settings

Settings Guide

Manage project-level settings in Cognigy via the manage_settings tool.

Voice Preview Settings

Configure a speech provider so voice endpoints (WebRTC) can synthesize and recognize speech.

Quick Start

{
  "operation": "set_voice_preview",
  "projectId": "<24-char hex>",
  "provider": "microsoft"
}

This auto-detects an existing speech connection for the provider. If none is found, you'll get instructions to upload a package containing one.

With explicit connection

{
  "operation": "set_voice_preview",
  "projectId": "<24-char hex>",
  "provider": "microsoft",
  "connectionId": "<connection referenceId>"
}

Supported Providers

No speech connection found?

Speech connections are typically installed via Cognigy packages. To add one:

1. manage_packages { operation: "upload_and_inspect", projectId, filePath: "<path to package.zip>" }
2. manage_packages { operation: "import", projectId, packageId }
3. Retry manage_settings { operation: "set_voice_preview", projectId, provider }

Typical Full Workflow

1. create_ai_agent → get projectId
2. setup_llm → configure LLM
3. manage_settings { operation: "set_voice_preview", projectId, provider: "microsoft" } → configure speech
4. manage_voice_gateway { projectId, flowId } → create voice endpoint with WebRTC

Knowledge AI Settings

Configure the project-level settings used by Knowledge Search and document parsing.

This is separate from the embedding model used by manage_knowledge to build the knowledge-store index. Do not confuse these settings:

• Embedding model: required for the knowledge store itself
• Knowledge Search model: configured here via knowledgeSearchModelId
• answerExtractionModelId is also supported by the tool, but it is usually not needed for normal AI-agent knowledge-store setups.
• knowledgeSearchModelId must reference an llm_model from the same project
• The accepted model type for knowledgeSearchModelId is instance-dependent