Skip to content

Start Here

Use this page to orient a sponsor conversation before diving into the full platform documentation.

What Bast Is

Bast is a platform for building governed AI assistants around an organization's approved knowledge, operating boundaries, and review process.

The core product object is a CAT: an AI system configured for a specific audience, purpose, knowledge boundary, behavior pattern, and evidence requirement. A CAT can combine retrieval, agents, workflows, tools, memory, and runtime records without treating every product decision as a source-code change.

  1. Name the use case. Define the audience, job to be done, approved sources, out-of-scope requests, and measurable outcome.
  2. Confirm the trust boundary. Identify data ownership, organization scope, access roles, source owners, model/provider constraints, and human approval points.
  3. Review the platform shape. Read the platform architecture, feature overview, and relevant API or frontend documentation.
  4. Pilot with evidence. Test representative questions, missing-knowledge cases, refusal cases, and reviewer workflows against a named CAT configuration.
  5. Decide how to operate. Agree on owners, monitoring, change control, support expectations, and re-review triggers before expanding.

What To Read First

Need Start with
Product orientation Bast AI Platform
Capability map Features & Capabilities Overview
Architecture review Platform Architecture
Engineering lens Bast Core Principles
API surface API Reference
Admin and frontend context Frontend Documentation

Questions To Answer Before A Pilot

  • Who is the sponsor owner and who signs off on the use case?
  • Which sources are authoritative, current, and approved for model use?
  • What should the CAT refuse, escalate, or hand back to a person?
  • What evidence must be visible to reviewers or end users?
  • Which users, roles, and organizations should have access?
  • What would make the pilot successful enough to expand?

Boundary Of This Guide

This sponsor documentation is a public-safe product and technical briefing. It intentionally omits credentials, customer data, infrastructure identifiers, internal topology, and controlled security evidence. It is not a certification, contract, service-level commitment, or substitute for environment-specific diligence.