,
"facetable": true,
"key": false,
"synonymMaps": []
},
{
"name": "sourceDoc",
"type": "Edm.String",
"searchable": true,
"filterable": true,
"retrievable": true,
"stored": true,
"sortable": true,
"facetable": true,
"key": false,
"synonymMaps": []
},
{
"name": "contentVector",
"type": "Collection(Edm.Single)",
"searchable": true,
"filterable": false,
"retrievable": false,
"stored": true,
"sortable": false,
"facetable": false,
"key": false,
"dimensions": 3072,
"vectorSearchProfile": "default-profile",
"synonymMaps": []
},
{
"name": "parentId",
"type": "Edm.String",
"searchable": false,
"filterable": true,
"retrievable": true,
"stored": true,
"sortable": false,
"facetable": false,
"key": false,
"synonymMaps": []
}
],
"scoringProfiles": [],
"suggesters": [],
"analyzers": [],
"normalizers": [],
"tokenizers": [],
"tokenFilters": [],
"charFilters": [],
"similarity": {
"@odata.type": "#Microsoft.Azure.Search.BM25Similarity"
},
"semantic": {
"configurations": [
{
"name": "default",
"flightingOptIn": false,
"rankingOrder": "BoostedRerankerScore",
"prioritizedFields": {
"titleField": {
"fieldName": "id"
},
"prioritizedContentFields": [
{
"fieldName": "content"
}
],
"prioritizedKeywordsFields": []
}
}
]
},
"vectorSearch": {
"algorithms": [
{
"name": "default-hnsw",
"kind": "hnsw",
"hnswParameters": {
"metric": "cosine",
"m": 4,
"efConstruction": 400,
"efSearch": 500
}
}
],
"profiles": [
{
"name": "default-profile",
"algorithm": "default-hnsw"
}
],
"vectorizers": [],
"compressions": []
}
}
Create a Skillset
In Search Management → Skillsets Click Add a Skillset . Paste the JSON below, Updating these text fields:
INSERT_NAME_HERE: Feel free to put any name that defines your skillset (I’d suggest index-skillset. So mine would be docsindex-skillset
INSERT_EMBEDDINGS_ENDPOINT_HERE: The endpoint copied from the embeddings provider in step 2.
INSERT_YOUR_API_KEY_HERE: The API key copied from the embeddings provider in step 2
INSERT_INDEX_NAME_HERE: The index name from step 4.
Click Create
{
"name": "INSERT_NAME_HERE",
"description": "Splits document text into token-based chunks and generates embeddings.",
"skills": [
{
"@odata.type": "#Microsoft.Skills.Util.DocumentExtractionSkill",
"name": "extractText",
"description": "Extract text from PDFs / Office docs / etc.",
"context": "/document",
"parsingMode": "default",
"dataToExtract": "contentAndMetadata",
"inputs": [
{
"name": "file_data",
"source": "/document/file_data",
"inputs": []
}
],
"outputs": [
{
"name": "content",
"targetName": "content"
}
],
"configuration": {}
},
{
"@odata.type": "#Microsoft.Skills.Text.SplitSkill",
"name": "SplitSkill",
"description": "Split document text into pages based on Azure OpenAI tokens.",
"context": "/document",
"defaultLanguageCode": "en",
"textSplitMode": "pages",
"maximumPageLength": 1400,
"pageOverlapLength": 150,
"maximumPagesToTake": 0,
"unit": "azureOpenAITokens",
"inputs": [
{
"name": "text",
"source": "/document/content",
"inputs": []
},
{
"name": "languageCode",
"source": "/document/language",
"inputs": []
}
],
"outputs": [
{
"name": "textItems",
"targetName": "pages"
}
],
"azureOpenAITokenizerParameters": {
"encoderModelName": "cl100k_base",
"allowedSpecialTokens": [
"[START]",
"[END]"
]
}
},
{
"@odata.type": "#Microsoft.Skills.Text.AzureOpenAIEmbeddingSkill",
"name": "EmbeddingSkill",
"description": "Generate embeddings for each page chunk.",
"context": "/document/pages/*",
"resourceUri": "INSERT_EMBEDDINGS_ENDPOINT_HERE",
"apiKey": "INSERT_YOUR_API_KEY_HERE",
"deploymentId": "text-embedding-3-large",
"dimensions": 3072,
"modelName": "text-embedding-3-large",
"inputs": [
{
"name": "text",
"source": "/document/pages/*",
"inputs": []
}
],
"outputs": [
{
"name": "embedding",
"targetName": "contentVector"
}
]
}
],
"indexProjections": {
"selectors": [
{
"targetIndexName": "INSERT_INDEX_NAME_HERE",
"parentKeyFieldName": "parentId",
"sourceContext": "/document/pages/*",
"mappings": [
{
"name": "content",
"source": "/document/pages/*",
"inputs": []
},
{
"name": "contentVector",
"source": "/document/pages/*/contentVector",
"inputs": []
},
{
"name": "sourceDoc",
"source": "/document/metadata_storage_name",
"inputs": []
}
]
}
],
"parameters": {
"projectionMode": "skipIndexingParentDocuments"
}
}
}
Add Indexer
Add Indexer (JSON) Copy and paste the JSON below. Fill in these fields:
INSERT_NAME_HERE: Insert a name for your indexer (I would put the name of the datasource-indexer. So here it would be docsindex-indexer
INSERT_DATASOURCE_NAME_HERE: Datasource name from Step 4
INSERT_SKILLSET_NAME_HERE: Skillset name from Step 5
INSERT_INDEX_NAME_HERE: Index name from Step 3{
"name": "INSERT_NAME_HERE",
"description": null,
"dataSourceName": "INSERT_DATASOURCE_NAME_HERE",
"skillsetName": "INSERT_SKILLSET_NAME_HERE",
"targetIndexName": "INSERT_INDEX_NAME_HERE",
"disabled": null,
"schedule": {
"interval": "PT5M",
"startTime": "2025-12-02T18:51:24.38Z"
},
"parameters": {
"batchSize": null,
"maxFailedItems": null,
"maxFailedItemsPerBatch": null,
"configuration": {
"dataToExtract": "storageMetadata",
"parsingMode": "default",
"allowSkillsetToReadFileData": true,
"failOnUnsupportedContentType": false,
"failOnUnprocessableDocument": false
}
},
"fieldMappings": [
{
"sourceFieldName": "metadata_storage_path",
"targetFieldName": "id",
"mappingFunction": {
"name": "base64Encode",
"parameters": null
}
},
{
"sourceFieldName": "metadata_storage_name",
"targetFieldName": "sourceDoc",
"mappingFunction": null
}
],
"outputFieldMappings": [
{
"sourceFieldName": "/document/pages/*",
"targetFieldName": "content",
"mappingFunction": null
},
{
"sourceFieldName": "/document/pages/*/contentVector",
"targetFieldName": "contentVector",
"mappingFunction": null
},
{
"sourceFieldName": "/document/sourceDoc",
"targetFieldName": "sourceDoc",
"mappingFunction": null
}
],
"cache": null,
"encryptionKey": null
}
In Search Service → Overview, Copy down the Search Endpoint listed under URL
In AIPortal, go to the Admin Panel → RAG Config and Add a New Config:
Name: The user friendly name for your RAG in AI Portal
Document Return Count: Number of Returned Documents from the RAG Model
Embedding Deployment Endpoint: The Endpoint for the Embeddings Deployment (should be something like https://{NAME}-aiportal-embeddings.cognitiveservices.azure.com/ ).
Embedding Deployment: text-embedding-3-large
Search Endpoint: The Endpoint copied from Step 6
Search Index: What you named your index in Step 3
In Your AI Portal, In Admin → Provider Config, attach the RAG Search to an existing provider, or create a new provider (e.g. copy the GPT-5.4 provider and under RAG Search select your new search)
Your RAG Provider is now Ready to use!
Notes:
