> ## Documentation Index > Fetch the complete documentation index at: https://knowledge.cloudquant.com/llms.txt > Use this file to discover all available pages before exploring further. # What's New in Liberator 2.3 > Release notes for the CloudQuant Data Liberator 2.3 release — AI-driven reporting, AI-assisted discovery, HubSpot connectivity, invite-via-email onboarding, SharePoint wildcards, and reliability improvements # What's new in Liberator 2.3 Liberator 2.3 centers on **AI-driven reporting and discovery**, **expanded connectivity**, and **smoother account onboarding**. Headline capabilities include an Admin UI AI Reporting panel for natural-language usage and entitlement reports, AI-generated dataset column descriptions, a HubSpot connector for CRM analytics, wildcard support for OneDrive and SharePoint file feeds, invite-via-email user provisioning with a forced password change on first login, and a round of reliability fixes across date parsing, query workers, and packet-capture ingest. ## Highlights Ask plain-English questions about usage, users, and entitlements in the Admin UI, with quick-action reports and Excel exports. Onboarding and metadata edits can suggest human-readable column descriptions, ranked by how often each description is used in your environment. Connect HubSpot CRM objects into Liberator and query them through the same Liberator surface as market and file-backed datasets. Create users with an email invite, send password-reset mail from the Users page, and require a new password on first login. Pattern-match multi-file and dated folder feeds without enumerating every path. Connect ChatGPT, Copilot Studio, and Claude Desktop to Liberator over MCP. ## AI-driven reporting Liberator 2.3 adds an **AI Reporting** panel (BETA) to the Admin UI. Admins can ask about platform usage, users, permissions, and entitlements in plain English, or request a downloadable report, instead of building manual queries or filters. What lands in 2.3: * **Quick-action reports** — one-click presets for yesterday's usage, today's user report, current entitlements, and month-to-date / last-month usage. * **Free-form natural language** — ask for a user's usage over a date range; the assistant resolves datasets and returns per-dataset totals (queries, records, bytes, GB). * **Excel exports** — downloadable `.xlsx` reports from both quick actions and free-form questions. * **Clarifying prompts** — if a username does not match, the assistant suggests likely accounts instead of failing silently. See [AI-based reporting](/administration/ai-based-reporting) for the admin workflow. ## AI-generated column descriptions When you onboard a dataset or edit column metadata, Liberator can **suggest descriptions** grounded in descriptions already used on similar columns in your environment. What lands in 2.3: * **AI-assisted suggestions** during the Columns step of dataset onboarding and edit flows. * **Frequency-based ranking** so the most commonly used descriptions appear first. * **Tighter relevance filtering** so broad substring matches no longer crowd out useful suggestions. See [Column description suggestions](/administration/column-description-suggestions) for the admin workflow. ## HubSpot connector Liberator 2.3 adds a first-class **HubSpot** connection type for CRM objects. Once connected, HubSpot-backed datasets use the same query API and entitlements model as file- and database-backed sources, so analysts can join commercial and market data without a separate export pipeline. See [HubSpot](/datasource-config/hubspot) for connection setup. ## Invite-via-email and first-login password change User provisioning no longer requires handing temporary passwords around out of band: 1. **Invite via email** — when creating a user, Super Admins can send an invite email so the recipient completes account setup themselves. 2. **Send reset email** — from the Users page, trigger a password-reset email for an existing account. 3. **Forced password change on first login** — newly provisioned users must set their own password before using Liberator. See [Managing users and entitlements](/administration/managing-users-and-entitlements#invite-via-email). ## SharePoint and OneDrive wildcards Building on the share-link paste flow introduced in 2.2, Liberator 2.3 adds **wildcard matching** for OneDrive and SharePoint sources. Recurring or partitioned file feeds (for example, `reports/2026-*/trades_*.csv`) can be onboarded without listing every folder or file. See [SharePoint and OneDrive](/datasource-config/sharepoint-onedrive). ## AI onboarding and MCP clients 2.3 also deepens Liberator's other AI surfaces: * **AI onboarding flow** — guided dataset setup with AI assistance for naming and column documentation. * **ChatGPT** — connect ChatGPT to Liberator with OAuth. See [Connecting Liberator to ChatGPT](/integrations/chatgpt). * **Microsoft Copilot Studio** — add Liberator as an MCP tool with OAuth 2.0 Dynamic discovery. See [Microsoft Copilot Studio](/integrations/copilot). * **Claude Desktop** — install the Liberator `.mcpb` extension bundle. See [Claude Desktop](/integrations/claude). All three MCP client guides live under the **Integrations** tab. ## Connection and admin polish Smaller usability improvements in the same release: * **`liberator.metadata()`** — Python SDK helper for `POST /liberator/metadata`, returning concordance-backed row/symbol counts and date spans. See [Dataset metadata](/api-reference/concepts/dataset-metadata). * **Disable data sources** — administrators can hide deprecated or unused source types from the connection selection list. * **Original source file name** — the Dataset & Connection modal shows the original file name for better traceability. * **Whitespace trimming** — leading and trailing spaces are stripped from connection and dataset form fields automatically. * **Sort resets to page one** — sorting the Connections & Datasets list returns you to the first page of results. * **System Monitoring charts** — legends stay contained and hover tooltips sort by value for easier reading. ## Reliability improvements * **Date/time format handling** — fixes for format selection UX and regressions around forward-slash date fields (including Oracle sources). * **Query worker recovery** — workers that die abruptly no longer leave clients hanging indefinitely on incomplete responses. * **PCAP Linux SLL2** — packet captures with Linux cooked-mode v2 link-layer headers (`link_type=276`) decode correctly for FIX tick ingest. * **Fine-grained entitlements enforcement** — end-to-end entitlement checks tightened so UI-visible access matches what the query path allows. ## Compatibility There are no breaking changes to the client libraries (Python, REST, JavaScript, Excel, etc.) or to the dataset query surface. Existing code and queries continue to work unchanged. New capabilities in 2.3 are additive through the Liberator UI, connection types, and AI assistant surfaces.