Roadmap

Stage and direction for NexusRAG

Connecting

Current stage

Where this project sits today.

Live in production

Now — Live in production

now

The platform serves real workload through a streaming LangGraph agent. Tier-A telemetry exposes workload counters, latency percentiles, and uptime. Audit log is tamper-evident and always on.

Multi-tenant /v1/run streaming endpoint with SSE
RBAC + ABAC + document ACLs (default-deny)
Bedrock KB and Vertex AI Search retrieval routing live
Envelope encryption + KMS key rotation enforced
SOC 2 evidence bundles persisted under var/evidence
Public Tier-A /api/stats with workload metrics

Next — Reliability + governance hardening

Active themes from the CHANGELOG Unreleased section: notification receiver contract v1.0, governance retention proofs, and compliance evidence automation. These tighten the operational surface without changing the public contract.

Notification Receiver Contract v1.0 (typed headers, signature parsing, dedupe)
DSAR / governance retention-proof workflows
Tenant-scoped notification routing with DLQ replay
Admin API key lifecycle endpoints + keyring rotation tooling
Operability evaluator worker with distributed locking

Later — Federated + edge-native retrieval

later

Where the platform is heading once the current reliability themes close. The architecture (multi-cloud retrieval routing, audit-evident logs, kill-switched feature flags) is already shaped for these.

Federated retrieval across customer-owned data planes
Active failover with geo-replicated audit log
Self-serve compliance attestation export
Edge-cached embeddings with regional TTLs
Customer-supplied LLM provider plug-in surface