← Back to Technical blog

Technical article

Embedding Agentic BI in Business Systems: HENGSHI SDK, iframe, and API Delivery Paths

Distinguish Agent and Workflow execution modes and organize identity, interaction, and delivery responsibilities across iframe, JS SDK, and API integrations.

Nov 16, 2026Technical blogHENGSHI13 min read
Agentic BIHENGSHI SENSE

Article body

Full article

When enterprises integrate Agentic BI into business systems, they typically want users to view business dashboards, ask follow-up questions, and connect analysis results to their own processes within one portal. Full pages, local interactions, and backend calls have different integration requirements. Projects must also connect host identities with analytics resources so people can use these capabilities from existing business entry points.

HENGSHI provides page embedding, a JS SDK, and APIs for these needs: iframes reuse complete analytics pages, the JS SDK integrates conversations and result rendering, and APIs support server-side calls and workflow orchestration. Integration teams can choose the appropriate layer around existing datasets, metrics, dashboards, and permissions, then configure authentication, resource access, and page interactions for the actual entry point.

Founded in 2016 for the Data+AI era, HENGSHI has pioneered a new generation of Agentic BI. Its enterprise analytics platform, HENGSHI SENSE, works with the HENGSHI JARVIS digital employee, the HENGSHI CLI execution engine, and HENGSHI BOX, a privately deployed control hub, to form an integrated service portfolio. With a decade of BI expertise and its “Powered by Hengshi” ecosystem strategy, HENGSHI empowers software vendors across industries and provides enterprise groups in manufacturing, finance, retail, government, and other sectors with integrated AI analytics for business management and intelligent decisions, built on trusted data within their private environments.

1. Choose the execution mode before the integration method

HENGSHI Data Agent integration distinguishes Agent and Workflow execution modes. Agent mode advances conversations and progressively reports execution; in browser embedding, a frontend runtime participates. Workflow mode advances the conversation through a backend workflow service. Enterprises needing backend integration with Agent mode can also evaluate the Node.js Agent service.

Iframes, the JS SDK, and APIs are integration methods. API integration can support different execution modes, and the SDK does more than display generated text. Separating these dimensions clarifies responsibility for conversations, execution, presentation, and troubleshooting.

Design dimensionQuestion to answer firstDelivery implications
Execution modeDoes a frontend Agent runtime, Node.js Agent service, or backend Workflow advance the analysis?Determines execution, status feedback, and service deployment responsibilities
Integration methodReuse a full page, conversation and result components, or backend capabilities only?Determines the scope of interface and interaction development
Identity and dataWho calls, which tenant do they belong to, and which resources can they access?Determines authentication, authorization, and acceptance conditions

2. Iframes: reuse the full analytics experience and connect it to the host

2.1 From complete pages to individual analytics objects

An iframe suits projects that want to reuse HENGSHI’s existing analytics pages. A project can embed a full ChatBI page or a selected application, dashboard, or chart. Users retain HENGSHI’s question-answering and chart interactions, while the integrator focuses on business pages and entry points.

Customer-facing products can also use iframes. Dashboard embedding supports configuration of headers, pagers, scrolling, backgrounds, title colors, and chart spacing. Themes and OEM options can further align the experience. The achievable fit depends on the target page, supported parameters, and customization scope; verify each item against actual designs.

2.2 Bring business context into the analytics page

When a customer-detail page switches customers, its analytics area should update too. HENGSHI provides URL parameters and page messaging so hosts can pass parameters, update embedded URLs, or trigger dashboard refreshes according to the documentation. Hosts can also listen for application, directory, clicked-object, and URL changes to connect analysis with business actions.

Distinguish business filters from data authorization. Date, store, and product parameters identify the current analytical task. Data permissions depend on the authenticated user and platform grants. Hiding a filter or passing a customer ID in a URL cannot replace data permission configuration.

3. JS SDK: retain your interface and reuse conversation and result capabilities

The HENGSHI Copilot JS SDK supports two common forms: a complete chat window and question-answering result cards placed inside a business page. The former suits a sidebar assistant; the latter suits fixed layouts such as order-detail pages and business cockpits.

Integrators can control the mounting container, visibility, window position, dimensions, language, and supported styles and wording. HENGSHI provides frontend question-answering and chart-result capabilities. In Agent mode, the SDK also handles corresponding frontend execution. In Workflow mode, the backend advances the task while the SDK handles the relevant interaction and presentation.

The SDK’s value: reuse analytics interactions inside your business pages and reduce repeated work on chat, progress feedback, and chart-result presentation. Integrators still need to understand supported configuration and execution modes.

SDK configuration can specify data packages, datasets, or subject areas for question-answering. This helps locate resources; actual access still requires an authenticated identity and resource grants. When customers, tenants, or accounts change, the host should recheck identity, data scope, and conversation ownership so previous business context does not carry into a new page.

Projects should maintain SDK and server version compatibility and validate static asset domains, API addresses, and cross-origin settings. The SDK does not automatically convert HENGSHI results into a custom charting library’s protocol. Fully custom rendering requires separate evaluation of result parsing, interactions, and ongoing adaptation costs.

4. APIs: bring data services and intelligent analysis into your processes

4.1 Distinguish data service APIs from Data Agent integration

HENGSHI data services can publish connections, datasets, or chart results as interfaces. They support aggregated data, detail data, custom queries, and registered APIs, with publishing, authorization, rate limiting, and monitoring. This suits business pages that consume existing analysis results directly.

Data Agent API integration addresses question-answering and analysis execution. Enterprises with an AI gateway, multi-agent orchestration, or backend business processes can integrate HENGSHI as a data analysis capability. A project must identify its execution mode and responsibilities for requests, status, result presentation, and error handling. Both API types can work together, but calling a data endpoint does not by itself deliver the complete Agentic BI experience.

4.2 Connect calling identities to authorization configuration

Integrators should manage API client credentials on trusted servers, obtain the required Access Token through the configured process, and verify the actual execution identity, authorization targets, and data scope. Backend service credentials and page sessions need separate management. A business user’s host login does not make a backend administrator token represent that user.

The total cost of API integration depends on the work the project assumes. Consuming aggregate results may be straightforward; building multi-turn chat, charts, linked filters, and recovery requires more development. List the capabilities to reuse before estimating effort.

5. Combined delivery: a parent-page assistant and embedded dashboards

HENGSHI supports placing the Copilot JS SDK in the parent page while embedding a dashboard through an iframe. The parent-page Copilot can handle the dashboard’s intelligent assistant entry point. Users can view charts, request interpretation, and continue analysis through exposed business actions and objects within one business page.

This combination suits enterprises with an existing portal framework. The parent page retains navigation and business actions, the iframe supplies the dashboard, and the parent assistant centralizes question-answering. Integration requires the correct parent origin, appropriate login sessions, and validation of user permissions for dashboards and charts. Agent access to iframe content uses exposed business capabilities for the corresponding reads or changes.

Project needPath to evaluate firstIntegrator’s main responsibilities
Launch a complete data question-answering entry pointChatBI page iframeAuthentication, container dimensions, entry point, and data scope
Add a sidebar assistant or result cards to an existing pageCopilot JS SDKPage layout, conversation continuity, and SDK configuration
Unify dashboards and an intelligent assistant in a portalParent-page SDK plus iframe dashboardCross-page communication, context handoff, and permission acceptance
Supply existing analytics results to a custom interfaceData service APIResult consumption, interface, and interactions
Connect an internal backend analytics processData Agent API integrationRuntime services, orchestration, status, and error handling

6. Example design: analytics in a retail-chain operations portal

The following is a hypothetical design. A retail chain wants regional managers to view monthly sales and investigate changes in returns within its operations portal. The page retains the enterprise’s navigation and store-management entry points, uses a HENGSHI iframe for the dashboard, and adds an assistant through the Copilot JS SDK in the parent page.

  1. Prepare the model. The data team organizes orders, refunds, and stores, confirms metrics and reporting periods, and builds the base dashboard.
  2. Prepare identities. The integration team connects SSO or JWT, maps each regional manager to a specific HENGSHI user, and configures regional data permissions.
  3. Connect tasks in the portal. Pass the current store and date as analytical context and update them when pages change. Users can continue with related questions after viewing charts.
  4. Deliver validated results. Use dashboards or data services for fixed displays and retain conversational interaction where follow-up analysis is needed.

Acceptance should cover account switching, region switching, follow-up questions, sign-out and sign-in, and page reloads. Compare dashboard and conversation results under the same metric definition and data scope. Investigate differences through time conditions, refresh status, execution identity, and calculation definitions.

7. Reusable analytics delivery is the foundation of embedding

HENGSHI connects data modeling, application authoring, intelligent question-answering, and external services in one platform. Enterprises can reuse complete pages, conversational components, or backend capabilities as projects require. Delivery teams can focus more effort on business processes, data definitions, and identity authorization while retaining their product’s interface framework and customer relationships. Clear responsibility for execution, presentation, and authorization underpins maintainable Agentic BI integration.

HENGSHI SENSE

Resources, ecosystem, and implementation stories

Explore how teams design and ship analytics with HENGSHI.

Request a trial

Enterprise deployment, embedded delivery, and trial requests can all be handled quickly.