Knowledge Base Indexer (Azure AI Search)

Overview

This solution keeps an Azure AI Search index continuously in sync with a SharePoint knowledge library. It ships as two cloud flows + one Dataverse sync ledger in a single solution: 1. Knowledge Base Indexer (event-driven) — when a document is created or modified in the library, it extracts the text, builds a search document, and upserts it into the index (IndexDocuments, @search.action = mergeOrUpload), recording the document in the ledger. 2. Knowledge Base Index Reconciliation Sweep (scheduled) — every 6 hours it compares the library against the ledger and deletes orphaned documents from the index (DeleteDocuments) when their source file has been removed.

Why two flows? SharePoint has no reliable "file deleted" trigger, so create/update is handled in real time by the event flow while deletion is handled by a periodic reconciliation sweep. Together they give an always-current index that can power an app, bot, or RAG pipeline. Both flows are connector-first and ship Off.

Implementation note: the correct connected connector is shared_azureaisearch (Azure AI Search). The index-write operations (IndexDocuments / DeleteDocuments) are used here for the first time in FlowLibs.

Use Case

A team building a chatbot or enterprise search experience over their document corpus needs the AI Search index updated as soon as a document changes in SharePoint — not on the next built-in indexer cron — and needs deletions to flow through so stale results never surface.

Flow Architecture

Environment Variables

Connectors & Connections

Connector	API name	Actions used
Azure AI Search	shared_azureaisearch	IndexDocuments DeleteDocuments
SharePoint	shared_sharepointonline	GetOnUpdatedFileItems GetFileItems GetFileContent
Microsoft Dataverse	shared_commondataserviceforapps	ListRecords CreateRecord UpdateRecord

Note — All connections are referenced as solution connection references; the flow is portable between environments as long as a connection is mapped at import time.

Customization Guide

Almost every realistic variant of this flow can be implemented by changing environment variable values. A few cases require small edits inside the flow definition — those are called out explicitly below.

Vector / RAG search

Add an Azure OpenAI embeddings call before Index Document and push the vector into a Collection(Edm.Single) field for semantic search.

Binary documents

base64ToString extracts text from text/markdown/HTML. For PDF/Word, insert a text-extraction step (Azure AI Document Intelligence prebuilt-read / prebuilt-layout) before Compose Content.

Chunking

Split large documents into chunks and index each as a separate doc with a parent id (extend the search key, e.g. sp-42-chunk-3).

Indexer frequency

Tune the event trigger's polling interval (5 min) and the sweep's recurrence (6 h) to your freshness vs. cost needs.

Real-time deletes

If your environment exposes a reliable delete signal, replace the sweep with an event-driven delete flow keyed on the same ledger.

Key Expressions

The flow is intentionally light on Power Fx / WDL gymnastics — the heaviest expressions are the branch-name concatenation and the approval outcome check. They are listed below in the order they appear in the flow.

@concat('sp-', string(triggerOutputs()?['body/ID']))

@if(greater(length(string(outputs('Compose_Raw_Text'))),32000),substring(string(outputs('Compose_Raw_Text')),0,32000),string(outputs('Compose_Raw_Text')))

[{ "@search.action":"mergeOrUpload", "id": ..., "title": ..., "content": ..., "url": ..., "library": ..., "modified": ..., "tags": @createArray(parameters('EnvVar_KbLibraryName')) }]

[{ "id": "@{items('For_Each_Ledger_Row')?['flowlibs_name']}" }]

@equals(string(item()?['ID']), items('For_Each_Ledger_Row')?['flowlibs_sharepointitemid'])

Customize & download