GPTGuard

Overview

GPTGuard expands enterprise readiness of SharePoint-based context ingestion, enhanced observability for chat latency, new Gemini 3 and Nano Banana model options, and higher configurability in the chatbot UI.

1. SharePoint integration for contextual grounding

Connect SharePoint Online as a source so the chatbot can ground answers on bank-approved content.

Key capabilities

• SharePoint Online only for this release

• Supported file types: PDF, DOC, DOCX, JPG, PNG

• Maximum file size: 10 MB per file

• Change handling: ignores deletions and metadata-only changes; processes content updates only

• Admin UI integration with three new screens: Source Connectors, Source Folder Discovery, and Source Sync Monitoring

Admin UI screens

• Source Connectors

  • Configure SharePoint Online connections

  • Store and manage connector credentials

  • Enable or disable specific connectors

• Source Folder Discovery

  • Browse and select SharePoint sites, document libraries, and folders

  • View folder-level metadata and item counts

  • Control which folders participate in sync

• Source Sync Monitoring

  • Track sync job status and history

  • Review item-level ingest statistics and errors

  • Trigger manual re-syncs when needed

Key benefits

• Keeps chatbot responses aligned with the latest internal SharePoint content

• Reduces manual upload effort by automatically ingesting approved documents

• Improves governance by separating connector setup, folder selection, and sync monitoring

Notes

• Support is limited to SharePoint Online; on-premises SharePoint is out of scope for this release.

• Deleting a file in SharePoint does not immediately remove it from the chatbot context; plan retention policies accordingly.

2. Health monitoring and observability for GPT Guard

Gain better visibility into GPT Guard performance, with a focus on chat latency.

Key capabilities

• New Latency Statistics dashboard tab in the GPT Guard monitoring view

• Displays the current week’s average latency for chat interactions

• Breaks out latency by key dimensions (for example, endpoint or use case) where available

• Supports time window switching and basic filtering for focused analysis

Key benefits

• Helps SRE and platform teams detect emerging performance issues early

• Supports SLA reporting for internal stakeholders

• Simplifies troubleshooting by separating latency from other health metrics

Notes

• Historical latency beyond the current week remains available via existing logs or reports where configured.

3. Gemini 3 model support (Flash and Pro)

Add the latest Gemini 3 models for text chat while reusing your existing configuration patterns.

Key capabilities

• Gemini 3 Flash support for low-latency, high-throughput text chat

• Gemini 3 Pro (Preview) support for richer reasoning in text chat flows

• Model selection via existing configuration mechanisms; no new config surface required

• Backwards-compatible defaults that preserve current model behavior unless changed

Key benefits

• Improves response quality and speed by adopting the newest Gemini 3 families

• Lets you experiment with Gemini 3 Pro in preview while keeping Gemini 3 Flash for production flows

• Reduces operational overhead by reusing existing model configuration and routing logic

Notes

• Gemini 3 Pro is available in preview only and should not be treated as a long-term production default yet.

4. Nano Banana image generation

Enable prompt-based image generation for chatbot experiences using the Nano Banana flow.

Key capabilities

• Integrates an image generation model based on Gemini-2.5-flash

• Supports prompt-based image creation directly from chatbot conversations

• Returns generated images as part of the chatbot response payload

• Respects existing guardrails and content policies where applicable

Key benefits

• Allows richer, visual responses for customer support and internal knowledge flows

• Keeps image generation within the same governance boundaries as text-based GPT Guard flows

• Reduces the need for separate image-generation tooling for common scenarios

Notes

• Image-generation usage may have separate cost and quota implications; coordinate with your platform team.

5. Customizable chatbot tab name

Rename the chatbot browser tab without code changes to align with internal branding.

Key capabilities

• Configurable chatbot tab name from administrative configuration

• Applies to the browser tab title shown to end users

• Takes effect on the next page reload after configuration change

Key benefits

• Aligns the chatbot experience with Bank Muscat naming conventions

• Reduces dependency on engineering changes for minor branding updates

Notes

• This is a no-code-change capability; admins can update the name directly in configuration.