GNOSTRA Project Roadmap
GNOSTRA is a complete, multi-stage autonomous kernel. Our development is structured into four milestones, all of which are now structurally complete and operational for the closed beta.
M0: The Kernel
COMPLETE- Goal: Prove the core logic.
- Status: COMPLETE.
Key Features:
- NeedIndex: The 12 Pillars of Need objective function.
- CP-SAT Allocator: The core solver that generates optimal plans.
- Trust Gate: The
validator,attestation, andprivacyservices (M0.3).
M1: The Distributed Backbone
COMPLETE- Goal: Scale the kernel into a persistent, distributed network.
- Status: COMPLETE.
Key Features:
- Data Plane (M1.1): Transitioned to a persistent Neo4j Knowledge Graph and NATS Event Bus.
- Orchestrator (M1.2): Autonomous, event-driven service (
gnostra_orchestrator) that dispatches plans via gRPC. - Mock Adapter: A functional gRPC agent that can receive and acknowledge tasks.
M2: The Trust & HMI Layer
COMPLETE- Goal: Make the network verifiable, auditable, and human-usable.
- Status: COMPLETE.
Key Features:
- PII Vaulting (M2.1): PII is no longer redacted, but replaced with verifiable cryptographic hashes.
- Verifiable Planning (M2.2): The Allocator cryptographically signs every
AllocationPlan. - GNOSTRA Qt Client (M2.3): The complete HMI, featuring the Bottleneck Map, Pillar Gauges, Plan Explainer, and Data Submission forms.
M3: The Learning Loop
COMPLETE- Goal: Make the system intelligent and autonomous.
- Status: COMPLETE.
Key Features:
- Simulator (M3.1): The "Digital Twin" that generates synthetic crisis scenarios.
- Learner (M3.2): The Epsilon-Greedy (MAB) agent that observes simulations and learns optimal policies.
- Evaluator (M3.3): The audit layer that runs counterfactual simulations to calculate "regret" and verify the Learner's intelligence.
M4: Post-Beta & Future Operations
IN PROGRESS- Goal: Harden the network for production and expand the learning model.
- Status: IN PROGRESS.
Next Steps:
- Deployment Packaging: Finalize the
docker-composeenvironment for testers. - Security Hardening: Implement full gRPC authentication and rotate all development keys.
- M3.3 Integration: Feed the "Regret Metric" from the Evaluator back into the Learner's decision model.
- Policy Expansion: Expand the Learner's "policy knobs" from one (w_strictness) to dozens (e.g., cost, speed, priority weights).