Bast AI Platform¶
Enterprise conversational AI with explainable knowledge architecture.
Start here
New to Bast or reviewing the platform as a sponsor? Begin with the Start Here guide to frame the use case, trust boundary, pilot path, and first documents to read.
What is Bast AI?¶
Bast AI enables organizations to build sophisticated conversational AI systems through a configuration-driven architecture. The platform supports everything from simple single-agent chatbots to complex multi-agent orchestration systems, with enterprise features built in:
- Multi-tenant data isolation via MongoDB Atlas with per-tenant database partitioning
- OAuth2/OIDC + MFA authentication with WorkOS, Google, and Microsoft providers
- W3C PROV-O provenance lineage anchoring every AI generation to curated expert knowledge
- Cloud-managed infrastructure with automated provisioning and environment isolation
Platform at a Glance¶
| Category | Count | Highlights |
|---|---|---|
| API Services | 13 | CAT, Cat-Runtime, Platform, Security, Analytics, Org-Services |
| AI Models | 13 | Intent, QA, Context, Greeting, Early Exit, OpenAI Assistant |
| Pipelines | 19 | PDF, MongoDB, OpenAI, Bedrock, Azure Index, Agno, Analytics ETL |
| Core Libraries | 33 | Base, Config, Crypto, Mongo, Neo4j, Classifier, NLP |
| Agent Configs | 15+ | Core, Memory, Session, Knowledge, Tools, Reasoning |
| MCP Servers | 3 | Base, MongoDB, Ontology |
| Frontend Apps | 4 | Admin UI, BFF (tRPC), Chat UI, Bast UX |
Key Capabilities¶
Multi-Agent Orchestration¶
Teams and workflows with conditional routing, sequential execution, and autonomous coordination between specialized agents — powered by the Agno Framework.
Knowledge Retrieval (RAG)¶
Hybrid search combining text and vector similarity, with entity filtering and score normalization. The retrieval layer surfaces provenance metadata so every answer can be traced to its source.
Explainable AI via PROV-O¶
Every knowledge entity carries W3C PROV-O lineage — prov:wasDerivedFrom and prov:wasGeneratedBy annotations that trace from AI-generated answers back through the curated expert ontology to the original source documents.
Enterprise Security¶
JWT authentication, role-based access control, AES encryption, MFA enrollment, and comprehensive audit logging across all platform operations.
Configuration-Driven Architecture¶
The entire platform is configuration-driven — agents, models, knowledge domains, and workflows are defined in YAML and managed through the platform API, enabling rapid deployment of new conversational AI use cases.
Technology Stack¶
| Layer | Technologies |
|---|---|
| Backend | Python 3.10-3.12 |
| Frontend | TypeScript, React 18+, Tailwind CSS v4 |
| Build | Pantsbuild, Poetry (backend); pnpm, Vite (frontend) |
| API | FastAPI, Uvicorn, tRPC (BFF) |
| Agents | Agno Framework, OpenAI |
| Databases | MongoDB Atlas, SQL Server, Neo4j, Milvus |
| Cloud | AWS, Azure, Docker, Terraform |
Explore the Platform¶
-
Three-tier service model, data architecture, and multi-tenant isolation patterns.
-
Complete platform capabilities from chat processing to analytics.
-
Agent configuration, workflows, knowledge retrieval, and shared memory.
-
33 Python libraries, 13 APIs, 19 pipelines — fully documented.
-
Admin UI, BFF architecture, design system with shared tokens.
-
Our engineering methodology: 173 capability specs across 18 domains, spanning bast-admin-ui (165) and bast-ai (8).