v1.0.0
First official releaseMarch 2026
GenAI Studio
- Code Agent persistent sandbox runtime: warm pool, spot failover, session resume.
- 92 native tools: agents, flows, identity, storage, DataHub and ML Workshop CRUD.
- Three-layer safety filter: input, output and execution; environment-aware enforcement.
- 13 LLM calibrations: Claude Sonnet 4 and Haiku 4, GPT-4o, Llama 2/3, Mistral, Bedrock.
Data & ML
- Data Engineering: 184 ready-made building blocks, Pandas/DuckDB/Spark auto-selection, embedded dbt.
- Stream Processing: five-engine portfolio (Kafka Streams, Flink, Spark SS, NiFi, Kafka).
- ML Workshop: five training adapters, four HPO algorithms, five fairness metrics.
- Model Ops: four-test drift ensemble (KS/PSI/Wasserstein/JS), automatic retraining trigger.
Compliance & security
- Seven-system erasure cascade with the 30-day SLA monitored by Prometheus.
- Hash-chained audit; every row references the previous and current hash.
- ISO 27560 signed consent receipts; retroactive policy mutation impossible.
- Turkish PII default: national ID, tax ID, IBAN and phone, all checksum-validated.
Deployment & ops
- Five deployment modes: k3s, air-gapped USB, AWS single-node, AWS EKS, Marketplace SaaS.
- Four capacity profiles (Small/Medium/Large/Enterprise); profile upgrade without data migration.
- License engine: seven-layer verification, 30-day offline grace, production endpoint immunity.
- Hybrid GPU scheduler (K8s + ECS scale-to-zero), idle timeout 5–480 minutes.