> For the complete documentation index, see [llms.txt](https://docs.zenlytic.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zenlytic.com/changelog/readme.md). # Product updates {% updates format="full" %} {% update date="2026-08-16" tags="improvements,fixes" %} ## Claude Opus retired from the model picker Claude Opus is no longer offered for new chats, alongside more relevant field search, more reliable chat titles, and restored access to Workspace Settings. ### Improvements * **Claude Opus retired from the model picker** — Opus models are no longer available for new chats. Existing conversations and Proactive Agents pinned to Opus continue to work, and Sonnet 5 is the system default. * **More relevant field search** — Zoë's field search drops zero-relevance matches and duplicates, so results better reflect your model. * **More reliable chat titles and citations** — Titles are generated from your messages, and citation generation is bounded to reduce failures and off-topic titles. * **More usable MCP connections table** — The MCP section of settings has clearer column widths and horizontal scrolling for wide tables. ### Bug fixes * **Workspace Settings and chat admin reachable again** — Workspace Settings, account settings, All Chats, and feedback are available from navigation again. {% endupdate %} {% update date="2026-08-09" tags="new-features,improvements,fixes" %} ## Attach artifacts from chat Attach saved artifacts directly in chat, scoped database-tree search, artifact folder polish, and fixes across deep links, deletion dialogs, chat filters, and Slack installs. ### New features * **Attach artifacts from chat** — Search for and attach saved artifacts directly in the chat composer, with previews and thumbnails. * **Filter chat history by source** — Filter your chat list by where a conversation originated. * **Default MCP connectors in Slack chats** — New Slack conversations automatically use connectors marked enabled by default, and show configuration errors in the thread. (Slack) ### Improvements * **Scoped database-tree search** — Limit search to selected databases in Context Manager, with better pagination and tree state. (Context Manager) * **GPT-5.6 Luna replaces GPT-5.5** — GPT-5.6 Luna is now the highlighted OpenAI model in the model picker. * **Improved artifact folders** — Folder breadcrumbs, scoped table navigation, and better drawer behavior for organizing artifacts. * **Human attribution on Zoë model commits** — Zoë-authored data model commits now record the initiating human as co-author. (Context Manager) ### Bug fixes * **Deep links open the right workspace** — Resource links now open in the workspace that actually owns the resource. * **Safer, consistent deletion dialogs** — Standardized confirmation dialogs and fixed loading and error handling when deleting agents, folders, artifacts, schedules, and credentials. * **Block conflicting Slack installations** — A second Zenlytic workspace can no longer silently take over an already-connected Slack workspace; you'll see a clear error instead. (Slack) * **Teradata pagination fixed** — Table browsing and previews use Teradata-compatible pagination. (Context Manager) * **Clear Git permission errors** — Save Model now explains revoked deploy keys and credential denials instead of a generic failure. (Context Manager) * **Mobile chat filters fixed** — Chat source-filter menus position, close, scroll, and respond to Escape correctly on mobile. * **Artifacts wait for chat completion** — The artifact drawer no longer opens prematurely while a chat is still streaming. * **Signup name fixed** — New signups no longer fail due to a missing required name field. {% endupdate %} {% update date="2026-08-02" tags="new-features,improvements,fixes" %} ## Claude Opus 5 and bulk user management Claude Opus 5 in the model picker, bulk workspace user management, and fixes across chat, deliveries, and Proactive Agents. ### New features * **Claude Opus 5** — Opus 5 is available in the chat model picker; Opus 4.8 remains available. * **Bulk user management** — Admins can select multiple workspace users and change their roles or remove them in one action. ### Improvements * **Verified-field styling** — Modeled fields use a verified style across queries and dashboards, making them easier to distinguish from dynamic fields. * **Artifact links open the exact version** — Links to an artifact version open the right conversation, select the matching version, and scroll it into view. * **Background indexing control** — Admins can turn off scheduled background search-index queries for a workspace. * **Microphone beside Send** — The voice transcription button now sits next to the send button in chat. ### Bug fixes * **Markdown tables render in delivery emails** — Scheduled-delivery emails convert Markdown tables, links, bold, and italic text into styled HTML. * **Chat failures end cleanly** — Provider errors, oversized inputs, interrupted streams, and failed agent or MCP connection setup now show a clear error instead of leaving the chat hanging. * **Files preserved during agent runs** — Generated and uploaded files no longer disappear during Proactive Agent runs. * **Proactive Agent feedback and filtering** — Thumbs-down feedback and status filtering work correctly in the Proactive Agents list. * **Consistent workspace connection creation** — Connections created from Workspace Manager apply the same credential defaults and validation as Workspace Settings. * **Memory setting fully respected** — When a workspace turns memory off, the memory button, memory creation, and memory retrieval are all disabled. * **Git provider errors shown on model save** — When a Git provider rejects a data model save, Context Manager shows the provider's error instead of a generic failure. (Context Manager) * **Model-editing toggle hidden without permission** — Proactive Agent configuration no longer shows the data-model editing tool toggle to users without model-edit permission. This was a display-only fix — the underlying capability was already permission-gated. * **Clearer SQL-permission errors** — Zoë's permission-emulation error names the exact setting involved and explains the effect of disabling it. * **Web search hidden when disabled** — Proactive Agent configuration no longer offers web search when the workspace has disabled it. {% endupdate %} {% update date="2026-07-26" tags="new-features,improvements,fixes" %} ## Proactive Agents for all workspaces Proactive Agents are now available to all workspaces, plus refreshed signup screens, HTML uploads, and chat and query fixes. ### New features * **Proactive Agents for all workspaces** — Schedule agents that pull context from your MCP connections and data sources, decide whether a result is worth sending, and alert you by email or Slack only when something matters — not on every run. * **Create a Proactive Agent from chat** — Save a chat as a Proactive Agent using the conversation title, prompts, and attachments. * **Share Proactive Agents** — Share agents with users or groups and assign viewer, editor, or owner access. * **Cancel running agent runs** — Cancel an active Proactive Agent run directly from run history. * **MCP connections enabled by default for new chats** — Admins can configure an MCP connection to start enabled in each new conversation for users who have access. * **HTML file uploads** — Upload .html files as context in chat and Proactive Agents. ### Improvements * **Refreshed signed-out and signup screens** — Signup, workspace invitation, and unauthenticated pages use a cleaner, consistent account flow, and campaign or partner signup links can show tailored copy. * **More tables in chat answers** — Zoë now uses compact Markdown tables more often when they communicate query results more clearly. ### Bug fixes * **Welcome-message edits save reliably** — Editing a workspace's chat welcome message enables Save immediately. * **Wide SQL queries no longer time out** — Very wide queries, including large SELECT \* results, skip a detection step that could previously delay or fail responses. {% endupdate %} {% update date="2026-07-19" tags="improvements,fixes" %} ## Refreshed signup flow and smarter documentation search A refreshed signup flow, smarter Zoë documentation searches, and fixes across artifact sharing, Slack deliveries, and warehouse queries. ### Improvements * **Updated signup flow** — New-user signup has a revised email/password flow, clearer terms links, and moves profile personalization to after signup. * **Fewer repeated Zoë documentation searches** — Zoë stops searching product docs once it has an answer, instead of repeating searches. ### Bug fixes * **Fixed Microsoft Teams sign-in** — Teams users now get the Entra login flow and no longer see a blank screen after signing out. * **Fixed case-sensitive warehouse queries** — Queries no longer lowercase quoted mixed-case identifiers on Fabric and Synapse, which caused failures on case-sensitive warehouses. * **Accurate upload-type validation** — The upload picker now matches supported file types, adds .webp and .ppt uploads, removes unsupported RTF, XML, and GIF options, and shows readable errors. * **Web search preserved across run steps** — Follow-up steps in a run keep the web-search setting instead of dropping it. * **Artifact sharing dialogs stay open** — Sharing and access-management popovers no longer close prematurely, and focus and scrolling behave correctly. * **Slack respects workspace web-search policy** — Slack conversations now honor whether web search is enabled for the workspace. (Slack) * **Longer BigQuery Data Project IDs** — The Data Project ID(s) field accepts the full 256 characters. * **No duplicate artifact refreshes** — Repeated scheduling actions no longer create duplicate refresh entries. * **Cleaner new-chat drawers** — Explore and context drawers initialize and clear correctly for deep-linked questions, without leaking prior-chat state. {% endupdate %} {% update date="2026-07-12" tags="new-features,improvements,fixes" %} ## Zoë looks up product docs before answering Workspace switching from Add Data, Zoë checking product docs before answering how-to questions, a refreshed login and signup experience, and fixes across MCP connection errors, data model reviews, daily refresh scheduling, SQL generation, and large-workspace exports. ### New features * **Zoë looks up product docs** — Zoë checks Zenlytic's product documentation before answering how-to questions, so answers reflect what's actually documented. (Slack, docs) ### Improvements * **Switch workspaces from Add Data** — Users with access to multiple workspaces can now switch workspaces directly from the Add Data flow. * **Refreshed login and signup pages** — Login and signup pages now show an updated brand panel with customer logos and security badges. * **Admin chat filter for Proactive Agent runs** — Admin chat lists can now include or exclude Proactive Agent run conversations when filtering. ### Bug fixes * **Clearer MCP connection errors in chat** — When an MCP connection is stale or misconfigured, chat now shows a clear error telling you to refresh the connection instead of hanging or retrying silently. * **Data model reviews skip deprecated Topics** — Data model review guidance no longer recommends legacy Topics unless the project actually contains topic files. (Data Model Editor) * **Daily refresh no longer blocks chat** — Background daily refresh jobs are now capped so they don't crowd out chat tasks, reducing queued or expired requests. * **Fixed duplicate-column SQL errors** — Access-filter SQL generation no longer produces ambiguous-column errors from duplicate quoted or cased references to the same column. (Data Model Editor) * **Fixed a Claude runtime error** — Resolved a runtime error that could affect some Claude-powered chat and agent runs. * **More reliable large-workspace exports** — Screenshot, PDF, and CSV rendering for large dashboards and questions is more reliable and less likely to crash. {% endupdate %} {% update date="2026-07-01" tags="new-features,improvements" %} ## Artifact folders are generally available Artifact folders are now generally available, along with table-selection, skill-error, and skill-metadata improvements. ### New features * **Organize artifacts into folders** — Browse artifacts in table or gallery view, organize them into workspace folders, and manage folder-based access. ### Improvements * **More visible table selection** — Row-selection checkboxes now have stronger contrast and stay visible on mobile gallery cards. * **Clearer skill and code error messages** — Zoë now tells you when a failure comes from a workspace-authored skill or uploaded code, including the skill or file and the actual error. (Data Model Editor) * **Git-backed skill metadata** — Git-backed workspace skills now show their name, description, required status, and file location in Zoë's skills list. (Data Model Editor) {% endupdate %} {% update date="2026-06-30" tags="new-features,improvements" %} ## Claude Sonnet 5 in the model picker Claude Sonnet 5 in the model picker, plus clearer MCP connection error messages. ### New features * **Claude Sonnet 5** — Sonnet 5 is now the default active model in the chat model picker; existing chats are unaffected. ### Improvements * **Clearer MCP setup errors** — MCP server connection and setup failures now show specific, useful error messages instead of a generic failure. {% endupdate %} {% update date="2026-06-26" tags="improvements,fixes" %} ## Expanded file uploads and smarter warehouse location handling Expanded file upload support, smarter warehouse location handling, and SSO and MCP fixes. ### Improvements * **Expanded file uploads** — Upload macro-enabled Excel files (.xlsm) and structured text and code files (.py, .json, .md) across chat, agents, workflows, document preview, and file analysis. * **Warehouse location handling** — Zoë treats "schema," "database," "catalog," and "dataset" more flexibly and retries alternate warehouse hierarchy levels before reporting data as unavailable. ### Bug fixes * **Vanity-host SSO redirects** — Vanity-domain auth redirects work correctly for EU and other non-app.zenlytic.com environments. * **MCP tool discovery** — MCP server tool discovery completes the initialization handshake and parses SSE responses correctly, fixing cases where connected MCP servers showed no tools. {% endupdate %} {% update date="2026-06-25" tags="improvements,fixes" %} ## Shared-chat and MCP connection fixes Shared-chat copy and MCP connection fixes. ### Improvements * **Shared chat copy navigation** — Copying a shared chat now opens the copied conversation through in-app navigation. ### Bug fixes * **MCP protocol version header** — MCP connections send the negotiated MCP-Protocol-Version header after initialization, fixing tool calls against servers that require it. (API) {% endupdate %} {% update date="2026-06-24" tags="new-features,improvements,fixes" %} ## Source-data citations on dashboards Source-data citations on dashboards, Context Manager in chat, plus upload and artifact fixes. ### New features * **Source-data citations for dashboards** — On request, dashboard visuals can show source-data badges, an underlying-rows view, and CSV downloads. ### Improvements * **Context Manager in chat** — Context Manager now opens in a chat-side drawer by default, so you stay in the conversation. ### Bug fixes * **Managed MotherDuck uploads** — Stale MotherDuck secrets are cleared automatically, so managed database uploads stop failing as credentials accumulate. * **Artifact thumbnails in sandboxes** — Thumbnails now generate correctly for artifacts created in agent sandboxes. {% endupdate %} {% update date="2026-06-23" tags="new-features,improvements,fixes" %} ## Context Manager reaches more surfaces Context manager reaches more surfaces, plus sandbox and chat-accuracy fixes. ### New features * **Context manager across more surfaces** — Context manager now works in chat, the Data Model Editor, and artifact citation and editing. ### Improvements * **Numeric answer guardrails** — Stronger guidance against reporting numbers that didn't come from tool results. ### Bug fixes * **Schema visibility in context** — Schemas that weren't appearing in context selection now show up. (Data Model Editor) * **Sandbox timeouts** — Sandbox soft timeouts now surface instead of being swallowed, so chats don't hang. * **Nested sandbox restore** — Nested directories are restored correctly during sandbox file restore. {% endupdate %} {% update date="2026-06-19" tags="new-features,fixes" %} ## Org-admin management and SQL interpretability Org-admin management and an optional SQL interpretability toggle. ### New features * **Organization admin management** — Workspace Manager now lets org admins view and manage other organization admins. * **SQL interpretability toggle** — New workspace setting to turn off SQL interpretability in chat. ### Bug fixes * **Discover fixes** — Resolves runtime issues in Discover. {% endupdate %} {% update date="2026-06-17" tags="improvements,fixes" %} ## Scheduling, picker, and permissions improvements Scheduling, picker, and permissions improvements, plus chat and SSO fixes. ### Improvements * **Model and database pagination** — Large model and database lists page in instead of loading all at once. (Data Model Editor) * **Larger scheduled artifact modal** — More room in the schedule-artifact modal. * **Restricted-permission messaging** — Clearer message for users with restricted permissions instead of a blank or failed state. ### Bug fixes * **Deleted chats filter** — Deleted conversations now appear under All Chats → Deleted. * **Sandbox file re-uploads** — Re-uploading a file with the same name overwrites the old sandbox copy instead of keeping stale content. * **SSO group assignment** — Re-adding SSO users to groups no longer creates duplicate or errored adds. {% endupdate %} {% update date="2026-06-16" tags="new-features" %} ## Try Zenlytic with sample data A quicker way to try Zenlytic with sample data. ### New features * **Demo datasets in self-serve** — Self-service setup now offers demo datasets to start from. {% endupdate %} {% update date="2026-06-15" tags="improvements,fixes" %} ## Connection setup polish Connection setup polish. ### Improvements * **Trino HTTP scheme field** — Clearer placeholder text for the Trino HTTP scheme field. ### Bug fixes * **BigQuery dataset field** — Fixes the dataset textbox in the BigQuery browse step. (Data Model Editor) {% endupdate %} {% update date="2026-06-12" tags="new-features,improvements,fixes" %} ## Live editing and onboarding updates Live editing, onboarding, sharing, and connection fixes. ### New features * **Select/inspect toggle in the HTML editor** — Switch between select and inspect modes in the live HTML editor. * **Department in onboarding** — Onboarding now collects the user's department. ### Improvements * **Connection name validation** — Tighter validation for connection names. * **Artifact save/present behavior** — Clearer guidance on when artifacts are presented versus saved. * **Regional processing routing** — Workspace processing routes to the matching Vertex region. (API) ### Bug fixes * **Shared artifact file access** — Files shared through an artifact are now accessible via that artifact. * **Delete group with shared artifacts** — Deleting a group that has shared artifacts now works. * **File-link rewriting** — Removes incorrect /mnt/data signed-URL rewriting in conversation display and citations. * **Fabric quoting** — Fixes quoting issues in Fabric connections. (Data Model Editor) {% endupdate %} {% update date="2026-06-10" tags="improvements,fixes" %} ## Recent Updates and citations redesign Recent Updates and citations redesign, plus several display fixes. ### Improvements * **Recent Updates redesign** — A reworked Recent Updates experience. * **Collapsed citations** — Citations collapse into a single, cleaner display. ### Bug fixes * **Artifact popups** — Artifact iframe popups can open as normal top-level windows. * **Uploaded image previews** — Fixes broken-image display for images uploaded in chat. * **Workflow final state** — Fixes a race where a workflow run could show the wrong final state. * **Remove all users** — Fixes the remove-all-users option in group and user management. * **Managed table creation** — Prevents query races right after a managed database table is created. (Data Model Editor) * **Duplicate table error** — Clearer error when YAML defines duplicate tables. (Data Model Editor) {% endupdate %} {% update date="2026-06-09" tags="new-features,improvements,fixes" %} ## MCP, model, and reliability updates A large batch of MCP, model, and reliability updates. ### New features * **MCP for proactive agents** — Proactive agents can now use MCP connections. * **Font uploads in skills** — Skills accept .ttf and .otf font uploads. ### Improvements * **Chat starts responding sooner** — Reduced time to first token when a chat begins. * **GPT model lineup** — Updated GPT model choices. * **MCP tool descriptions** — Clearer connection context in MCP tool descriptions. * **MCP tool-selection help** — Added tooltip help for MCP tool selection. * **Stale sandbox warning** — The agent is warned when sandbox context may be stale. * **Scheduled report context** — Clearer scheduled-report instructions for the agent. * **Scrollable errors and warnings** — Long error and warning panels are now scrollable. (Data Model Editor) * **API conversation tagging** — Conversations created through the API are tagged source=api. (API) ### Bug fixes * **Oversized prompts** — Oversized inputs fail cleanly instead of retrying without effect. * **MCP connection selector** — Chat MCP connection options use the correct non-admin endpoint. * **Clarity setup routing** — Clarity workspaces route through the setup-workspace flow. * **SQL Server casing** — Fixes table-name normalization for SQL Server while preserving Synapse behavior. (Data Model Editor) * **Context manager tooltip** — Fixes tooltip positioning in the context manager. (Data Model Editor) * **Artifact CSV recovery** — Missing CSVs are regenerated before an artifact rebuilds. {% endupdate %} {% update date="2026-06-04" tags="new-features,improvements,fixes" %} ## Connectors settings redesign Connectors settings redesign, BigQuery improvements, and context manager rework. ### New features * **Connectors settings page** — A redesigned connectors and MCP connections settings page. * **MCP access management** — Access-management controls in the MCP modal. * **BigQuery project selector** — Choose a project when browsing BigQuery data. (Data Model Editor) * **Fabric service-principal auth** — Fabric connections support Microsoft Entra service-principal authentication. ### Improvements * **Context manager redesign** — A reworked context manager experience. (Data Model Editor) * **BigQuery multi-project** — Handles multiple BigQuery project IDs when browsing data. (Data Model Editor) * **Branch selector limit** — Caps branch selector choices so large repos stay usable. (Data Model Editor) * **Agent list pagination** — Agent lists paginate so they load more reliably. * **Scheduled artifact copy** — Updated scheduled-artifact and proactive-agent descriptions. ### Bug fixes * **Long chat titles** — Retries or truncates chat-title generation when a conversation exceeds context limits. {% endupdate %} {% update date="2026-06-03" tags="new-features,improvements,fixes" %} ## Opus 4 and scheduled-delivery run history Opus 4, scheduled-delivery run history, and context editing settings. ### New features * **Claude Opus 4** — Opus 4 is available as a model option. * **Scheduled delivery run history** — A run-history tab for scheduled artifact deliveries. * **Context editing setting** — A settings-page toggle for context editing. ### Improvements * **Email attachment limits** — Clearer maximum attachment size in scheduled-delivery emails. * **Proactive agent test runs** — Better storage and display of proactive agent test runs. ### Bug fixes * **Artifact citations** — Restores citation generation for artifacts. {% endupdate %} {% endupdates %} --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.zenlytic.com/changelog/readme.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.