Agent Landing Zone Data Ingestion

The Python 3.12 ingestion service for Agent Landing Zone. It chunks documents, generates embeddings, writes Azure AI Search content, and runs Blob, SharePoint, NL2SQL, and maintenance jobs. It also exposes ingestion APIs, an operator dashboard, and a separately gated hosted retrieval API. It does not execute conversational agents or store their full chat history.

Documentation and release scope

• Product documentation owns deployment topology and shared setup.
• Canonical documentation source is on Azure/AI-Landing-Zones main, not a legacy documentation branch.
• Runtime, API, security, scheduling, and audit guide covers this service.
• Contributor quality commands and operator dashboard development cover local engineering.

Agent Landing Zone 4.3.0 pins v3.1.1 at 72efd8d7d51e10d38bf8b89f03b7eb42e1ed8ff2. Repository main also contains later naming and App Configuration label changes. The runtime guide marks the label difference explicitly. main or develop is not evidence that a change is in the platform's shipped image.

Run this service locally

Obtain compatible infrastructure and a development App Configuration store through Agent Landing Zone first. This repository's Bicep file does not provision a standalone environment. The selected sources, models, storage, Search index, identities, and network access must already be configured.

Starting the service can write and delete data. Startup jobs default to enabled, and cron scheduling remains active even when immediate startup jobs are disabled. Review the scheduling defaults before connecting to any environment.

From this repository's root, with Python 3.12 and Azure CLI installed, run:

python -m venv .venv
.\.venv\Scripts\python -m pip install -r requirements.txt
az login
$env:APP_CONFIG_ENDPOINT = "https://<app-configuration-name>.azconfig.io"
.\.venv\Scripts\python -m uvicorn main:app --host 127.0.0.1 --port 8080

These are service startup commands, not infrastructure deployment commands. See configuration and identity for RBAC, runtime values, and environment opt-in. GET /healthz and GET /readyz are process probes, not proof that a source was indexed.

Optional deployment

The ingestion Container App is optional, independently of Container Apps or Foundry hosted orchestration. Fresh supported native Foundry IQ deployments default its platform deployment control to false. Existing and custom selections are preserved. Native Blob ingestion does not replace conversation attachments or all source types.

Enabling deployment does not change retrieval backends or sources, enable POST /retrieve, or remove API-key requirements from ingestion APIs. See deployment and retrieval boundaries.

Contributing

Follow the engineering contract and contribution guidelines. Audit schemas and their integrity pins are in contracts/.

Trademarks

Microsoft trademarks and logos are subject to Microsoft's guidelines. Modified versions must not imply sponsorship. Third-party trademarks follow their owners' policies.