v4.0.0

Released October 2026

A major release that makes it easier to create reusable prompts, organize your work, manage access, and understand your usage — with improvements to chat reliability, source transparency, document handling, and accessibility. The modern Ask Sage interface is now the only interface.

Forward‑only upgrade. This release includes database changes and cannot be rolled back to the previous version once applied. Self‑hosted and Ask Sage Edge deployments should plan accordingly; managed (SaaS) tenants are upgraded by Ask Sage, and no customer action is required.

Confirming this release. The modern Ask Sage interface replaces the legacy one for everyone, and a persistent sidebar gives you recents and a single search across chats, canvases, agents, and workbooks. Some capabilities require enablement for your deployment — contact your Ask Sage administrator to confirm availability.

✨ Create and organize your work more easily

AI‑assisted prompt templates. Create a reusable prompt template from a short description, then review and refine the suggested title, description, and prompt. AI assistance is also available when editing existing templates.
More recognizable dataset names. Dataset selection shows more recognizable names.
Bulk workbook management. Select and delete multiple workbooks at once, with related content and sharing removed as part of the deletion.
Unified navigation and search. Reach recent chats, canvases, agents, and workbooks from a persistent sidebar, and search across all four from one place.

🔎 Get clearer answers and better visibility into sources

Clearer source attribution. Source information better reflects the datasets actually queried for a response, with retrieval errors shown separately.
Better response details. See which model answered your request; saved conversations also retain the datasets, persona, and agent used.
Cleaner Deep Research reports. Completed reports show a clear completion status.
More control over references. Supported deployments can configure up to 100 references per request for more comprehensive retrieval. Availability varies by deployment.
More reliable web search.

💬 Work more reliably with chats and attachments

Improved handling of large conversations.
Clearer response messages. When a model request can’t be completed, you’ll see a helpful explanation.
Better progress feedback. Long‑running responses show elapsed time and updated status messages.
Improved image handling. Supported government‑hosted models process images directly, and models that don’t accept images tell you so clearly.
Direct PDF support for Anthropic models. Attach supported PDFs directly to Anthropic models, preserving the document rather than converting it to text alone.
More reliable Word attachments. Improved handling of Word documents with supported Anthropic models.

🛠 Manage access and organization settings with greater control

Organization policy management. Administrators can manage available models, features, API‑key permissions, default member roles, and token limits from organization settings.
Easier policy updates. Search and select allowed models, restore individual settings to their defaults, and retry an unsuccessful save without losing edits.
IP‑based access restrictions. Administrators can manage permitted IP addresses and network ranges, with validation and a warning before saving a restriction that would exclude their own IP address.
Clearer integration controls and visibility. Authorized administrators can deactivate and reactivate OAuth‑connected integrations, and more easily tell organization availability from an individual user’s connection status.
Simpler API‑key creation. Clearer permission labels and guidance make it easier to choose the access a new key should have.

📊 Understand and manage usage more easily

Personal usage visibility. Members of organizations with pooled billing can view their own token consumption.
Clearer limit notifications. Members receive more relevant guidance when they reach an administrator‑defined token limit.
Local model usage (Ask Sage Edge only). Inference on local, on‑appliance models is exempt from billing and quota; usage counts remain available for monitoring, and cloud‑model billing is unchanged.
Streamlined reporting. The redundant “Users with Tokens” report and the “Label Usage” report have been retired; the main Users report remains available.

♿ A clearer, more accessible experience

Accessibility improvements. Updates across the interface improve contrast, readability, and accessibility, including dates, source cards, and account initials.
Clearer classification labels. Dataset selectors show more legible classification indicators, and web search is blocked while a Controlled Unclassified Information (CUI) dataset is selected.
More useful sign‑in and permission messages. Smart‑card sign‑in messages now point clearly to a certificate or network cause, and restricted administration screens show clearer access messages.
Smoother account and administration workflows. Refinements to registration, activity‑log access, user‑list filtering and sorting, and administrator access.
V1 interface retired. The legacy V1 interface has been retired, and the V2 interface is now the only Ask Sage interface available.

🔌 For API users

This release improves reliability for Anthropic model requests routed through Vertex, provides clearer validation messages for missing required request fields, and adds an optional setting for more consistent error‑status handling on the v1 completion endpoint. The optional error‑status behavior is deployment‑dependent; default behavior is unchanged.

🧪 Early access — available via API (beta)

These capabilities are available through the API as of this release; a management interface is planned for a future release.

Chat folders. Organize conversations into folders.
Rename workbooks and datasets. Update names without recreating your work.
Member management (administrators). Remove a member from an organization, and transfer a user between organizations with approval and an audit record.
Administrator reporting. Export dataset statistics and additional organization reports.

🔒 Security & platform

Ask Sage services are delivered as FIPS‑hardened, digest‑pinned, and cosign‑signed images within the FedRAMP Class D certification boundary, and are continuously rebuilt against the latest patched base images as part of Ask Sage’s continuous‑monitoring commitment.