Using Ask Sage on Edge
Chat, work with documents, and use the tools available on your Edge deployment
Table of Contents
About this guide
Sign in to your organization’s Edge workspace to chat, work with documents, and use the tools enabled for your deployment. Most workspace interactions follow the standard Ask Sage guides. The available models, network connections, and installed release determine what you can use on Edge.
Chat with a local model
- Start a new conversation.
- Choose an available model from the model picker. Ask your administrator which models are approved for your task and data.
- Enter your prompt and send it.
- Review the answer before using it, especially names, numbers, citations, and instructions.
Local models share the appliance’s capacity. Longer prompts, larger documents, and simultaneous requests can increase response time. If a model disappears or stops responding, an operator may be loading a replacement or performing maintenance.
The full cloud model catalog is not automatically available on Edge. A visible model also may not support images, audio, tool calling, or the same context length as another model. Use its model card and your administrator’s guidance.
Work with documents and datasets
- Upload an approved document or select a dataset you can access.
- Wait for processing to finish.
- Select the document or dataset as context for your conversation.
- Ask a specific question about its contents.
- Inspect the retrieved material or citations to verify the answer.
Edge uses local embedding for its local retrieval path. The selected chat model determines where the subsequent response is generated. If a cloud model is selected, prompt content and retrieved excerpts used for that request can be sent to the configured provider. Follow your organization’s rules for model selection.
Use the dataset guide and dataset management guide for the workspace controls. Processing and supported file types depend on the installed services; a successful upload does not by itself confirm that retrieval is ready.
Voice, images, and other tools
Where installed, local speech-to-text and text-to-speech services support microphone input and spoken responses without an internet connection. Allow microphone access in your browser when prompted. Voice transcription is a separate capability from a chat model’s ability to understand audio directly.
Image understanding requires a compatible vision model. Agent workflows and integrations require both a capable model and access to the tools or services the workflow calls. A workflow that uses a remote website or cloud connector will still need that connection when its chat model runs locally.
Related workspace guides
| Task | Guide |
|---|---|
| Learn the chat interface | Getting Started |
| Choose models and prompts | Models, personas, and prompt templates |
| Manage personal preferences | Account Settings |
| Build supported workflows | Agent Builder |
| Manage workspace users | Enterprise User Management |
Examples in the general guides may use a cloud URL. Stay on your Edge deployment’s address and use the features enabled there. Cloud subscription and token-purchase instructions do not define your organization’s Edge license or usage policy.