Release notes

This page tracks new capabilities, enhancements and bug fixes across DataQI platform releases.

v2.4.3

Date: 10 September 2026

This release includes v2.4.1 and v2.4.2. The summary below covers changes since v2.4.0.

New features

• Headless DataQI: Integrators can build custom user interfaces and unattended automation through a supported API and official TypeScript and C# SDKs. Custom interfaces retain the signed-in user’s access, while application clients use short-lived credentials and explicit workflow grants. Diagnostics provides client registration and access administration.
• Workflow identities: Workflows now run as the DataQI system identity, with the initiating user or application recorded separately for attribution. Administrators can authorise dedicated DataQI connections, enabling workflows to use MCP tools and connected data without inheriting the initiator’s personal access. See DataQI workflow connections.
• Spreadsheet queries (beta): Agents can query prepared CSV and Excel (XLSX) data through the SQL query capability and return datasets for inspection and follow-up questions. This capability is not yet fully tested; validate results before relying on them.
• Legacy Word documents: Document ingestion now supports Word .doc files alongside existing formats.

Bug fixes

• Chat & UI: Fixed new messages not appearing correctly when sent while viewing earlier conversation history, and improved conversation input focus.
• Agent feed: Restored access for Basic Access users and limited live event subscriptions to visible agent cards.
• Administration: Corrected workflow grant status reporting, removed application-only roles from the user role picker and improved connection table accessibility.
• Workflows: Corrected audit attribution across user and application requests.
• Data & ingestion: Preserved empty spreadsheet date values during extraction.
• Authentication: Prevented email-based account linking and corrected handling of absent identity claims.

Workflow integration upgrade

The legacy global workflow API key has been removed. Existing integrations must move to headless application-client authentication and explicit pipeline grants. Administrators must authorise the connections required by workflows for the DataQI system identity. During migration, unfinished legacy runs without an initiating identity are marked as failed and must be started again if needed.

v2.4.0

Date: 31 August 2026

New features

• MCP connections: Administrators can add anonymous or OAuth 2.1 Model Context Protocol (MCP) servers through the standard Connections experience. Agents can discover and invoke tools using the current user’s authorised connection, subject to the MCP server URLs allowed for that agent.
• Google Drive connections: Users can authorise Google Drive, synchronise documents they can access and open referenced files in the native Google Drive interface.
• Charts: Agent responses can now include interactive bar, line and donut charts, with improved formatting for currencies, dates, percentages and multiple data series.

MCP migration required

This release replaces the legacy global MCP server configuration. Existing MCP servers and agent assignments are not migrated. Administrators must recreate each integration as an MCP Connection, allow its canonical server URL on the relevant agents and let the Connection synchronise its tool catalogue.

Workflows cannot use MCP connections

MCP connections cannot be used by workflows in this release because workflows execute without an identity. The next release is planned to introduce system identities for workflows, allowing workflows to run as DataQI and use shared connection instances owned by DataQI.

Review MCP tool permissions

MCP tools are not restricted to read-only actions, and DataQI does not yet provide a human-in-the-loop approval step before an agent invokes a tool. Review each server’s capabilities and the permissions granted to its credentials before connecting it. For systems where an incorrect action could cause material harm, use least-privilege or read-only credentials where available, isolate the integration, or defer connecting the system until appropriate controls are in place.

Enhanced features

• Chat & UI: Conversation timestamps now update relative to the current time.
• Chat & UI: Conversation navigation now distinguishes unread responses from conversations where an agent is still working, and page titles provide more useful conversation context.
• Chat & UI: Responses can use more diverse document search results generated from multiple query variants.
• Connections: OAuth sign-in is more consistent across connection types, and interrupted connection synchronisations recover automatically after service restarts or transient failures.
• Data & ingestion: Excel workbooks can be processed into persistent structured tabular data and schema metadata, preparing them for future analytical querying. This release does not make spreadsheet-backed data available to agents for querying.
• Authentication: Session renewal is more resilient to transient failures and is coordinated across browser tabs.

Bug fixes

• Chat & UI: Fixed race conditions that could leave a rapidly switched conversation blank, incomplete or showing a generic working status.
• Connections: Improved inline feedback for connection form and authentication errors.
• Connections: Fixed concurrent validation of MCP server URLs and made endpoint changes invalidate stale tool catalogues safely.
• Security: Hardened connection authorisation, service-host validation and sensitive audit filtering.

v2.3.0

Date: 31 July 2026

New features

• SQL query tool: The agent can now generate and execute safe, read-only SQL queries against connected relational data sources. Queries produce standard DataQI datasets for interactive table rendering and follow-up dataset inspection. For more information, see SQL query.
• PostgreSQL database connections: DataQI supports PostgreSQL as a first-class connection type. Users authenticate with their own credentials, which the platform encrypts and stores securely, providing a per-user, isolated view of the database schema. For more information, see relational database credentials.

Enhanced features

• Chat & UI: Interactive table artefacts now support the dataset outputs produced by the SQL query tool.
• Chat & UI: Conversation history is now sorted by last activity rather than creation time, keeping active conversations at the top.
• Connections: Improved the UX for SMB connection authentication, streamlining the overall connection flow.
• Data & Ingestion: Improved feedback and status reporting when document extraction encounters unsupported states.

Bug fixes

• Chat & UI: Fixed an issue where search results would repeat in chat.
• Chat & UI: Corrected the DataQI logo link behaviour.
• Connections: Fixed an issue where modal buttons did not correctly return users to the connections page.
• Connections: Hid cleared lifecycle messages in the connections table to reduce noise.
• Data & Ingestion: Prevented complete agents from making authoritative claims about partial dataset results when they don’t have all of the facts.
• Security: Enforced scoped diagnostics and protected assistant response audits from exposing sensitive data.