Skip to content

README.md

Latest commit

 

History

History
126 lines (110 loc) · 5.91 KB

README.md

File metadata and controls

126 lines (110 loc) · 5.91 KB

OntoLoop Sovereign OS (Engineering Runtime)

OntoLoop is an AI sovereign runtime oriented to Governance + Execution + Learning + Replay. The current version has entered a production-ready engineering phase, featuring closed-loop capabilities for governance, execution, trust verification and continuous learning.

Core Capability Modules

  • Governance Pipeline：Policy / Tenant / Approval / Risk / Budget / Rollout
  • Execution Pipeline：Capability Admission → Trust Admission Kernel → Runtime Guard → Layered Tool Stack → Execution Fabric
  • Trust Pipeline：Evidence Tagger → Flow State Engine → Trust Evidence Ledger (Hash-chain)
  • Learning Pipeline：Learning Proposal → Promotion Gate → Private Memory / Org Knowledge
  • Observability Pipeline：Telemetry Collector → Policy Signal Aggregator → Query/Replay/Explanation Plane
  • TrustKernel (NoumenonCore) Integration：Identity Authentication, Attestation, Supply Chain Validation, Strong Isolation, Replayable Audit

Architecture Pipeline (Compact Version)

[User/Trigger]
    -> [Structured Transport + Session Bridge]
    -> [Query Turn State Machine + Context Compiler/Compactor]
    -> [Requirement Clarification]
    -> [Policy & Governance Context]
    -> [Knowledge Context + SuperMemory]
    -> [QueryEngine + Orchestrator + Capability Router]
    -> [Capability Admission]
    -> [Trust Admission Kernel]
    -> [Runtime Guard + Permission Mode]
    -> [Layered Tool Execution + Execution Fabric + Hook Runtime]
    -> [Evidence Tagger]
    -> [Flow State Engine]
    -> [Trust Evidence Ledger]
    -> [Verifier & Audit]
    -> [Learning + Promotion]
    -> [Memory / Org Knowledge Update]
    -> [Unified Query / Replay / Explanation]
    -> [Observability + Reports]
    -> [Next Iteration]

Quick Start

1) Environment Requirements

  • Rust toolchain
  • Optional: StateStore CLI
  • Optional: Docker / Docker Compose

2) Local Execution

cargo run --manifest-path .\Cargo.toml -- --message "Build a governed autonomous loop" --swarm

3) Local Validation

cargo check --workspace --manifest-path .\Cargo.toml
cargo test --workspace --manifest-path .\Cargo.toml

Common CLI Commands

cargo run --manifest-path .\Cargo.toml -- system health
cargo run --manifest-path .\Cargo.toml -- --session demo system replay-report
cargo run --manifest-path .\Cargo.toml -- trigger list
cargo run --manifest-path .\Cargo.toml -- --session demo focus board
cargo run --manifest-path .\Cargo.toml -- --session demo org context
cargo run --manifest-path .\Cargo.toml -- bridge status
cargo run --manifest-path .\Cargo.toml -- knowledge batch-export --anchor-list .\deploy\anchors.txt --type graph
cargo run --manifest-path .\Cargo.toml -- trigger webhook --anchor-id cli:focus --topic order.created --payload "{\"order_id\": \"A-1001\"}" --run-now
cargo run --manifest-path .\Cargo.toml -- --session cli:focus system export
cargo run --manifest-path .\Cargo.toml -- --session cli:focus frontend status
cargo run --manifest-path .\Cargo.toml -- --session cli:focus frontend events --format pretty --limit 20

The CLI frontend is designed additively. Existing frontend directories such as dashboard-ui/ are fully reserved for subsequent application interface expansion.

Acceptance Scripts

powershell -ExecutionPolicy Bypass -File .\deploy\scripts\p95_acceptance.ps1
powershell -ExecutionPolicy Bypass -File .\deploy\scripts\pq9_acceptance.ps1
powershell -ExecutionPolicy Bypass -File .\deploy\scripts\week6_acceptance.ps1
powershell -ExecutionPolicy Bypass -File .\deploy\scripts\trigger_supermemory_acceptance.ps1

Linux Environment:

bash ./deploy/scripts/week6_acceptance.sh

Output Artifacts:

  • deploy/runtime/p95-acceptance.log
  • deploy/runtime/p95-acceptance.json

Directory Structure

  • src/: Core runtime & governance logic
  • src/query_engine/: Turn state / Continuation / Context Compactor / Loop Scheduler
  • src/runtime/: Runtime Guard / Admission Control / Execution / Evidence Collection / Flow Engine
  • src/security/: Policy Control / Permission Mode / Capability Admission System
  • src/session/: Checkpoint / Resume / Runtime Session Management
  • src/memory/: Private Memory + SuperMemory Pipeline
  • src/observability/: Telemetry Collector / Query Plane / Replay & Trace Explanation
  • src/transport/: Structured Transport / Cross-environment Session Bridge
  • src/plugins/: Plugin Lifecycle Management
  • src/skills/: Skill Registry / Capability Build Pipeline
  • src/services/: Service Mediation & Orchestration Spine
  • tests/: E2E Testing & Regression Test Suite
  • deploy/scripts/: Automated Acceptance & Operational Scripts
  • docs/: Protocol Specifications / Architecture Design / Acceptance Standards

Current Status

The current release is in a high-maturity engineering stage: core pipeline runnable, data traceable, audit log explainable. It is recommended to complete full regression verification via deploy/scripts/ first, then gradually scale production workloads.

CLI Specification Document:

  • docs/CLI_SPECIFICATION.md

Signal Whitebox Commands (D9/D10/D11)

The internal signal pipeline is exposed via white-box CLI commands for debugging and inspection:

cargo run --manifest-path .\Cargo.toml -- system signal status
cargo run --manifest-path .\Cargo.toml -- system signal explain --trace-id <trace-id>
cargo run --manifest-path .\Cargo.toml -- system signal drain

Implementation Governance Rules:

  • All business-side signal writes must pass through the unified SignalFacade layer.
  • Direct bypass write operations are blocked by static scan tests.
  • Signal pipeline acceptance is integrated into week6_acceptance full regression scripts.

Rule Reference Policy

Third-party reference materials under the rule/ directory are only used for architecture benchmarking and theoretical abstraction. These files are excluded from build dependencies and runtime execution.