How microsoft/content-processing-solution-accelerator works
Project type and purpose
• This is a full-stack, event-driven content processing accelerator built on Azure.
• It ingests unstructured multi-modal content, applies ML/OCR and schema mapping via Azure AI services, scores results, stores outputs, and exposes APIs and a web UI for integration and human validation.
File structure
• Top-level directories:
• infra – Bicep templates and scripts provisioning Azure resources (Container Apps, Blob, Queue, Cosmos DB, Key Vault, App Configuration, AI services, etc.)
• src/ContentProcessor – Python worker application in a container that processes queue messages, calls AI services, writes to storage/Cosmos
• src/ContentProcessorAPI – Python web API container exposing REST endpoints for ingestion and status, interacting with storage, queue, Cosmos
• src/ContentProcessorWeb – React/TypeScript single-page app with MSAL authentication, calling the API
• CI/CD and development configs: GitHub workflows, Dev Containers, Dockerfiles, linting, testing.
README for additional insights
• Architecture uses Azure AI Foundry, Content Understanding, OpenAI, Blob Storage, Queue Storage, Cosmos DB, Container Registry, and Container Apps.
• Processing pipeline is event-driven: upload → queue → worker → database → API → UI.
• Human-in-the-loop review when confidence thresholds are not met.