-
Notifications
You must be signed in to change notification settings - Fork 0
Home
github-actions[bot] edited this page Mar 8, 2026
·
35 revisions
DeepSigma is an execution-governance and coherence-operations platform for AI and human decision systems. It applies pre-execution authority gates, policy-bound execution contracts, and sealed decision evidence to ensure governance is enforceable before state changes, then routes outcomes into the Drift → Patch learning loop.
If you only read one thing: governance must constrain execution, not just describe it afterward.
| Page | Purpose |
|---|---|
| Wiki Index | Canonical wiki structure and navigation |
| Quickstart | Get a supervised agent running in minutes |
| Concepts | The problems Coherence Ops solves and the primitives that solve them |
| Architecture | System diagram and component map |
| FAQ | Common questions |
The runtime enforces four contracts on every decision before it is sealed. After sealing, outcomes flow through the CERPA adaptation loop (Claim → Event → Review → Patch → Apply).
| Page | What it covers |
|---|---|
| Contracts | All four contract types: DTE, Freshness, Action, Verification |
| DTE Schema | Time budgets, stage limits, degrade triggers |
| Action Contract Schema | Blast radius, idempotency, rollback, authorization modes |
| Degrade Ladder | Six rungs from warn → block; how the supervisor degrades under pressure |
| Verifiers | Postcondition checks (read_after_write, invariant_check, custom) |
| Sealing & Episodes | Immutable DecisionEpisode envelope + SHA-256 seal |
| Runtime Flow | Step-by-step request lifecycle |
| Policy Packs | Versioned bundles of DTE + Action constraints |
| Authority Boundary Primitive | Pre-runtime governance declaration — what's allowed/denied/required before enforcement |
| Context Envelope | Ambient context (WHO/WHEN/WHERE/WHY/CONSTRAINTS/SCOPE) wrapping CERPA cycles — NOT a primitive |
| Page | What it covers |
|---|---|
| CERPA | Foundational adaptation loop: Claim → Event → Review → Patch → Apply |
| Drift → Patch | How drift signals become structured Patch Packets |
| Drift Schema | 10 drift types, 3 severity levels, fingerprint dedup |
| Coherence Ops Mapping | DLR / RS / DS / MG — the four governance artifacts |
| IRIS | Operator query engine: WHY / WHAT_CHANGED / WHAT_DRIFTED / RECALL / STATUS |
| Unified Atomic Claims | Claim primitive — the unit of institutional memory |
| Canon | Blessed claim memory and canon entry lifecycle |
| Retcon | Retroactive claim correction with full audit trail |
| LLM Data Model | How LLM interactions map to the governance schema |
| Page | What it covers |
|---|---|
| Creative Director Suite | Dataset, workbook template, generator, quickstart |
| Excel-First Governance | BOOT protocol, 7 table schemas, 6-lens prompting, writeback contract |
Six executable domain mode modules (93 function handlers total) with cross-domain cascade propagation, event contracts, deterministic replay, portable DecisionSurface runtime, and Drift Radar operational surface.
| Page | What it covers |
|---|---|
| IntelOps | 12 handlers (INTEL-F01–F12): claim ingest → validate → drift → patch → MG update → canon promote |
| FranOps | 12 handlers (FRAN-F01–F12): canon propose/bless/enforce, retcon assess/execute/propagate, inflation |
| ReflectionOps | 19 handlers (RE-F01–F19): episodes, gates, killswitch, severity, audit, coherence, IRIS, replay + institutional memory (precedent, fingerprint, knowledge consolidation, temporal recall, decay) |
| AuthorityOps | 19 handlers (AUTH-F01–F19): action intake, actor/resource resolve, policy, DLR presence, assumption validate, blast radius + simulation + propagation + seal, kill-switch, decision gate, audit, delegation, authority drift + history + cross-domain correlation, assumption sweep |
| ParadoxOps | 12 handlers (PDX-F01–F12): tension set create, pole manage, dimension attach/shift, pressure/imbalance compute, threshold evaluate, drift promote, inter-dimensional drift, seal, patch, lifecycle |
| ActionOps | 19 handlers (ACTION-F01–F19): commitment intake/validate, deliverable track, deadline check, compliance evaluate, risk assess, breach detect, escalation, remediation, adjust, complete, report + decision accounting (cost, time-to-decision, value, debt, ROI, budget) |
| DecisionSurface | Portable Coherence Ops runtime — adapter ABC, claim-event engine, notebook/CLI/Vantage adapters |
| Drift Radar | Cross-domain drift intelligence surface — correlation, trending, forecasting, remediation prioritization |
| Cascade Engine | 27 cross-domain rules with depth-limited propagation |
| Event Contracts | Routing table: 79 functions + 91 events → FEEDS topics/subtypes/handlers |
| Money Demo v2 | 10-step end-to-end pipeline exercising all 6 domain modes |
Log-agnostic coherence engine for external telemetry with cross-environment federation.
| Page | What it covers |
|---|---|
| JRM Pipeline | 3 adapters (Suricata, Snort, Copilot), 5-stage pipeline, JRM-X packet builder, CLI |
| JRM Federation | Gate validation, hub drift detection, advisory engine, HMAC-SHA256 signing |
Event-driven pub/sub connecting governance primitives (TS, ALS, DLR, DS, CE) via file-based bus with manifest-first ingest, authority validation, triage state machine, and canon versioning.
| Stage | What it does |
|---|---|
| Event Envelope | 7 schemas, 6 golden fixtures, SHA-256 payload hashing, two-phase validation |
| File-Bus | Atomic publisher, poll subscriber, DLQ + replay, multi-worker safety |
| Ingest | Manifest-first, hash verification, atomic staging, PROCESS_GAP drift on failure |
| Consumers | Authority gate (DLR vs ALS), evidence completeness, SQLite triage store |
| Canon | Append-only store, claim validator, MG writer, supersedes chain |
Captures AI interaction exhaust (prompts, completions, tool calls, metrics) and routes it into the governance pipeline automatically.
| Page | What it covers |
|---|---|
| Exhaust Inbox | Full feature docs: adapters, API, LLM extraction, coherence scoring |
Adapters: LangChain · Anthropic direct · Azure OpenAI batch
API: 10 REST endpoints — ingest → assemble → refine → commit
LLM extraction: EXHAUST_USE_LLM=1 enables Anthropic-backed bucket extraction with rule-based fallback
| Schema | Description |
|---|---|
| Episode Schema | Sealed DecisionEpisode — the core output |
| DTE Schema | Decision Timing Envelope |
| Action Contract Schema | Safe action constraints |
| Drift Schema | Drift signal structure |
| Policy Pack Schema | Policy bundle format |
| Claim Schema | Unified Atomic Claim |
| Canon Schema | Canon entry format |
| Retcon Schema | Retroactive correction record |
| ABP Schema | Authority Boundary Primitive v1 |
| Schemas | Full index of all JSON Schema specs |
| Integration | Page |
|---|---|
| SDK Packages | SDK-Packages |
| MCP (Model Context Protocol) | MCP |
| LangChain | LangChain |
| Palantir Foundry | Palantir-Foundry |
| Microsoft Power Platform | Power-Platform |
| OpenTelemetry | OpenTelemetry |
| Page | What it covers |
|---|---|
| Operations | Deployment, configuration, environment variables |
| SLOs & Metrics | Latency, drift rate, verification pass rate targets |
| Replay & Testing | Episode replay harness and test fixtures |
| Security | Threat model, seal integrity, authorization |
| Asset | Link |
|---|---|
| Latest release notes | v2.1.0 |
| KPI composite radar | release_kpis/radar_composite_latest.png |
| KPI delta table | release_kpis/radar_composite_latest.md |
| TEC summary (C-TEC v1.0) | release_kpis/TEC_SUMMARY.md |
| Mermaid canonical index | docs/mermaid/README.md |
| Mermaid archive index | docs/archive/mermaid/ARCHIVE_INDEX.md |
enterprise/
├─ artifacts/ # Sealed runs, templates, verifier bundles, authority ledger
├─ charts/ # Helm and deployment chart assets
├─ core_output/ # Core demo/runtime output pipeline stages
├─ dashboard/ # UI and dashboard services
├─ docker/ # Container build and runtime definitions
├─ docs/ # Enterprise documentation set
│ ├─ wiki/ # GitHub wiki source pages
│ ├─ mermaid/ # Canonical architecture/runtime diagrams
│ ├─ archive/ # Historical/archived design materials
│ ├─ release/ # Release notes and release documentation
│ ├─ examples/ # Demo and reference examples
│ └─ cookbook/ # Operational and integration recipes
├─ governance/ # Governance specs and operating artifacts
├─ ops/ # Monitoring/ops config (Grafana, Prometheus, etc.)
├─ pilot/ # Pilot data: assumptions, decisions, drift, patches, reports
├─ prompts/ # Prompt packs and prompt OS material
├─ release_kpis/ # KPI exports and telemetry summaries
├─ roadmap/ # Issue and milestone planning assets
├─ schemas/ # JSON Schema catalog (core + prompt_os + reconstruct)
├─ scripts/ # Tooling and automation scripts
├─ specs/ # Legacy/compatibility specs
├─ src/ # Enterprise Python packages and modules
└─ tests/ # Enterprise test suites and fixtures
Source of truth:
- Wiki source files:
enterprise/docs/wiki/*.md - Enterprise docs root:
enterprise/docs/
Exportable single-file HTML apps — zero dependencies, work offline, governance-ready.
| Page | What it covers |
|---|---|
| EDGE Modules | Full module listing, ABP gate enforcement, delegation review |
| JRM EDGE | Browser-based JRM pipeline explorer (v1.0.7): events, packets, health, test lab, stream mode, So What panel, policy drawer |
| RFP Co-Pilot | AI-assisted RFP extraction → JSON → Excel Power Query → 6 role action packets |
| RFP Co-Pilot Exec Brief | 1-page executive summary with Print/PDF |
| Page | What it covers |
|---|---|
| Glossary | All terms defined |
| Roadmap | Near-term and mid-term priorities |
| Contributing | How to contribute |
Σ OVERWATCH — Coherence Ops Platform • Current release: v2.1.0 • DeepSigma
- Start
- Core
- Schemas
- FEEDS + Exhaust
- Integrations
- Reference Layer
- Ops
- Excel-First
- EDGE + ABP
- Domain Modes
- Governance
- Meta