Roadmap¶

This page outlines planned features for tobira. Timelines are estimates and may change based on community feedback and development priorities.

Want to influence priorities?

Join the waitlist — early members get priority access and input on feature prioritization.

Current: Community Edition (v0.5)¶

The open-source Community edition is production-ready and includes:

All inference backends (FastText, BERT, ONNX, Ollama, LLM API, Ensemble, Two-Stage)
All MTA plugins (rspamd, SpamAssassin, Haraka, Postfix milter)
Full CLI toolset (init, doctor, monitor, train, evaluate, demo, distill, hub-push/pull, ab-test, active-learning)
A/B testing and active learning
Web dashboard
Knowledge distillation
AI-generated text detection
HuggingFace Hub integration
GDPR-aware PII anonymization
Docker Compose deployment with health checks
Kubernetes-ready health probes (readiness / liveness)
PostgreSQL and Redis storage backends

These features are in the planning stage. No implementation has started yet. Target availability: second half of 2026.

Feature	Description	Status
Multi-tenant management	Isolated configurations per tenant with centralized administration	Planning
RBAC + audit logs	Role-based access control with full audit trail	Planning
OpenTelemetry integration	Metrics and traces export via OpenTelemetry protocol	Planning
Grafana dashboard templates	Pre-built dashboards for prediction monitoring and drift detection	Planning
SSO (SAML / OIDC)	Enterprise identity provider integration	Planning
Priority support + SLA	Dedicated support channel with service-level agreement	Planning

Timeline disclaimer

Target dates are estimates based on current plans and may shift depending on community feedback, waitlist demand, and development capacity.

These features depend on Enterprise features and are in early concept stage. Target availability: 2027.

Feature	Description	Status
Cloud-based model training	Upload anonymized data, train on managed GPUs, download results	Concept
Managed GPU resources	No GPU procurement or maintenance required	Concept

Feature priority is determined by:

Waitlist demand — Features most requested by waitlist members ship first
Community feedback — GitHub Issues and Discussions influence the backlog
Technical dependencies — Some features (e.g., Cloud) depend on others (e.g., Enterprise auth layer)