Modular open-source platform for Retrieval-Augmented Generation (RAG) with specialized verticals for e-learning and beyond.
Vektra is designed as infrastructure, not as a consumer application. It provides building blocks for developers and organizations that need to integrate RAG capabilities into their systems, with a specialized vertical for learning management and knowledge delivery.
vektra-stack is the entrypoint for understanding Vektra. It does not contain core implementation code. Instead, it serves as:
- A central reference for the overall architecture
- Documentation hub for the multi-repo ecosystem
- A guide to understanding how components relate to each other
The actual implementation lives in separate, focused repositories.
Vektra is organized in layers:
┌───────────────────────────────────────────────────────────────────┐
│ DEPLOYMENTS Branded instances for specific institutions │
├───────────────────────────────────────────────────────────────────┤
│ VERTICALS Domain solutions (vektra-learn for e-learning)│
├───────────────────────────────────────────────────────────────────┤
│ LMS ADAPTERS Platform integrations (vektra-moodle, etc.) │
├───────────────────────────────────────────────────────────────────┤
│ CORE RAG engine, ingestion, indexing, analytics │
├───────────────────────────────────────────────────────────────────┤
│ OPERATIONS System administration (vektra-admin) │
├───────────────────────────────────────────────────────────────────┤
│ INTEGRATION SDKs (Python, JavaScript) │
└───────────────────────────────────────────────────────────────────┘
For detailed architecture documentation, see ARCHITECTURE.md.
- Separation of concerns — Each repository has a single, well-defined responsibility
- No tight coupling — Components communicate through stable interfaces, not internal dependencies
- Infrastructure-first — Built as platform components, not end-user applications
- Composability — Use only what you need; components work independently or together
- Deployment parity — Same architecture for cloud and on-premises installations
- Workflow orchestration — Pipeline components designed for external orchestration (n8n)
vektra-stack/
├── README.md # This file
├── ARCHITECTURE.md # High-level system design
├── ROADMAP.md # Development phases and milestones
└── CONTRIBUTING.md # How to contribute
Vektra is in early development (Phase 0: Foundation). We are:
- Defining component boundaries and interfaces
- Documenting architecture and data flows
- Preparing the foundation for individual component repositories
See ROADMAP.md for planned phases.
| Repository | Description | Status |
|---|---|---|
| vektra-core | RAG engine, LLM abstraction, conversation management, safeguards | Planned |
| vektra-ingest | Document processing pipeline (PDF, OCR, PPT, Word) | Planned |
| vektra-index | Vector store abstraction, embedding, semantic search | Planned |
| vektra-analytics | Metrics aggregation, reporting API, alerting | Planned |
| Repository | Description | Status |
|---|---|---|
| vektra-learn | E-learning vertical (chatbot, instructor dashboard) | Planned |
| vektra-moodle | Moodle LMS adapter (SSO, content sync, embed) | Planned |
| Repository | Description | Status |
|---|---|---|
| vektra-admin | System administration interface | Planned |
| vektra-sdk-py | Python SDK | Planned |
| vektra-sdk-js | JavaScript/TypeScript SDK | Planned |
We welcome contributions. Please read CONTRIBUTING.md before submitting issues or pull requests.
This project is open source. License details will be finalized as the project matures.
Vektra is maintained by VektraLabs.