@cube-dev/platform-client 0.22.0 → 0.23.0

package/CHANGELOG.md

@@ -9,6 +9,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 <!-- Curate public-facing entries here before the next release. -->
 
+## [0.23.0] - 2026-10-22
+
+### Added
+
+- `outputPath` and `namePrefix` on `StartDbtSyncInput` (`POST /api/v1/deployments/{deploymentId}/dbt-sync`) override the deployment's saved Output path and Name prefix for one sync, so several dbt projects can sync into one deployment, each into its own folder with its own prefix.
+
+### Changed
+
+- A dbt sync that commits to a branch now deletes stale cube files only directly inside the folder it writes its cubes into, and leaves that folder's subfolders alone, so another project's folder nested under it survives the sync. An output path written with a trailing slash, such as `/model/cubes/dbt/`, no longer skips that cleanup: files there that no longer match a dbt model are now deleted.
+- `POST /api/v1/deployments/{deploymentId}/dashboard-exports` now requires **read** access to the dashboard's workbook and the **download data** permission on the deployment, instead of **manage** access to the workbook. **Manage** access to the workbook still passes, so callers that were allowed before keep working.
+
 ## [0.22.0] - 2026-10-20
 ### Added
 

package/dist/schema.d.ts

@@ -354,7 +354,7 @@ export interface paths {
          *
          *     Every user's AI chat threads in the deployment, most recently updated first, for analyzing how people use the agent. Filters combine with AND; a filter given several values matches any of them.
          *
-         *     Set `includeTurns=true` to get each thread's turns in the same response. A turn is one prompt and everything the agent did to answer it: the final answer, the tools it called, and each query it ran with its SQL, status, row count and any thumbs-up/down feedback. Query result rows are never included. A prompt, answer or SQL longer than 20,000 characters, or an error longer than 2,000, is cut and ends in `…`. A page holds up to 200 threads, or 50 with turns; a larger `first` is lowered to that.
+         *     Set `includeTurns=true` to get each thread's turns in the same response. A turn is one prompt and everything the agent did to answer it: the final answer, the tools it called, and each query it ran with its SQL, status, row count and any thumbs-up/down feedback. A prompt, answer or SQL longer than 20,000 characters, or an error longer than 2,000, is cut and ends in `…`. Each turn's steps and each query's result rows come only with a single thread. A page holds up to 200 threads, or 50 with turns; a larger `first` is lowered to that.
          *
          *     For an incremental export, keep the newest `updatedAt` you have seen and pass it as `updatedAfter` next time.
          *
@@ -380,7 +380,7 @@ export interface paths {
          * Get a chat thread
          * @description **🔒 Admin only.** Requires administrator privileges — the authenticated principal (API key, embed JWT, or any bearer token) must belong to a user with the admin role.
          *
-         *     One chat thread with all of its turns, oldest first. A turn is one prompt and everything the agent did to answer it: the final answer, the tools it called, and each query it ran with its SQL, status, row count and any thumbs-up/down feedback. Query result rows are never included. A prompt, answer or SQL longer than 20,000 characters, or an error longer than 2,000, is cut and ends in `…`.
+         *     One chat thread with all of its turns, oldest first. A turn is one prompt and everything the agent did to answer it: the final answer, the tools it called, and each query it ran with its SQL, status, row count and any thumbs-up/down feedback. A prompt, answer or SQL longer than 20,000 characters, or an error longer than 2,000, is cut and ends in `…`. Each turn also lists its `steps`: every reply the agent wrote, with its reasoning, and every tool it called, with its input, in the order they began; their texts are cut the same way. Each query carries up to 100 of its result `rows`.
          *
          *     Chat history export is in preview: when it is not enabled for the account, `404` is returned.
          */
@@ -446,7 +446,7 @@ export interface paths {
          *
          *     Pass `layout: print` with `format: pdf` for the paginated A4 document the dashboard's "Download print version" menu item produces — one column, every table row, the filter state as a summary line — instead of the default `screenshot` of the board. `print` with `format: png` returns `400`.
          *
-         *     By default the dashboard renders under the calling user's own security context, so the file shows what that user would see. Requires **manage** access to the dashboard's workbook — the same permission the console's own "Download as PNG / PDF" actions require, because a whole-dashboard render is a bulk data export.
+         *     By default the dashboard renders under the calling user's own security context, so the file shows what that user would see. Requires **read** access to the dashboard's workbook and the **download data** permission on the deployment — the same rule the console's own "Download as PNG / PDF" actions follow. **Manage** access to the workbook also passes, as it did before the download data permission was required.
          *
          *     To export on someone else's behalf, pass `renderAs`: the queries behind the image then run under that user's security context instead of yours. **Requires the caller to be a tenant administrator.** Name a console user by `userId` or `email` (`type: USER`), or an embed user by `embedTenantName` + `externalId` (`type: EMBED_USER`). A deactivated console user is not a valid subject and returns `404`; an email is matched case-insensitively.
          *
@@ -4285,6 +4285,8 @@ export interface components {
         /** @enum {string} */
         BulkUserActionResponseAction: "DEACTIVATE" | "ACTIVATE" | "DELETE";
         ChatHistoryQuery: {
+            /** @description The column name of each `rows` cell. Returned only when reading a single thread. Null when `rows` is. */
+            columns?: string[] | null;
             /** @description The data source a `runSql` query ran on. */
             dataSource?: string | null;
             error?: string | null;
@@ -4292,8 +4294,10 @@ export interface components {
             feedback?: components["schemas"]["ChatHistoryQueryFeedback"] | null;
             /** @description The tool-call message id. */
             messageId: string;
-            /** @description Rows the query returned. The rows themselves are never included. */
+            /** @description How many rows the query returned. */
             rowCount?: number | null;
+            /** @description The result rows the agent saw, at most 100; `rowCount` says how many the query returned. Returned only when reading a single thread. Null when the query returned no result, as when it failed or was stopped. */
+            rows?: any[][] | null;
             /** @description The SQL the agent generated. */
             sql?: string | null;
             /** @description `incomplete` when the query never returned: the turn was stopped or failed. */
@@ -4310,6 +4314,28 @@ export interface components {
         };
         /** @enum {string} */
         ChatHistoryQueryStatus: "success" | "error" | "incomplete";
+        ChatHistoryStep: {
+            /** @description Tool steps: the arguments the agent passed, as JSON. */
+            input?: string | null;
+            /** @description The message id; a query's step shares its `messageId`. */
+            messageId: string;
+            /** @description Model steps: the model's reasoning, as the model reported it; some models report only a summary, and a stopped step may hold only its start. Null when it reported none. */
+            reasoning?: string | null;
+            /** @description When the step began. Null for steps recorded before Cube stored step times. */
+            startedAt?: string | null;
+            /** @description Tool steps: `incomplete` when the tool never returned. */
+            status?: components["schemas"]["ChatHistoryStepStatus"] | null;
+            /** @description Model steps: what the agent wrote, such as a note before calling tools or its answer. */
+            text?: string | null;
+            /** @description Tool steps: the tool the agent called. */
+            tool?: string | null;
+            /** @description `model` for a reply the agent wrote, `tool` for a tool it called. */
+            type: components["schemas"]["ChatHistoryStepType"];
+        };
+        /** @enum {string} */
+        ChatHistoryStepStatus: "success" | "error" | "incomplete";
+        /** @enum {string} */
+        ChatHistoryStepType: "model" | "tool";
         ChatHistoryThread: {
             /** @description Null for the Auto agent. */
             agentId?: number | null;
@@ -4355,6 +4381,8 @@ export interface components {
             queries: components["schemas"]["ChatHistoryQuery"][];
             /** @description When the user sent the prompt. Null for turns recorded before Cube stored turn times. */
             startedAt?: string | null;
+            /** @description Everything the agent did in the turn, in the order it began: each reply with its reasoning, and each tool call. When any step has no `startedAt`, the steps are in the order they last changed instead. Returned only when reading a single thread. */
+            steps?: components["schemas"]["ChatHistoryStep"][] | null;
             /** @description Every tool the agent called in the turn, in first-call order. */
             tools: string[];
         };
@@ -7156,8 +7184,12 @@ export interface components {
             manifest?: {
                 [key: string]: any;
             } | null;
+            /** @description Prefix for this sync’s cube names and file names, e.g. `sales_`. Overrides the Name prefix saved in the deployment’s dbt settings for this sync only; an empty string means no prefix. Cube names must be unique across the data model, so give each dbt project its own prefix when two projects can share a model name. */
+            namePrefix?: string | null;
             /** @description What to do with the generated cubes. `branch` (the default) commits them to a new Cube branch for review. `files` commits nothing and creates no branch — the files are returned from `GET /dbt-sync/{syncJobId}/generated-files` instead, so your own CI can commit them to your repository through its normal review. Independent of `source`. */
             output?: components["schemas"]["StartDbtSyncInputOutput"] | null;
+            /** @description Repository folder for this sync’s generated cube files, e.g. `model/cubes/dbt/sales` (a leading `/` is optional). Overrides the Output path saved in the deployment’s dbt settings for this sync only. A sync replaces the files directly inside this folder and leaves its subfolders alone, so syncing several dbt projects into one deployment works when each gets its own folder and its own `namePrefix`. */
+            outputPath?: string | null;
             /** @description Git ref in the dbt repository to sync from — a branch or a tag, NOT a commit SHA. Overrides the branch saved on the deployment’s dbt git integration for this sync only, so CI can sync the ref under review (e.g. a pull request’s head branch) instead of the tracked branch. The generated Cube branch is named after it. */
             ref?: string | null;
             /** @description Where the dbt metadata comes from. `git` (the default) clones the deployment’s dbt repository and runs dbt to produce a manifest. `manifest` converts a manifest you supply in `manifest` below, so Cube never needs access to the dbt repository — and the sync takes seconds rather than minutes, because no sandbox is provisioned and no dbt command runs. */