skills.rest
memasanz
memasanzCOMMUNITY
Explore all skills by this author →
GitHub

azure-content-understanding

Extracts structured RAG-ready data from documents, images, audio, and video via Azure AI Content Understanding.

1|Updated May 29, 2026
Search Tags:#ocr#python#rag#azure#multimodal#document extraction#content understanding
One-click install
npx skills add https://github.com/memasanz/agent-harness --skill azure-content-understanding-memasanz
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: azure-content-understanding
Source: https://github.com/memasanz/agent-harness/tree/main/.github/skills/azure-content-understanding
Command: npx skills add https://github.com/memasanz/agent-harness --skill azure-content-understanding-memasanz

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires azure-ai-contentunderstanding, azure-identity, aiohttp.

What problem does it solve? Turning unstructured PDFs, images, audio, and video into structured, machine-readable output for RAG pipelines and workflow automation requires choosing the right analyzer, configuring a Microsoft Foundry resource correctly, and handling async long-running operations — mistakes in any of these steps cause every call to fail. ## Core Features & Use Cases - Prebuilt Analyzer Selection: Guides choosing among documentSearch, imageSearch, audioSearch, videoSearch, invoice, receipt, layout, read, and domain analyzers for identity, tax, mortgage, and contract documents. - Custom Analyzers: Builds field-schema-based extractors on top of prebuilt base analyzers across all four modalities. - Foundry Resource Setup: Covers region selection, Cognitive Services User RBAC, required model deployments (gpt-4.1, gpt-4.1-mini, text-embedding-3-large), and the one-time default model mapping. - Use Case: Ground a RAG pipeline by running prebuilt-documentSearch over a folder of PDFs to get markdown, tables, figures, and per-item summaries ready for embedding and retrieval. ## Quick Start Ask the agent to extract structured fields from an invoice PDF using the azure-ai-contentunderstanding SDK with the prebuilt-invoice analyzer.

Frequently Asked Questions about azure-content-understanding

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I extract structured data from PDFs with Azure AI Content Understanding?▼

Install azure-ai-contentunderstanding, create a ContentUnderstandingClient with DefaultAzureCredential, and call begin_analyze with prebuilt-documentSearch for RAG markdown output or prebuilt-invoice for typed fields. Call poller.result() to get the AnalysisResult.

Which prebuilt analyzer should I use for documents, images, audio, or video?▼

Use prebuilt-documentSearch for PDFs, Office docs, and text-bearing images; prebuilt-imageSearch for standalone photos; prebuilt-audioSearch for transcription with diarization; and prebuilt-videoSearch for video frames plus transcript. Domain analyzers like prebuilt-invoice return typed fields instead.

Why do prebuilt analyzer calls fail with DeploymentNotFound?▼

The one-time default model mapping was never run on the Foundry resource. Run sample_update_defaults.py once per resource after deploying gpt-4.1, gpt-4.1-mini, and text-embedding-3-large so prebuilt analyzers can resolve their model deployments.

Does Azure Content Understanding support real-time audio or video streaming?▼

No. Content Understanding is a batch, asynchronous long-running operation service. Every analysis returns an LROPoller that you resolve with poller.result(); it is not designed for real-time streaming input.

Why do I get 403 errors when configuring default model deployments?▼

The calling identity lacks the Cognitive Services User role on the Foundry resource. Being the resource owner is not sufficient — the API explicitly checks this RBAC assignment, so grant it to your user, service principal, or managed identity.

When should I use Document Intelligence instead of Content Understanding?▼

Use Document Intelligence prebuilt-read when you only need plain OCR without markdown, layout, or RAG-oriented output. Content Understanding is the better fit when you need multimodal analysis, summaries, or custom field schemas.

Related Skills

Explore More Software Engineering Skills→
microsoftmicrosoft
1.4k

deploy-model

Deploys Azure OpenAI models with intent-based routing across preset, customize, and capacity discovery modes.

Official
Advanced
shalevaminshalevamin

rag-architect

Designs and implements RAG systems with chunking, embedding generation, vector stores, hybrid search, reranking, and evaluation.

Community
Advanced
InnerScopeHearingInnerScopeHearing

fleet-secret-custodian

Manage Azure Key Vault secret lifecycles with autonomous key rotation and tamper-evident audit logging.

Official
Advanced
MoonAxisMoonAxis

azure-ai-vision-imageanalysis-py

Analyze images to extract captions, tags, objects, OCR text, and people using the Azure AI Vision SDK for Python.

Official
Intermediate