AI
Use models, organizational knowledge, documents, voice, and agents through governed interfaces.
AI is the product domain for building and using governed AI experiences. It provides stable application interfaces while keeping model runtimes, storage engines, and workflow components behind platform boundaries.
Choose an outcome
Access models
Discover and use platform-hosted or connected models.
Build with the BSQAI API
Follow an application journey from authentication to operation.
Ground responses with RAG
Ingest organizational knowledge and retrieve relevant context.
Connect tools with MCP
Make Model Context Protocol servers available to an organization.
Agents
Create reusable assistants with instructions, knowledge, and tools.
Skills
Reuse, author, and share task instructions in text chat.
Voicebot
Start a live spoken conversation and continue in the same chat.
Deep Research
Plan a web investigation and reuse its cited report.
Process documents
Convert documents for extraction and retrieval workflows.
Capability map
Model access and routing
Model as a Service covers model discovery, registration, cloud-model connections, and the hand-off into inference. Applications integrate through the BSQAI API instead of depending directly on internal gateways or serving engines.
Each model has one route to its server: a KServe deployment in the cluster, or a connection to an external model server that a platform administrator configures. Before the BSQAI API forwards a request, it checks that the caller may use the model, checks the monthly BCU caps that apply to the caller, and records the call as usage. A route does not fall back to another model. See Cloud and external models for routes and connections.
Usage, budgets and access
A model is usable only when an allow rule lets the caller use it, apart from the model's owner. Model pickers and /v1/models list only the models the caller may use. See Model access.
The BSQAI API records every model call as usage, with its cost in BCU, by user, API key, application, model and group. Members see their own consumption in My usage. Tenant administrators see the organization's consumption in Analytics and set monthly BCU caps on members and on the groups that fund API keys in Budgets.
Applications authenticate with an API key, which carries the permission scope chosen when it was created, or with a service account, which gives the application its own identity in the organization.
Conversational AI and knowledge
Conversational experiences combine model access, conversation state, organizational knowledge, and permission-aware retrieval. The Portal provides the user-facing experience; RAG explains ingestion, chunking, embeddings, vector search, and grounded responses.
Use the AI and knowledge guided path when the goal is to help users work with approved models and enterprise content.
Documents, voice, and multimodal input
Document processing converts files into structured content for extraction, classification, and retrieval workflows. Voicebot combines live transcription, model replies, speech playback, and interruption in one conversation. Separate dictation and read-aloud controls support editable speech input and playback of existing replies. These capabilities share authentication, workflow, storage, and observability services with other AI experiences.
Use Docling for document conversion details and RAG for the complete ingestion and retrieval flow.
Agents and tool use
Agentic applications add orchestration, governed tools, connectors, and secure execution to model interactions. BullSequana AI exposes approved tool interfaces and carries the calling identity's tenant and team boundary into those integrations.
Connectors is the tool-integration surface. A Platform Admin makes a Model Context Protocol server available to an organization, chooses which of its tools may be called, and declares its credential mode. Each member links their own credential for the system behind the server, except where that system needs no account and the server is registered as unauthenticated. Every call passes through the MCP Gateway.
Start with the application-building path, then use How MCP works on the platform for the request flow and the MCP Lifecycle Operator reference for managed workload reconciliation.
Where the platform team enables the agent sandbox, the Build mode of OmniChat gives the agent an isolated workspace in which it runs commands and writes files, and, where the site enables previews, serves a web app preview. See Build and the Agent Sandbox component.
Agents save a model and instructions with optional knowledge sources and Connectors. Their retrieval uses the caller's access, not the creator's credentials.
Skills add reusable task instructions and supporting files to Portal conversations without changing the selected model or Agent. A skill does not grant tool access. See Use skills in chat for the member workflow.
Model lifecycle
Experiment tracking, registration, sharing, deployment, and serving form one lifecycle. The Global organization uses the installation's MLflow. A customer organization tracks experiments and registers models in its own private MLflow, and deploys them into its own namespace within the GPU guarantee and cap that a platform administrator sets for it. AI exposes approved models through the BSQAI API; Foundation runs the serving infrastructure.
The Data and ML guided path connects those responsibilities.
Typical application flow
- Identify whether the application needs direct generation, organizational knowledge, document processing, or tools.
- Choose a model that an allow rule lets your users, service account, or API key use, and check the monthly BCU caps that apply to them.
- Authenticate to the BSQAI API with a service account or an API key scoped to the permissions it needs, and implement the smallest end-to-end request through
/v1/chat/completions,/v1/responses, or the Anthropic Messages API. - Add retrieval, documents, voice, or tools only when the use case requires them.
- Deploy the application with health checks, resource limits, telemetry, and a rollback path.