# Single Sign-On (SSO) Source: https://docs.push.ai/admin/single-sign-on Configure SSO authentication for your Push.ai organization # Single Sign-On Integration Push.ai supports using Identity Providers (IdP) such as Okta to control authentication through single sign-on (SSO). We also support provisioning and deprovisioning Push.ai users along with managing their role assignments through SCIM (directory sync). ## Setting up SSO Push.ai supports the SAML protocol for single sign-on authentication. To initiate setup, contact our support team at [support@push.ai](mailto:support@push.ai). We'll provide specific instructions and a test environment for setting up Push.ai with your preferred IdP. ### Post-Integration Changes After completing the SSO integration: * Users will no longer be able to log in with their previous Push.ai authentication method (username & password or Google) * Attempting to use previous authentication methods will redirect users to the SSO login page * Users will need to input their email, which will use the domain to redirect to your IdP for sign-in * Users can log in directly from your IdP by clicking on the Push.ai application integration **Email Requirements** The primary email of each user assigned to Push.ai in your IdP must match their Push.ai user's email. Otherwise, they won't be able to log in to their Push.ai account. ## SCIM Configuration ### Overview SCIM (System for Cross-domain Identity Management) enables you to: * Create and delete Push.ai users directly within your IdP * Update user information (including email) * Manage Push.ai role assignments ### Required Attributes In addition to standard attributes (`id`, `emails`, `first_name`, `last_name`), Push.ai requires the following custom attribute: `roles` → `` Examples: `Admin`, `Data`, or `NormalUser` **Important Notes about Roles** - Role assignments must be non-empty - SCIM-enabled organizations do not have default roles - All role assignments must be explicitly managed through your IdP - Role names are case-insensitive ### Initial Synchronization During initial SCIM setup: 1. We match IdP users with existing Push.ai users by primary email 2. Matched users will be linked to their IdP identity 3. Future updates (including email changes) will sync automatically 4. New IdP users without matching Push.ai accounts will have accounts provisioned automatically ### User Management * User assignments in your IdP automatically sync to Push.ai in real-time * Creating, editing, and deleting users must be done through your IdP * Direct user management in Push.ai is restricted for SSO-enabled organizations ### Role Management * Creating and editing roles must still be done within Push.ai by Admins * Role assignments to users must be managed in your IdP using the `roles` attribute * Available roles include: Admin, Data, and Normal User For detailed setup instructions specific to your IdP or additional support, please contact our support team at [support@push.ai](mailto:support@push.ai) # Users and Roles Source: https://docs.push.ai/admin/users-and-roles Manage users and roles within your Push.ai organization ## Adding Users Users can be added to your organization through an invitation flow or automatically according to their email domain. All users' email addresses are verified during user creation. ### Automatic Assignment Any user who signs up with an email domain that matches the organization's domain will be automatically added to their organization. Common domains such as gmail, hotmail, etc. are excluded from this policy. ### Invitation Any user who is invited via an invite flow will be added to the organization regardless of their email domain. This allows you to include team members or external collaborators who may not share your organization's email domain. ## Roles and Permissions Push.ai offers different user roles with varying levels of permissions to help manage team members' access levels and responsibilities. Here's an overview of each role: ### Admin Role The admin role is the highest level of authority within Push.ai, possessing comprehensive access and control over all settings and functions. * Can manage integrations, members, organization, and billing in the settings page * Modify team settings, including billing and member roles * Access to all Data Role and Normal User permissions ### Data Role * Can create and edit metrics and metadata * Access to all Normal User permissions ### Normal User * Can view metrics and subscribe to them * Can manage individual user preferences # Create a run Source: https://docs.push.ai/api-reference/endpoint/create-run POST /agents/{agent_id}/runs Invoke a published agent asynchronously. Returns a run ID that you can poll for status and results. Pass `?dry_run=true` to preview the compiled prompt without executing. # Get run status Source: https://docs.push.ai/api-reference/endpoint/get-run-status GET /agents/{agent_id}/runs/{run_id} Poll for the status and result of an agent run. Returns the output when the run succeeds, or an error message if it fails. # List agents Source: https://docs.push.ai/api-reference/endpoint/list-agents GET /agents Returns all published agents for your organization, with their accepted parameters. # API Reference Source: https://docs.push.ai/api-reference/introduction Programmatic access to Push AI agents The Push AI API lets you list your published agents, invoke them with parameters, and poll for results — all from your own code, scripts, or backend services. ## Base URL ``` https://api.push.ai/api/v1 ``` ## Authentication All endpoints require an API key passed in the `x-api-key` header. ```bash theme={null} curl https://api.push.ai/api/v1/agents \ -H "x-api-key: YOUR_API_KEY" ``` API keys are scoped to your organization. Every key you create can access all published agents in your org. ### Creating an API key API key management is only available to **Admin** users. If you don't see the API Key option in Settings, contact your organization's admin. 1. Open **Settings → API Key** from the left sidebar in the Push AI dashboard 2. Click **Create Key** 3. Copy the key immediately — it is only shown once API Key settings page ### Regenerating or revoking a key From the same **Settings → API Key** page, you can: * **Regenerate** — creates a new key and invalidates the old one * **Revoke** — permanently disables the key Revoked or regenerated keys return `401` immediately. Make sure to update any scripts or integrations using the old key. ## Errors All errors return a consistent JSON envelope: ```json theme={null} { "error": { "code": "agent_not_found", "message": "Agent 123 not found or not published.", "request_id": "req_abc123" } } ``` | Field | Description | | ------------ | -------------------------------------------------------------------------------------------------------------------- | | `code` | Machine-readable error code for programmatic handling. | | `message` | Human-readable description. | | `request_id` | Unique ID for this request. Include it when contacting support. Also returned in the `X-Request-Id` response header. | ### Common error codes | Code | Status | Meaning | | ------------------------- | ------ | ------------------------------------------------------------------------ | | `missing_api_key` | 401 | No `x-api-key` header provided. | | `invalid_api_key` | 401 | Key is invalid or has been revoked. | | `agent_not_found` | 404 | Agent doesn't exist, isn't published, or belongs to another org. | | `run_not_found` | 404 | Run ID not found or belongs to another org. | | `unknown_parameter` | 400 | Parameter name doesn't match any defined on the agent. | | `invalid_parameter_value` | 400 | The value provided for a parameter doesn't match any known option. | | `internal_error` | 500 | Unexpected server error. Retry or contact support with the `request_id`. | ## Rate limits Rate limiting is enforced per API key. Current limits are generous for normal usage. If you hit a limit, you'll receive a `429` response — back off and retry. # Use Cases Source: https://docs.push.ai/business-graph/business-graph Populate your Business Graph™ to improve the relevancy of Push.ai's AI systems. By defining Business Objects, you've already established a strong foundation of data context. The Business Graph™ is used to map additional context to the data, enhancing the relevancy of AI systems. Over time, we'll expand functionality to include and integrate additional sources of business context. Below is a list of the capabilities available today. ## Data Context ### Metric Relationships Define relationships between metrics to include additional context and leverage additional analaysis capabilities. Metric relationships are defined through input and output metrics, in what is commonly known as a metric tree. ### Metric-Dimension Relationships **Example:** If a user is Subscribed to a metric such as Revenue, Metric-Dimension relationships allow our systems to pull in related context, like the Dimensions we can use to slice Revenue (e.g., Revenue by Product, Revenue by Country, etc.) Metric-Dimension relationships are generated as a result of defining Business Objects like metrics and dimensions. They serve a crucial role in the Business Graph™, allowing our systems to pull in context from related Business Objects. ## Business Context ### Metadata for Business Objects Attach metdata to Business Objects, such as metrics, dimensions, and artifacts. For metrics and dimensions defined in Push.ai, you can use descriptions, labels and owners to attach metadata and organize resources. Artifacts can include rich metadata such as document types, authors, and creation dates. Semantic layer integrations offer additional metadata capabilities that can be integrated with Push.ai. ### Goals **Note:** If you'd like to input a large list of goals or would like to define goals at the dimension level, you can [reach out to us](mailto:support@push.ai) and someone from our team can help you get set up. Define goals to keep teams accountable and identify where teams are on or off track. Goals can be defined on metrics at a monthly time grain in the UI. ### Organizational Context Organizational context is an open text field that can be configured in [Organization Settings](https://app.push.ai/settings). This allows you to add additional context around your organization that can be used to enhance the relevancy of Push.ai's AI systems. # How It Works? Source: https://docs.push.ai/business-graph/how-it-works Map data with business context using a graph-based model of your business. Business Graph The Business Graph™ is a graph-based representation of your business that integrates data with business context. You can think of the Business Graph™ as a model of your business which encompasses: * the data and measurement system used to measure success * the people in the business and their individual roles in driving success * the projects and goals that the business is working to achieve ## Data Context The Business Graph™ allows you to model your data beyond tables and columns. Through centrally defined metrics and dimensions, you can explicitly define a measurement system for sucess. For example, by defining metric-relationships and dimension-relationships, you gain visibility into how different parts of the business relate to each other. ## Business Context The Business Graph™ increases the knowledge of your organization by capturing and organizing information outside your measurement system. Conversations, documents and other sources of unstructured data can be mapped back to the individual Business Objects in your Business Graph™. This is valuable in two major ways: * improves the ability for organizations to share information and collaborate around the correct actions * improves the relevancy of AI systems using context not available in the data directly For example, a marketing team might be discussing a spike in performance and determine that the increase was due to a cultural holiday within a specific region. This information, whether it be in a Slack message, weekly business review, or logged in Push.ai can be mapped back to the Business Object for that region and referenced by others exploring that same region and in AI systems. # Artifacts Source: https://docs.push.ai/business-objects/artifacts Artifacts are one of the core primitives of knowledge in Push.ai. They represent various types of content and provide valuable context that enhances strcutured data analysis. ## Features * Process and analyze content from multiple sources including documents, chats, meeting minutes and more * Automatically categorize and tag content for better organization * Make content searchable and accessible across your organization ## Artifact Categories Artifacts are organized into abstract categories that help classify different types of content: | Category | Description | | ---------------- | -------------------------------------------------------------------------- | | **DOCUMENT** | Any text-based document, such as PDFs, Word documents, or text files | | **MESSAGE** | Any chat or discussion format, including Slack channels and Teams channels | | **PRESENTATION** | Any slide or presentation format, such as PowerPoint or Google Slides | | **MEETING** | Any meeting recording or transcript | ## Integration Artifacts can be integrated from various sources: * **File Upload** - Direct file uploads to the system * **API Integration** - Integration with third-party APIs * **Warehouse Integration** - Integration with data warehouses Each integration method supports different types of artifacts and provides different capabilities for processing and analysis. # Dimension Values Source: https://docs.push.ai/business-objects/dimension-values If Dimensions are like groups, Dimension Values are the individual parts that make up each group. For example, Marketing Channel is a Dimension, while Paid Search, Direct Traffic and Email are Dimension Values. ## How Dimension Values work in Push.ai? In Push.ai, Dimension Values are first-class citizens that get automatically generated, rather than being defined by the admin or user. Using omni-search, you can quickly find any Dimension Value and explore all of its related metrics. This is especially valuable for data teams because it allows business users to access data and metrics specific to their area of the business without relying on a data team to build out custom views. # Dimensions Source: https://docs.push.ai/business-objects/dimensions Dimensions are categories or attributes used to organize and analyze data within a business. Dimensions help break down data into specific segments, such as time, location, or product type, allowing a business to understand and compare different aspects of their operations. ## Features * Define dimensions centrally and use them across multiple metrics. * Enabled AI for dimensions that are the primary drivers of your metrics. * Define hierarchies across dimensions to add additional context to your data. *(coming soon)*. ## ✨ AI Enabled Dimensions This feature enables Dimensions for use in AI systems. We recommend enabling this for any Dimensions considered to be the primary drivers of metric changes. These are often lower cardinality dimensions, such as Marketing Channel, Country or Product. Selecting this feature across too many dimensions or high cardinality dimensions can result in noise in AI features and create unnecessary compute on your datasource. # Metrics Source: https://docs.push.ai/business-objects/metrics Metrics are the basis for measuring your business. They are times-series aggregations of data that can be defined inside of Push.ai or integrated using semantic layers such as [the dbt Semantic Layer](https://getdbt.com), [Looker](https://cloud.google.com/looker/) and [Cube](https://cube.dev). ## Features * Define metrics in one place and use them across a variety of use-cases without the need to maintain defintions. * Track historical results, see automated projections and compare them to business goals. * Use pre-built calculations to quickly explore how metrics change over time. ## Defining Metrics Our metric builder allows teams to easily input metric definitions and dynamically generate the yml used to store them in dbt. We also include some best practice metadata so that Push.ai and other systems can better organize and deliver value using metrics. Below is the complete list of properties that are needed to make metrics. ### Definition Types * **Metric Builder** - The easiest way to define a metric. Select a table and the fields used to configure a metric. * **Custom SQL** - A flexible solution with the ability to join tables, edit and paste in SQL from other sources. * **Derived Metric** - Use existing metrics to define metrics with parent-child relationships, like average order value. ### Definition Configuration | Field | Description | Example | Required? | | :-------------------- | :---------------------------------------------------------- | :---------------- | :--------------------------- | | **Table** | The table used to calculate this metric | dim\_customers | yes (no for derived metrics) | | **Calculation Type** | The method of calculation that is applied to the expression | count\_distinct | yes | | **Calculation Field** | The expression to aggregate/calculate over | user\_id | yes | | **Date Field** | The time-based component of the metric | signup\_date | yes | | **Dimensions** | A list of dimensions to group or filter the metric by | \[plan, country] | no | | **Filters** | A list of filters to apply before calculating the metric | status = 'active' | no | ### Properties Metric properties define how a metric is calculated in Push. | Field | Description | | :------------------------- | :---------------------------------------------------------------------------------------- | | **Goal Direction** | Defines whether the business goal is to increase or decrease the metric. | | **Aggregation Properties** | Configures whether this metric can be aggregated across the date dimension. | | **Metric Format** | Configures whether a metric will display as a number, percent or currency when displayed. | ### Derived Metrics Derived metrics use existing metric definitions to calculate a new metric with a *parent-child* relationship. For example, *Profit* is a derived metric that can be calculating using the parent metrics of *Revenue* and *Cost* (`profit = revenue - costs`). Using derived metrics makes maintaining metric definitions easier by defining the supporting metrics in one definition and then having the derived metrics inherit any metric changes. With derived metrics, dimensions are inherited. The dimensions available for derived metrics are the ones shared by all related metrics ## Calculations Calculations help us explore how metrics change over time. You are likely already familiar with them if you've used traditional Business Intellignece tools. In Push.ai, calculations are used to compare metrics across time periods. Each calculation in Push.ai is associated with two time periods, a **current period** and a **previous period**. Individual calculations will be relevant depending on the use-case (e.g., you want to track something on a daily v. monthly cadence) or the nature of the business (e.g., you run an ecommerce business and order volume has a strong day-of-week seasonality).
Calculation Current Period Previous Period
Previous Day Apr 23 Apr 22
Previous Week Apr 17-23 Apr 10-16
Previous Month Apr 1-30 Mar 1-31
Previous Quarter Q1 2023 Q2 2023
Previous Year 2021 2022
Calculation Current Period Previous Period
Week to date Apr 16: Sun - Thurs Apr 9: Sun - Thurs
Month to date Apr 1-23 Mar 1-23
Quarter to date Q2 Days 1-23 Q1 Days 1-23
Year to date Y 2023 Days 1-156 Y 2022 Days 1-156
Week on Week Prior Full Week Current Week to Date
Month on Month Prior Full Month Current Week to Date
Calculation Current Period Previous Period
Trailing 7 Day Apr 15 - 22 Apr 7 - 14
Trailing 30 Day Mar 24 - Apr 23 Feb 21 - Mar 23
Calculation Current Period Previous Period
Same Day Previous Week W15: Mon W14: Mon
Same Day Previous Month January 15 February 15
Same Day Previous Quarter Q2 W1 Monday Q1 W2 Monday
Same Day Previous Year Y2023 W15 Monday Y2022 W15 Monday
Calculation Current Period Previous Period
Week to Date, Prev. Year W2 2023: Sun - Thurs W2 2024: Sun - Thurs
Month to Date, Prev. Year Apr 1-16 2023 Apr 1-16 2024
Quarter to Date, Prev. Year Q2 Days 1-23, 2023 Q2 Days 1-23, 2024
Prev. Week, Prev. Year Y 2023 W2 Y 2022 W2
Prev. Month, Prev. Year Apr 2023 Apr 2024
Prev. Quarter, Prev. Year Q1 2023 Q1 2024
Calculation Current Period Previous Period
End of Week Progress Apr 16-23 Goal EOW W15
End of Month Progress Apr 1-23 Goal EOM Apr
End of Quarter Progress Apr 1-23 Goal EOQ Q2
End of Year Progress Jan 1 - Apr 23 Goal EOY 2023
Calculation Current Period Previous Period
End of Week Pacing Projection EOW W15 Goal EOW W14
End of Month Pacing Projection EOM Apr Goal EOM Apr
End of Quarter Pacing Projection Q2 Goal Q2
End of Year Pacing Projection 2023 Goal 2023
# Product Updates Source: https://docs.push.ai/changelog/product-updates New updates and improvements. ### New Feature: AI-Assisted Model Editing * The AI agent can now suggest changes to your semantic models — adding metrics, measures, and dimensions with inline patch suggestions * Live activity timeline shows each step the agent takes when building or updating a model * Modified rows are highlighted so you can quickly review what changed before accepting suggestions * Smarter validation catches issues like duplicate measure names and invalid metric configurations before they're saved ### New Feature: Learning Center * Interactive learning modules covering key concepts: data modeling, knowledge graphs, integrations, context windows, agent skills, and more * Each module walks through a topic in digestible sections with clear explanations * Mobile-friendly layout with responsive navigation ### Charts & Visualizations * New waterfall chart type with dynamic formatting and percentage support * Time series analysis charts now lock their chart type to prevent accidental changes * Fixed-schema chart types hide irrelevant configuration controls for a cleaner editing experience ### Context & Integrations * New Context overview page showing a unified summary of all connected sources — data warehouses, integrations, knowledge, and web sources * Redesigned context sidebar with per-card scrolling and clearer organization * Agent responses now link back to the specific integration sources they used, so you can see where each answer came from * Data table previews load faster with pre-fetched source details ### Semantic Layer & Models * Top-level metric filters let you scope any metric to a specific subset of data directly from the metric configuration modal * Percentile aggregation support for measures * Derived metrics now support per-reference filters on formula tokens for more precise calculations * Cascading validation: deleting or modifying a model element now correctly flags all dependent metrics and dimensions * Models can be created without measures for dimension-only use cases ### AI Agent Improvements * Agent now tracks and displays web page sources used during analysis * Improved search with pagination and per-type result counts for more complete answers * Better handling of complex metric references and object mentions in conversation ### Onboarding * Onboarding now includes a real query step so you can test your data connection immediately after setup * Data source auto-detection streamlines the warehouse connection flow ### Settings & UX * Column type icons are now consistently aligned across all data tables * Full internationalization pass across the product with improved translations * Fixed a settings panel flash when closing modals ### Performance & Reliability * Faster permission checks with cached authorization policies * Fixed streaming timeouts that could interrupt long-running agent sessions * Fixed issues with model creation forms silently dropping invalid configurations * Resolved errors when editing conversion metrics and derived metric references * Fixed query planner errors when referencing removed semantic objects ### New Onboarding Experience * Interactive product demo that walks through Push AI's capabilities in under 5 minutes — no setup required * Progressive demo starting with a basic data answer, then layering on web research and integrations to show the full power of multi-source AI * Streamlined flow to connect your data warehouse (Snowflake, BigQuery, Redshift, ClickHouse, Databricks) with connection testing and field grouping * After connecting tools like Slack, Notion, and GitHub via OAuth, onboarding shows what was discovered (channels, repos, pages) in a scrollable carousel * Clear 5-step progress indicator (Demo, Connect, Explore, Model, Integrate) keeps you oriented throughout setup ### Redesigned Home Dashboard * Completely redesigned landing experience with a query input overlay and live metric ticker bar with inline sparkline charts * Click any metric in the ticker to see a full-width trend chart powered by real data from your warehouse * See your active agents and recent activity at a glance from the home page ### Metric Formatting * Configure how each metric displays its values — choose number, currency, or percentage with decimal precision, compact notation (1.2K, 3.4M), and custom currency symbols * Formatting settings now apply consistently across charts (axes, tooltips, value labels), data tables, and CSV/JSON exports * Set format and goal direction directly from the metric settings modal * Update format and goal direction for multiple metrics at once ### Unified Data Display * Every data result now has a single toggle to switch between table view, chart view, and the underlying SQL * Smooth cross-fade animations when switching between table, chart, and code views * Toggle text wrapping in data table cells for better readability of long values * Tables with many columns now handle gracefully, preventing issues with wide datasets ### Charts in Notifications * When an agent sends a scheduled report, charts are now rendered as images and included inline in Slack, Teams, and Email * Data table results sent to Slack now include a downloadable CSV file attached to the message ### AI Agent Improvements * Live-updating todo list in the context sidebar as the agent works through complex analyses — see what's pending, in progress, and completed * Agents can now execute longer analysis runs for deeper, more thorough results * Upgraded report writing for more coherent and complete output * Data queries no longer cap at 100 rows — the agent retrieves all the data it needs for accurate analysis * Improved chart generation with more reliable field detection for visualizations ### Semantic Layer & Validation * New issues page showing all invalid objects with error details and one-click resolution paths * When a dimension, entity, or time field becomes invalid, all metrics that reference it are now automatically flagged * Models with dimensions but no metrics are now fully supported * Faster validation on large semantic layers ### Data Model & Connections * Redesigned business graph with a hub-and-spoke layout and dedicated panels for Sources, Metrics, Relationships, and Knowledge * Consolidated warehouse connection forms into a single consistent experience across all providers ### Task Progress * Integration discovery and metadata sync now report detailed progress (which schemas, tables, and columns are being processed) instead of just showing "pending" ### Performance & Reliability * Fixed several issues that could freeze the query page or prevent navigation * Fixed rendering issues with wide data tables containing many columns * Fixed chart rendering issues in scheduled Slack and Email reports ### Charts & Visualizations * New chart types: treemap, heatmap, pie, and donut charts with percentage labels and smart aggregation * Interactive charts: click to select data points, legend filtering, and reference lines/bands for annotations * Smart date axes with continuous temporal scale and automatic quarter label formatting * Small multiples: break a chart into a grid of subcharts by any dimension for easy comparison * Download charts as images directly from the chat interface * Simplified, cleaner tooltip design for source citations ### Data Tables * Completely rebuilt data tables with column sorting, drag-to-reorder columns, and column resizing * Tables with hundreds of rows now render instantly * Global search across all columns with instant highlighting and filtered row counts * Row grouping and aggregation: group rows by any dimension with collapsible sections and automatic numeric aggregation * One-click copy to clipboard, or export as CSV / JSON — respects any active search filters or row selection * Conditional formatting: toggle heatmap-style color coding for numeric columns to spot patterns at a glance * Sticky headers and optional row numbering ### Query & Chat Experience * New context sidebar showing all data sources and integrations referenced in your conversation * Click any data source in the context sidebar to open a resizable live data preview alongside your chat * Scroll-to-bottom button to quickly jump to the latest message in long conversations * Metrics built from other metrics (derived/composite) are now fully supported in queries * View the generated SQL behind any metric or query result ### Integrations * Connect third-party tools directly from settings with a streamlined OAuth experience * Name your connections, view details, and delete with inline confirmation * Push AI now automatically discovers your Slack channels, GitHub repos, Linear projects, Notion pages, and Asana workspaces so the AI can reference them in context * Improved search and navigation in the integrations settings panel ### Semantic Layer & Models * Create and edit semantic models using a rich YAML editor with inline validation and error messages with line numbers * View and manage derived/composite metrics that reference other metrics * More reliable syncing between your data warehouse schema and your semantic model definitions ### AI Agent * 11 new built-in analysis skills including trend analysis, cohort comparison, anomaly detection, and more * Human-in-the-loop tool approval — the agent asks for your approval before executing sensitive actions * Hybrid query mode: the agent can now fall back to direct SQL when the semantic layer doesn't cover a question * Agent actions collapsed by default with clear one-line summaries; expand for full detail ### Settings & UX * Redesigned modal-based settings with faster loading and cleaner navigation * Two-panel dashboard layout with improved schema browsing on the data page * Better UX for finding and managing saved queries * Cleaner agent history grouped by query instead of individual runs * Streamlined product experience with fewer options for a more focused workflow ### Performance * Faster data transfer with an optimized data pipeline * Query results, entity recognition, and join resolution now cached with automatic invalidation on model changes * Fixed inflated query view counts — now tracks unique viewers per query * Faster page loads through reduced bundle size ### New Feature: Multi-Model Queries * Query data across multiple semantic models in a single request * Smart disambiguation when objects exist in multiple models * Automatic discovery of join paths between models ### Complex Metrics * Create derived metrics based on other metrics * Use metrics in filter conditions * Support for queries without aggregation ### Model Validation * See exactly why a model or object is invalid * Automatic scheduled validation of models ### YAML Model Creation * Create semantic models using YAML syntax with a built-in code editor * Real-time validation with helpful error messages * Automatic label generation from definitions ### New Feature: Tool Approval * Review and approve tool executions before they run * Clear visual feedback on approval status * Dedicated approval flow for SQL execution ### New Feature: AI Skills * New skills system for higher-level agent orchestration * 11 new analysis skills available out of the box * Create custom skills for your team ### Hybrid Query Mode * Automatically fall back to SQL when the semantic layer can't answer a question * Combine semantic layer precision with SQL flexibility ### Time Field Improvements * Models can specify default time fields * Better handling of time grain throughout queries ### Performance * Faster query compilation and execution ### New Feature: Subagents * Main agent can now coordinate specialized subagents for complex analysis * Parallel subagent execution with result aggregation * View results from subagent runs in the main conversation * Control how many subagents run simultaneously ### New Feature: Agent Scheduling * Set up agents to run on a recurring schedule * Configure scheduling behavior with flexible policies * Track whether a run was triggered manually or by schedule ### Agent Configuration * Configure reasoning effort for more thorough or faster analysis * Control the level of detail in agent responses * Choose which AI model powers each agent ### Data Table Improvements * Better axis configuration for data visualizations * Richer metadata for table columns * Improved dimension value handling ### Knowledge Documents * Documents now influence query mode selection * Documents included in AI context based on query mode ### New Feature: SQL Explore Mode * New query mode for direct SQL exploration of your data * Choose between Explore, Analyze, and Deep Research modes * Select specific data sources for exploration * Configure which schemas, tables, and columns are available ### Data Source Objects * Reference schemas, tables, and columns in queries using @mentions * Data source objects now appear in the business graph * Detailed tooltips for database objects on hover ### AI-Powered Metadata * AI-suggested descriptions and synonyms for objects * Edit metadata directly in tables * Refresh schema with automatic column reconciliation ### Knowledge Documents * Link knowledge documents to metrics, dimensions, and other objects * Documents can influence which query mode is used * Full lifecycle management: publish, unpublish, and delete ### Agent History * View parameter values used in agent runs * See who created each agent run ### Search Improvements * Improved search including descriptions and synonyms * Option to search by name only for precise matching * Show complete object paths in selector ### Integrations * Updated authentication for improved compatibility with third-party providers ### New Feature: Knowledge Documents * Full-featured document editor with versioning and autosave * Save drafts and publish when ready * View and restore previous versions of documents * Compare changes between document versions ### Navigation Reorganization * "Knowledge" is now "Context" for clearer navigation * New dedicated Integrations tab under Context ### New Feature: Business Graph 2.0 * Completely rewritten visualization with an interactive canvas * Automatic graph layout with smart positioning * Web pages and knowledge documents now appear in the graph * Rich tooltips on graph nodes showing object details * Navigate the graph by expanding and collapsing nodes ### Agent Improvements * Agents can now have parameterized dimension values * New card UI with metric charts showing goal progress * View agent run history with status and timing * Dedicated tabs for monitoring agent status ### New Data Source: ClickHouse * Connect to ClickHouse databases * Full schema discovery for ClickHouse with proper case-sensitivity handling ### Integration Objects * Select which Slack channels and users to include * New UI for selecting integration objects ### New Feature: Report Builder * Add data queries directly to reports * Use relative time periods instead of fixed dates * Include agent analysis in reports ### New Feature: Agent Configuration * New rich text editor for configuring agent prompts * Reference integrations, metrics, and other objects directly in agent prompts using @mentions * Configure goal metrics for agents to track ### Object Mentions * Rich tooltips showing object details on hover * Integrations can now be mentioned like other objects in prompts ### New Feature: Metric Formulas * New UI for creating and editing metric formulas * Automatic measure extraction from data queries * Visual indicators when objects aren't valid for use in formulas ### Field Mapping * Redesigned field mapping step in model creation * Progressive field mapping during model setup ### Query Experience * Share queries at specific timestamps * Check for model references in queries ### New Chart Types * Waterfall charts for visualizing incremental changes over time * Area and scatter chart options * Improved axis formatting with smarter date and time labels ### New Feature: Query Sharing * Share queries with team members via a new sharing modal * Generate public links to share query results externally * View and edit permissions now enforced on shared queries ### Model Management * Improved model editing experience with better field management * Bulk operations for business objects * Merge duplicate or related business objects * Track and display active and inactive fields ### Metrics & Measures * Improved interfaces for managing measures including SQL preview * Better metric type selection and configuration * Support for distinct count and other aggregate functions * Improved time field handling in models ### Query Experience Overhaul * Redesigned query creation flow with a cleaner, more intuitive interface * New query grid layout for browsing and managing saved queries * More responsive query cancellation ### Data Exploration * Explore dimension values directly within the query interface * Real-time "Analyzing..." feedback while the AI processes your question ### Integrations * Integrations now seamlessly connect to the AI assistant * Better visibility into agent task execution status ### Permissions * New permission scoping for better access control * Fixed interaction issues with permission-gated elements ### Internationalization * Added support for Spanish and French locales ### New Feature: Slack as a Source of Context in Reviews * The agent uses what your team is already discussing in the channel to improve relevancy and respond to feedback. * This gives the agent context around what's important and a method of providing feedback * Use this to facilitate an asynchronous weekly business review style workflows ### Agent Updates * Latency improvements through faster tool initialization with integrations * Thread management in Slack across both channels and app DMs ### Agent Updates * Latency improvements through LLM model caching * Enhanced charting configuration * Enhanced search capabilities for faster and more relevant results ### New Feature: Record-level Data Queries * Support for record-level queries to drill into data further * Define Attributes for longer context or high cardinality fields (Beta) ### New Feature: Tools * Tools allow you to leverage context from and take action in your third-party applications * Tool integrations include Gmail, Hubspot, Google Calendar, Linear, Notion and 250+ additional integrations. * Over time, we'll work to more deeply integrate and connect the objects and resources from these integrations into the Business Graph (e.g., connect a message, meeting transcript, or document to a single, canonical customer record) ### Agent Updates * Query Agent: Improved Slack search performance and latency * Query Agent can now answer questions using Push.ai documentation * Added support for web tools, enabling real-time web and documentation search ### New Feature: Use the Query Agent in Slack and Microsoft Teams * Query the Agent directly in Slack channels, App DMs, and Microsoft Teams * Agent responses in Reviews now include deeper background analysis, chart generation, and custom destination formats ### New Feature: The Push.ai Query Agent * Ask questions in natural language using trusted semantic-layer queries * Agentic Business Graph exploration (i.e., it can do data analysis for you) * Brings explicity tagged and relevant untagged Business Objects into context * Support for citations, sources, charts, tables ### General Updates * Added support for dimensions with derived metrics ### New Feature: Knowledge * Support for generic knowledge processors of chat systems, meeting recording transcripts, and documents * Source specific and LLM-driven chunking * Added ability to integrate knowledge via data warehouses * Added ability to upload knowledge from files ### General Updates * Enhanced performance and latency of the dbt semantic layer integration ### General Updates * Introduced additional calculations for trailing periods and introduced calculations for future periods including "Next Period" and "Leading Periods" * Introduced Spanish translation support ### Bug Fixes * Fixed issue where calculation order was lost in Reviews * Fixed issue with displaying duplicate Dimensions in the Dimension table on metric pages ### New Feature: Outliers (experimental) * Introduced ability for users to receive outliers for any metrics they've subscribed to * Introduced sensitivity configuration for outlier detection ### Performance Improvements * Improved caching and performance of object selectors and omni-search * Introduced unlimited scroll for dimension-values, improving app performance with high cardinality dimension values ### UI Upgrades * Expanded use of iconography for metrics, dimension and dimension-values ### Bug Fixes * Fixed issues with Review creation and saving * Fixed ability to group-by dimensions and filter to dimension values # Caching Source: https://docs.push.ai/data-sources/caching Understand how caching and datasource queries work in Push.ai. Push.ai uses a caching strategy to optimize application performance, while reducing the load on your data sources. ## How it works? Push.ai queries datasources through two methods: 1. **Calculations**: metric and dimension queries using one of our [calculations](/business-objects/metrics#calculations). 2. **Time-series**: metric and dimension queries using a daily time grain. ## When is caching used? Caching is used when querying both calculations and timeseries data. Push.ai implements a caching strategy for each unique metric-dimension combination. For each combination, we maintain a daily cache, meaning that we will query that specific combination from the data source at most once per day. ## When are datasource queries triggered? Push.ai queries datasources using two methods: 1. **In-app Exploration**: Datasource queries are triggered when you interact with a metrics and dimension in the Push.ai UI. 2. **Scheduled Updates**: Datasource queries are triggered on a schedule defined in both Reviews and Subscriptions. With Subscriptions, datasource queries are triggered for each Business Object a user is subscribed to, in addition to any related Business Objects. Related objects include any related metrics and are limited to [AI-enabled dimensions](/business-objects/dimensions#ai-enabled-dimensions) only. If multiple users are subscribed to the same Business Object, the datasource query will be triggered once for each object, rather than once for each user. ## When is datasource compute not used? Datasource compute is not used in any of Push.ai's advanced analytics or AI applications. This includes time-series modeling, such as outlier detection and forecasting. In addition, Push.ai's AI systems combine the existing calculations with large language models, and do not use additional datasource compute. # Cube Cloud Source: https://docs.push.ai/data-sources/semantic-layers/cube-cloud Leverage you cubes, measures and dimensions in Push.ai by connecting with Cube Cloud. Learn about Push.ai