NexVigilant Core (nexcore)

The Vigilance Kernel — unified Rust workspace (~63K LOC) powering NexVigilant's pharmacovigilance platform, skill orchestration, and AI agent infrastructure.

nexcore is the proprietary computation kernel developed by NexVigilant LLC for drug safety signal detection, regulatory compliance, and intelligent automation.

Quick Start

# Build
cd ~/nexcore
cargo build --release

# Run tests (~1800+)
cargo test --workspace

# Start REST API (preferred for external integrations)
cargo install --path crates/nexcore-api
nexcore-api  # http://localhost:3030

# Use MCP tools (preferred for Claude Code)
# mcp__nexcore__pv_signal_complete(a=15, b=100, c=20, d=10000)

Workspace Structure

~/nexcore/
├── Cargo.toml                    # Workspace root
└── crates/
    ├── nexcore-vigilance/        # Core: ToV axioms, Guardian-AV, PV signals, 25+ modules
    ├── nexcore-brain/            # Working memory: sessions, artifacts, code tracking
    ├── nexcore-mcp/              # MCP server (90+ tools for Claude Code)
    ├── nexcore-api/              # REST API server (Axum, 33 routes)
    ├── nexcore-friday/           # FRIDAY orchestrator (voice, webhooks, scheduler)
    ├── nexcore-hooks/            # Claude Code hooks (quality, security enforcement)
    ├── nexcore-faers-etl/        # FAERS data pipeline
    ├── nexcore-config/           # Configuration types
    ├── nexcore-skill-verify/     # Skill validation CLI
    └── nexcore-epa1/             # EPA ECHO API integration

Workspace Dependency Direction

Use a one-way dependency flow to keep layering clear and avoid cycles:

1. foundation crates: shared primitives, utilities, types
2. domain crates: business logic, algorithms, PV, vigilance
3. orchestration crates: workflows, pipelines, schedulers
4. service/bin crates: API servers, CLIs, MCP, UI shells

Rules of thumb:

1. service/bin can depend on anything below it, but never the other way around.
2. domain should not depend on service/bin; keep boundaries clean.
3. foundation should be leaf-level: no dependencies on higher layers.

Crate Overview

Access Methods (Priority Order)

1. MCP Tools (Preferred for Claude Code)

mcp__nexcore__pv_signal_complete(a=15, b=100, c=20, d=10000)
mcp__nexcore__foundation_levenshtein(source="kitten", target="sitting")
mcp__nexcore__brain_artifact_save(name="task.md", content="...")
mcp__nexcore__skill_validate(path="~/.claude/skills/my-skill")

2. REST API (External systems, webhooks, scripts)

nexcore-api                        # Start on http://localhost:3030
PORT=8080 nexcore-api              # Custom port

curl -X POST http://localhost:3030/api/v1/pv/signal/complete \
  -H "Content-Type: application/json" \
  -d '{"a": 15, "b": 100, "c": 20, "d": 10000}'

Docs: http://localhost:3030/docs (Scalar UI)

3. CLI (Deprecated)

nexcore verify <path>              # Validate Diamond v2 compliance
nexcore skill scan <path>          # Build skill registry

Signal Detection Thresholds

Performance (vs Python)

Cognitive Hooks

Hooks auto-enforce code quality and security:

• Python Blocker: Enforces 100% Rust mandate
• Secret Scanner: Blocks commits with credentials
• Clippy Enforcer: Zero warnings policy
• Panic-Free Guarantee: Blocks unhandled unwrap()/expect()

Protocol: Exit 0 = allow, Exit 1 = warn, Exit 2 = block

Config: ~/.claude/settings.json

Development

# Run specific crate tests
cargo test -p nexcore-vigilance
cargo test -p nexcore-mcp -- test_foundation

# Verbose output
RUST_LOG=debug cargo test -p nexcore-mcp

# Lint (strict)
cargo clippy --workspace -- -D warnings

# Build release
cargo build --release --workspace

License

Proprietary. Copyright (c) 2026 NexVigilant LLC. All Rights Reserved.