-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Frequently asked questions about Voice Agent, getting started, and troubleshooting.
Voice Agent in Foundry Agent Service combines real-time voice interaction with agent capabilities such as instructions, tools, knowledge, and orchestration.
Voice Live provides the real-time voice capabilities used by Voice Agent.
This repository is a community hub containing documentation, samples, a local demo portal, and a place to report bugs or request features.
Start with the repository overview.
Choose the experience that fits your needs:
- Browser experience: Follow the portal setup guide.
- Code examples: Follow the samples guide.
Each guide describes its prerequisites, configuration, and startup steps.
Requirements vary by example. Generally, you need an authorized Azure Foundry project, Azure credentials, and the dependencies listed in the example's README.
Voice conversation examples also require a microphone and audio output device. Check the relevant guide for runtime versions and additional audio dependencies.
The repository documents Linux as its supported environment and recommends WSL2 for Windows. Individual samples may document additional platform support or exceptions.
Follow the platform instructions for the specific example you are running.
The portal runs locally, but its backend uses Azure Foundry. It requires network access and an authorized Foundry project.
See the portal guide for configuration details.
Yes. The samples demonstrate integrations such as MCP tools, local functions, and Foundry IQ.
Check the samples guide for an example matching your scenario and its setup requirements.
Check the following:
- Your Azure authentication and project permissions.
- The configured project endpoint and agent settings.
- The sample's required dependencies.
- Microphone permissions and selected audio devices.
- Network connectivity and the exact error message.
If the problem continues, report it with the diagnostic details below.
Search existing issues, then open a new issue.
Include:
- Steps to reproduce.
- Expected and actual behavior.
- Sample name, SDK version, and operating system.
- Azure region and model, when applicable.
- Failure time in UTC and available session or request IDs.
- Sanitized error messages or logs.
Remove credentials, tokens, and personal data before posting.
Open an issue with a title such as [Feature] <brief description>.
Describe the problem you want to solve, your proposed capability, an example workflow, existing workarounds, and what a successful implementation would look like.