SetFN
SetFNDocs2026년 10월 5일

Deploy an AI support chat sample

Enable advanced Webspace AI support and test the support flow in the generated chat.html. This setup uses the model and Agent execution pipeline already configured by an administrator instead of creating a separate model or Agent for each Webspace deployment.

Before you begin

  • You need permission to deploy schemas and APIs.
  • A server administrator must connect an AI provider, then register and enable a chat model.
  • Decide which members, levels, teams, or administrators receive the AI chat Use permission.
  • Members who open chat.html must be able to sign in to the same SetFN site.
  • Review the visibility of support data and the system instructions.

Prepare administrator AI models

Connect an AI provider in the administrator settings, then register and enable a chat model for support. If Primary model and Fallback model are blank in Webspace, the server automatically prioritizes enabled chat models that are ready to use. For requests that include an image, it also applies the administrator's OCR/VL model priorities.

Set Primary model and Fallback model under Advanced settings only when a Webspace needs explicit model overrides. Do not place provider keys or model secrets in Webspace pages or chat.html. The support page calls the deployed AI feature connection, while model selection and Agent execution use the existing server pipeline.

Configure advanced AI chat settings

Enable AI chat in the schema's feature connections, then open Advanced settings.

  1. Set AI chat to Enabled.
  2. Leave Primary model and Fallback model blank unless you need explicit model overrides.
  3. Enable Use tools only if the support flow needs tools. It is off by default.
  4. Enter the support scope and response rules in the system instructions.
  5. Configure the Use permission.
  6. Enable Generate a support chat sample.

The schema stores the setting in this structure:

{
  "apiBuild": {
    "features": {
      "ai-chat": {
        "enabled": true,
        "sampleChat": true
      }
    }
  }
}

sampleChat controls only sample generation. It does not copy or overwrite the existing administrator model and Agent settings.

Generate chat.html during deployment

Webspace generates chat.html when both enabled and sampleChat are true for AI chat and the deployment includes frontend samples. Select a client deployment path and deploy the schema. chat.html is stored in the current frontend sample output folder alongside index.html and test.html; deployment does not create a separate folder for AI support.

support-service/
├── index.html
├── test.html
├── chat.html
├── style.css
└── README.md

Sign in as a member on the same SetFN site, then open chat.html. The sample bootstraps member information from the current session, so do not assume that it works as anonymous public chat.

  1. Confirm that a member without the AI chat Use permission cannot send a request.
  2. Send a message from an account with permission.
  3. Confirm that the response arrives through the existing administrator model or Agent pipeline.
  4. If Use tools is enabled, confirm that only allowed tools can run.
  5. Disconnect the connection and confirm that the page shows an error state without duplicating the message.

Adapt the support page

The generated chat.html is a starting point for testing the connection and permissions. You can adapt the guidance, suggested questions, and visual style. Do not replace the connection, authentication, or server-managed model selection with client-controlled configuration.

In the system instructions, define the support scope and how to handle questions that cannot be verified. Before launch, test questions, tool calls, and error responses with accounts representing each permission level.

Notes

  • chat.html is a sample client. It does not create a separate AI model or Agent.
  • When Primary model and Fallback model are blank, the server selects a model automatically from the administrator configuration and available model state.
  • Sample generation and Agent protocol connectivity do not guarantee a response from a live AI provider. Before launch, send a real message through the provider, model, or Agent configured on the server.
  • The sample uses the existing AI Agent protocol and displays tool status, approval-pending, and input-pending notices. It does not provide controls that complete an approval or supply the requested input.
  • Use tools is off by default. Before enabling it, confirm that a separate UI can handle any approval or additional input required by the Agent.
  • The sample requires member sign-in and session bootstrap; anonymous public use is not guaranteed.
  • Review tool permission separately from the AI chat Use permission.
  • Add guidance that prevents users from entering personal data or secrets in support messages.
  • Check which files may be overwritten before redeploying a modified sample.