Skip to content

Garrus800-stack/genesis-agent

Repository files navigation


Genesis — Self-Modifying Cognitive AI Agent

A self-aware, self-verifying, self-evolving cognitive AI agent.
It reads its own source code. It fixes its own bugs. It builds its own features.
It verifies its own output programmatically. It thinks while you're away.
It feels the consequences of its actions. It pursues goals autonomously.
It learns what works for its specific model.


Version Tests Modules Services Phases Events MCP Languages Electron License CI


What is Genesis?

Genesis is not a framework for building agents. Genesis is the agent.

It's a self-contained Electron application that reads its own source code, modifies itself, verifies its own output programmatically, and learns from every interaction. It runs locally on your machine — no cloud platform required, no YAML configs, no SDK. Install, start, talk.

Think of it as an AI developer that lives on your desktop: it has a codebase (its own), a memory (episodic + semantic), emotions (that influence its decisions), and goals (that survive restarts). When idle, it dreams — consolidating memories into reusable schemas. When frustrated, it escalates to larger models. When curious, it explores its own architecture.

How it works

You:    "Build a REST API module for Genesis with tests"
Genesis: [PLAN] 6 steps via FormalPlanner (preconditions checked against WorldState)
         [EXPECT] P(success)=0.78 based on 47 prior code-gen outcomes
         [SIMULATE] Best path: direct (risk=0.22), vs. incremental (risk=0.31)
         [STEP 1/6] Analyzing existing code... ✓ verified programmatically
         [STEP 2/6] Generating module (87 lines)... ✓ AST parse valid, imports resolve
         [STEP 3/6] Writing tests... ✓ file created, syntax verified
         [STEP 4/6] Running npm test... ✓ exit 0, 12/12 passing
         [STEP 5/6] Git snapshot... ✓ committed
         [STEP 6/6] Wiring into AgentCore...
         [VERIFY] Goal complete. VerificationEngine: 5/5 PASS, 0 AMBIGUOUS.
         [SURPRISE] Low (0.12) — outcome matched expectation
         [SIGN] Module signed: HMAC-SHA256 (abc12def...)
         [LEARN] MetaLearning recorded: json-schema prompt @ temp 0.3 → success
         "Done. Module + 12 tests, all passing."

Every step is verified by the machine, not the LLM. AST parsing, exit codes, file validation, import resolution — the LLM proposes, deterministic checks verify. Only ambiguous quality judgments fall back to LLM evaluation.

What makes it different

Capability What Genesis does What typical AI tools do
Self-modification Reads its own AST, plans changes, tests in sandbox, snapshots with git, applies only if tests pass Run user-provided code
Verification 66 programmatic checks — AST, exit codes, imports, signatures — LLM is last resort Trust the LLM output
Memory 5-layer system — episodic, semantic, vector, conversation, knowledge graph — with intelligent forgetting Chat history window
Planning FormalPlanner with preconditions, mental simulation, probabilistic branching, failure taxonomy Sequential function calling
Learning Tracks success rates by model/prompt/temperature, auto-optimizes — A/B tests its own prompts Static prompts
Autonomy Pursues multi-step goals, survives restarts, graduates its own trust level (0–3) Single-turn responses
Cognition Expectations, surprise, dreams, working memory, autobiographical identity, emotional steering None

Capabilities at a glance

Autonomous execution — FormalPlanner with typed action steps, precondition checking against live WorldState, mental simulation with probabilistic branching, goal persistence across restarts, failure taxonomy with 4 recovery strategies, cooperative cancellation, working memory per goal.

Self-modification — reads its own source via SelfModel, plans changes via SelfModificationPipeline, tests in dual-mode sandbox (VM + process), snapshots with git, HMAC-SHA256 module signing, hot-reloads without restart.

Verification — 66-test VerificationEngine covering AST syntax, import resolution, dangerous patterns, test exit codes, file integrity, module signatures. The LLM proposes — the machine verifies.

Memory & learning — 5-layer memory (conversation, episodic, vector, unified, knowledge graph), adaptive forgetting (surprise amplifies retention 5×), DreamCycle consolidation during idle time, MetaLearning prompt optimization, PromptEvolution A/B testing, OnlineLearner real-time feedback (streak detection, model escalation, temperature tuning), LessonsStore cross-project persistent learning.

Cognition & consciousness — ExpectationEngine (quantitative predictions), SurpriseAccumulator (information-theoretic), PhenomenalField (unified awareness every 2s), TemporalSelf (past/future continuity), IntrospectionEngine (3-level meta-awareness), CognitiveWorkspace (9-slot transient working memory).

Organism — 5 emotional dimensions, homeostasis (6 vitals), 4 needs (social, mastery, novelty, rest), metabolism (500 AU energy pool), heritable genome (6 evolvable traits), epigenetic conditioning, immune system (anomaly detection).

Infrastructure — 13-phase DI boot, EventBus (279 events), MCP with adaptive discovery, CircuitBreaker per connection, CorrelationContext tracing, PeerNetwork (AES-256-GCM), 10-layer defense-in-depth security.

For the full feature list with version history, see CAPABILITIES.md.


Quick start

New to Genesis? Read the Quick Start Guide — it walks you through your first conversation, your first goal, and self-modification in under 5 minutes.

Option A — Cloud API (recommended for best results):

git clone https://github.com/Garrus800-stack/genesis-agent.git
cd genesis-agent
npm install
npm start

Then open Settings → paste your Anthropic API key or OpenAI API key. Genesis auto-detects and selects the best available model.

Option B — Local with Ollama (fully offline, private):

ollama pull qwen2.5:7b   # or gemma2:9b, deepseek-r1:8b, llama3.1:8b, etc.
ollama serve

git clone https://github.com/Garrus800-stack/genesis-agent.git
cd genesis-agent
npm install
npm start

Option C — Hybrid (best of both):

Run Ollama locally AND configure a cloud API key. Genesis uses cloud for complex reasoning tasks and auto-failovers to local when cloud is unavailable.

Boot profiles

Genesis supports three boot profiles for different complexity levels:

npm start                      # Full — all 113 services, all cognitive systems
npm start -- --cognitive       # Cognitive — skip consciousness (~90 services)
npm start -- --minimal         # Minimal — core agent loop only (~50 services)

Use --minimal to learn the architecture without cognitive overhead. Use --cognitive for development. Use --full (default) for production.

Requires Node.js 20+ (tested on 20, 22) and Git. Ollama is optional if a cloud API is configured. On Windows, double-click Genesis-Start.bat instead.

Supported backends

Backend Models Config
Anthropic Claude Opus 4, Sonnet 4, Haiku 4.5 Settings → models.anthropicApiKey
OpenAI-compatible GPT-4o, GPT-4, o1, or any compatible API Settings → models.openaiApiKey + models.openaiBaseUrl
Ollama (local) Any model Ollama supports (gemma2, qwen2.5, deepseek, llama, mistral, ...) Auto-detected on 127.0.0.1:11434

Genesis automatically selects the best model: user-preferred → cloud → local. Override via Settings → models.preferred.


Architecture

Thirteen layers with clear boundaries — star topology where every layer depends only on core/ and ports/, never on each other. The kernel is immutable. Critical safety files are hash-locked. Everything else is fair game for self-modification. v5.3.0: zero cross-layer violations, zero orphans, zero phantom late-bindings.

┌─────────────────────────────────────────────────────────────┐
│  🖥️  UI Layer          Chat + Monaco Editor + Dashboard     │
├─────────────────────────────────────────────────────────────┤
│  🌌 Consciousness [P13] PhenomenalField · TemporalSelf      │
│                         IntrospectionEngine · AttentionalGate│
│                         ConsciousnessExtension (6 subsystems)│
├─────────────────────────────────────────────────────────────┤
│  🔮 Hybrid [P12]       GraphReasoner · AdaptiveMemory       │
├─────────────────────────────────────────────────────────────┤
│  🌐 Extended [P11]     TrustLevels · Effectors · WebPercept │
│                         SelfSpawner · GitHubEffector          │
├─────────────────────────────────────────────────────────────┤
│  🏛️ Agency [P10]       GoalPersistence · FailureTaxonomy    │
│                         DynamicContextBudget · LocalClassifier│
│                         EmotionalSteering · FitnessEvaluator │
├─────────────────────────────────────────────────────────────┤
│  🧠 Cognitive [P9]     Expectations · Simulation · Surprise │
│                         DreamCycle (phase delegates) v5.2    │
│                         SelfNarrative · PromptEvolution v5.2 │
│                         CognitiveWorkspace (SA-P6)     v5.3 │
│                         OnlineLearner (SA-P5)          v5.3 │
│                         LessonsStore (SA-P7)           v5.3 │
├─────────────────────────────────────────────────────────────┤
│  ⚡ Revolution [P8]     FormalPlanner · AgentLoop + Cancel   │
│                         ModelRouter · VectorMemory v5.2      │
├─────────────────────────────────────────────────────────────┤
│  🧬 Organism [P7]      Emotions (5D) · Homeostasis (6 vitals)│
│                        Genome · Epigenetic · Fitness         │
│                         NeedsSystem (4 drives) · Metabolism   │
├─────────────────────────────────────────────────────────────┤
│  🛡️  Autonomy [P6]     IdleMind · Daemon · HealthMonitor    │
│                         HealthServer (optional) v5.2         │
├─────────────────────────────────────────────────────────────┤
│  🔗 Hexagonal [P5]     ChatOrchestrator · SelfModPipeline   │
│                         EpisodicMemory · PeerNetwork          │
├─────────────────────────────────────────────────────────────┤
│  📋 Planning [P4]       GoalStack · MetaLearning · SchemaStore│
├─────────────────────────────────────────────────────────────┤
│  🔧 Capabilities [P3]  ShellAgent · MCP (CodeExec delegate)  │
│                         PluginRegistry · HotReloader v5.2    │
├─────────────────────────────────────────────────────────────┤
│  🧩 Intelligence [P2]  VerificationEngine · CodeSafetyScanner│
│                         IntentRouter · ContextManager (table) │
│                         CircuitBreaker v5.2                   │
├─────────────────────────────────────────────────────────────┤
│  📦 Foundation [P1]     ModelBridge · Sandbox · WorldState    │
│                         KnowledgeGraph · ModuleSigner · Cache │
│                         CorrelationContext v5.2               │
├─────────────────────────────────────────────────────────────┤
│  🔗 Ports               LLM · Memory · KG · Sandbox ·       │
│                         CodeSafety (hexagonal boundary) v5.2 │
├─────────────────────────────────────────────────────────────┤
│  🔒 KERNEL (immutable)  SafeGuard · IPC Contract · Hashes    │
│      + 🔐 Hash-Locked   Scanner · Verifier · Constants       │
└─────────────────────────────────────────────────────────────┘

Kernel (immutable): main.js, preload.js, src/kernel/. SHA-256 hashed at boot, verified periodically.

Critical Safety Files (hash-locked): CodeSafetyScanner, VerificationEngine, Constants, EventBus, Container — locked via SafeGuard.lockCritical(). The agent cannot weaken the modules that enforce its own safety.

Agent Core: Self-modifiable modules — read, analyze, modify, hot-reload — but only after sandbox testing, safety scanning, and git snapshots.

Cognitive Layer (v4.0+v5.2): Expectation formation, mental simulation, surprise-driven learning, memory consolidation (with phase delegates), autobiographical identity, prompt evolution via A/B testing.


The Cognitive Loop

Every autonomous step follows a five-phase cycle:

Perceive (WorldState) → Plan (FormalPlanner) → Act (AgentLoop)
    → Verify (VerificationEngine) → Learn (MetaLearning + EpisodicMemory)

The VerificationEngine returns PASS, FAIL, or AMBIGUOUS. Only AMBIGUOUS falls back to LLM judgment. Everything else is deterministic.

What's checked How LLM?
Code syntax AST parse (acorn) No
Imports Filesystem check No
Dangerous patterns AST walk + regex No
Test results Exit code + assertion count No
Shell commands Exit code + timeout + patterns No
File writes Existence + non-empty + encoding No
Plan preconditions WorldState API No
Module integrity HMAC-SHA256 signature v4.0 No
Subjective quality AMBIGUOUS only

Phase 9: The Cognitive Meta-Loop v4.0

Expect → Simulate → Act → Surprise → Learn → Dream → Schema → better Expect

ExpectationEngine — Quantitative predictions using MetaLearning statistics and SchemaStore patterns. No LLM calls.

MentalSimulator — Plan sequences in-memory against cloned WorldState with probabilistic branching and risk scoring.

SurpriseAccumulator — Prediction error via information-theoretic surprise (−log₂P). High surprise amplifies learning up to 4×.

DreamCycle — Idle-time memory consolidation. Phases 1-4 are heuristics; Phase 5 uses one batched LLM call. v5.2: Schema extraction, value crystallization, and DreamEngine corroboration extracted to phase delegates.

SelfNarrative — Autobiographical identity summary injected into every LLM call (~200 tokens of metacognitive context).

PromptEvolution v5.2 — A/B testing framework for prompt sections. Runs controlled experiments (25+ trials per arm), auto-promotes variants with statistically significant improvement (≥5%). Identity and safety sections are immutable by design.

Phase 10: Persistent Agency v4.1

Goals that survive reboots. Errors that get classified. Context that adapts. Emotions that steer.

Goal created → Checkpoint → [Restart] → Resume → Continue
Error caught → FailureTaxonomy.classify() → Strategy (retry | replan | escalate | update-world)
Intent detected → DynamicContextBudget.allocate() → Optimized token distribution
Emotion shifts → EmotionalSteering.getSignals() → ModelRouter / FormalPlanner / IdleMind adjusts
LLM fallback → LocalClassifier.addSample() → Next time: no LLM needed (2-3s saved)

GoalPersistence — Active goals serialized to disk after every step. Crash recovery via step-level checkpoints. Resume prompt on boot.

FailureTaxonomy — Four error categories with distinct recovery strategies. Transient errors get exponential backoff. Deterministic errors trigger immediate replan. Environmental errors update WorldState. Capability errors escalate to a larger model.

DynamicContextBudget — Intent-based token allocation replaces fixed budgets. Code-gen gets 55% code context. Chat gets 40% conversation. Learns from MetaLearning success rates.

EmotionalSteering — Emotions become functional control signals. Frustration > 0.65 → ModelRouter tries larger model. Energy < 0.30 → FormalPlanner caps plans at 3 steps. Curiosity > 0.75 → IdleMind prioritizes exploration.

LocalClassifier — TF-IDF + cosine similarity classifier trained from IntentRouter's own LLM fallback log. After ~30 samples, handles 60-80% of classifications without LLM calls.

Phase 11: Extended Perception & Action v4.1

Genesis sees beyond the filesystem. Acts beyond its project directory. Graduates its own autonomy.

TrustLevelSystem — Four trust levels: Supervised (everything needs approval), Assisted (safe actions auto-execute), Autonomous (only high-risk needs approval), Full Autonomy (only safety invariants block). MetaLearning data can suggest auto-upgrades for specific action types with >90% success over 50+ attempts.

EffectorRegistry — Typed, verifiable, approval-gated actions for the outside world. Built-in: clipboard, OS notifications, browser open, external file write. Plugin: GitHubEffector (issues, PRs, comments). Each effector has risk level, preconditions, optional verification and rollback.

WebPerception — Lightweight HTTP fetch with HTML parsing (cheerio optional). Headless browser mode via Puppeteer (optional). Results cached, fed into WorldState.external for prompt context.

SelfSpawner — Fork-based worker processes for parallel sub-tasks. Each worker gets minimal context (ModelBridge config, focused goal) and runs independently with timeout and memory limits. Up to 3 concurrent workers.

Phase 12: Symbolic + Neural Hybrid v4.1

Not everything needs an LLM. Graph reasoning and intelligent forgetting.

GraphReasoner — Deterministic queries over the KnowledgeGraph without LLM calls. Transitive dependency chains, impact analysis ("if I change EventBus, what breaks?"), cycle detection, shortest path between concepts, contradiction detection. Integrated into ReasoningEngine — structural questions are answered in milliseconds instead of seconds.

AdaptiveMemory — Differentiated forgetting based on emotional valence, surprise magnitude, and access frequency. High-surprise memories decay 5× slower. Routine conversations decay quickly. Frequently accessed memories are reinforced. Consolidation runs during DreamCycle.


Security model

┌──────────────────────────────────────────────────────────────────┐
│  KERNEL (immutable) — SHA-256 hashes, periodic integrity checks  │
├──────────────────────────────────────────────────────────────────┤
│  CRITICAL FILE HASH-LOCK — Scanner, Verifier, Constants, Bus    │
├──────────────────────────────────────────────────────────────────┤
│  TRUST LEVEL SYSTEM — graduated autonomy (0-3)             v4.1  │
├──────────────────────────────────────────────────────────────────┤
│  MODULE SIGNING — HMAC-SHA256 for self-modified files       v4.0 │
├──────────────────────────────────────────────────────────────────┤
│  AST CODE SAFETY SCANNER — acorn AST walk + regex fallback       │
│  └─ CodeSafetyPort — hexagonal adapter (DI-injected)       v5.2  │
├──────────────────────────────────────────────────────────────────┤
│  VERIFICATION ENGINE — programmatic truth, 66 tests              │
├──────────────────────────────────────────────────────────────────┤
│  SANDBOX — dual-mode isolation (Process + VM)              v4.0  │
│  └─ WORKER THREAD — MCP code exec in resource-limited     v5.1  │
│     worker_thread (64MB heap, hard-kill on timeout)               │
├──────────────────────────────────────────────────────────────────┤
│  CIRCUIT BREAKER — per-MCP-connection failure isolation     v5.2  │
│  └─ CLOSED → OPEN → HALF_OPEN → CLOSED state machine            │
├──────────────────────────────────────────────────────────────────┤
│  IMMUNE SYSTEM — self-modification anomaly detection        v5.0  │
├──────────────────────────────────────────────────────────────────┤
│  EFFECTOR REGISTRY — risk-gated external actions           v4.1  │
├──────────────────────────────────────────────────────────────────┤
│  CORRELATION CONTEXT — causal tracing via AsyncLocalStorage v5.2  │
├──────────────────────────────────────────────────────────────────┤
│  UI ERROR BOUNDARY — global error/rejection handler        v4.0  │
├──────────────────────────────────────────────────────────────────┤
│  IPC RATE LIMITER — per-channel token bucket (kernel-space)      │
├──────────────────────────────────────────────────────────────────┤
│  SHELL AGENT — 4 tiers, blocklist, rate limiter, no injection    │
├──────────────────────────────────────────────────────────────────┤
│  PEER NETWORK — AES-256-GCM, PBKDF2 600K iterations             │
└──────────────────────────────────────────────────────────────────┘

Testing

npm test              # All tests (135 suites)
npm run test:coverage # With coverage report (c8)
npm run ci            # Full CI: tests + event audit + channel audit + fitness gate

All tests run without external dependencies (no Ollama, no API keys, no internet). Tested on Node 18, 20, 22. CI runs on Ubuntu + Windows via GitHub Actions.

Code Metrics by Layer

Layer Files LOC Purpose
Core 13 ~4,000 EventBus, Container, Constants, Logger, CorrelationContext, CircuitBreaker, CancellationToken
Foundation 29 ~8,400 ModelBridge, Backends, Sandbox, KG, WorldState, TrustLevels, Telemetry, LLMCache
Intelligence 16 ~6,200 Verification, Safety Scanner, Intent, Reasoning, ContextManager, PromptBuilder, PromptEvolution
Capabilities 18 ~5,400 Shell, MCP (Client + Transport + CodeExec + Worker + Server), HotReload, Skills, Plugins, WebPerception
Planning 9 ~2,900 Goals, GoalPersistence, MetaLearning, SchemaStore, ValueStore
Hexagonal 14 ~5,700 ChatOrchestrator, SelfModPipeline, Memory (Unified + Episodic + Adaptive), PeerNetwork, PeerConsensus
Autonomy 6 ~2,600 IdleMind, Daemon, HealthMonitor, HealthServer, CognitiveMonitor, ErrorAggregator
Organism 12 ~4,600 Emotions (5D), Homeostasis (6 vitals), Needs (4 drives), Metabolism, ImmuneSystem, Genome, Epigenetic, Fitness
Revolution 14 ~5,600 AgentLoop (+ Steps/Planner delegates), FormalPlanner, HTN, VectorMemory, NativeToolUse, ModelRouter
Cognitive 9 ~3,900 DreamCycle (+ phase delegates), ExpectationEngine, SurpriseAccumulator, MentalSimulator, SelfNarrative, CognitiveWorkspace
Consciousness 13 ~6,000 PhenomenalField (+ Computation delegate), TemporalSelf, IntrospectionEngine, AttentionalGate, EchoicMemory, DreamEngine
Ports 6 ~800 Hexagonal adapters (LLM, Memory, Knowledge, Sandbox, CodeSafety)
Total 196 ~68,700 (all src/, incl. UI + kernel)

Project Stats

Metric Value
Source modules 196 JS files in src/
Lines of code ~67k src + ~28k test
Manifest phases 13 (+ Phase 0 bootstrap)
DI services 107 (manifest + auto-discovery)
Late-bindings 174 (150 optional, 24 required)
Test suites 135 files, 2500 tests (coverage gates: 60/50/55)
Dependencies 4 production + 3 optional + 5 dev
LLM backends 3 (Anthropic, OpenAI-compatible, Ollama)
Anthropic models 3 (Opus 4, Sonnet 4, Haiku 4.5)
IPC channels 39 invoke + 1 send + 6 receive (rate-limited)
Event types 275 across 76 namespaces (catalogued in EventTypes.js)
Cross-layer event flows 42 (via EventBus, no direct imports)
Hexagonal ports 5 (LLM, Memory, Knowledge, Sandbox, CodeSafety)
Cognitive modules 8 (ExpectationEngine, MentalSimulator, SurpriseAccumulator, DreamCycle, SelfNarrative, CognitiveHealthTracker, CognitiveWorkspace, OnlineLearner, LessonsStore)
Consciousness modules 13 (PhenomenalField, TemporalSelf, IntrospectionEngine, AttentionalGate, EchoicMemory, PredictiveCoder, NeuroModulators, SalienceGate, DreamEngine, ConsciousnessState + 3 adapters/delegates)
Organism 5 emotional dimensions + homeostasis + allostasis + 4 needs + steering + metabolism + immune system + heritable genome + epigenetic conditioning + fitness evaluation
Safety layers 10 (kernel lock → AST scan → port → sandbox → worker → circuit breaker → immune → trust → validateWrite → blocklist)
Trust levels 4 (supervised → full autonomy)
Languages EN primary (+ DE, FR, ES via i18n)
Architectural fitness Cross-layer: 0, Orphans: 0, Phantoms: 0, CC>30: 21

Memory Architecture

Genesis has a five-layer memory system, unified through a facade:

┌─────────────────────────────────────────────────────────┐
│  UnifiedMemory (read facade over all layers)            │
├─────────────────────────────────────────────────────────┤
│  AdaptiveMemory [P12] Retention scoring + smart decay    │
│  VectorMemory        Embedding-based semantic search     │
│  EpisodicMemory      Timestamped experiences + causality │
│  ConversationMemory  Chat history + episode extraction   │
│  KnowledgeGraph      Concept nodes + typed relations     │
│  WorldState          Live snapshot of system state        │
├─────────────────────────────────────────────────────────┤
│  SchemaStore [P9]    Abstract patterns from DreamCycle   │
└─────────────────────────────────────────────────────────┘

All persistence goes through StorageService (write-queued, atomic JSON writes). The EmbeddingService is optional — without it, search degrades to keyword matching.


Known Limitations

  • No TypeScripttypes/core.d.ts and types/cognitive.d.ts exist for IDE support, but the codebase is pure CommonJS JS. Type-safety relies on JSDoc and convention. TypeDoc config is provided for API doc generation.
  • VM sandbox is not a true sandboxvm.createContext provides isolation for quick evals but is explicitly NOT security-grade for untrusted code. Untrusted code must use process-mode execute() or Linux namespace isolation.
  • sandbox: false on Electron <35 — CJS preload requires require() which is blocked by sandbox:true. contextIsolation:true is the primary security boundary. See preload.mjs for ESM preload implementation.
  • Single-instance storage — StorageService serializes writes but all autonomous systems share the same .genesis/ directory.
  • Consciousness metrics are heuristic — Phi (integrated information) in PhenomenalField is a simplified proxy, not a rigorous IIT implementation.

Dependencies

{
  "acorn": "^8.16.0",
  "chokidar": "^3.6.0",
  "electron": "^33.0.0",
  "monaco-editor": "^0.44.0",
  "tree-kill": "^1.2.2"
}

No LangChain. No LlamaIndex. Everything self-written.


Documentation

Architecture & Design

Document What it covers
QUICK-START.md Start here — first conversation, first goal, self-modification, boot profiles
ARCHITECTURE-DEEP-DIVE.md Technical deep-dive — boot sequence, DI container, service lifecycle, data flows
EVENT-FLOW.md EventBus event map — which modules emit/consume which events
CAPABILITIES.md Complete feature overview — what Genesis can do, organized by category
COMMUNICATION.md How Genesis instances communicate — IPC, EventBus, PeerNetwork, MCP
DEGRADATION-MATRIX.md What breaks if each service is missing — auto-generated

Cognitive & Consciousness

Document What it covers
phase9-cognitive-architecture.md Phase 9 design — DreamCycle, Expectations, SelfNarrative, MentalSimulator
phase9-integration-review.md Integration review and test results for Phase 9
consciousness-extension-architecture.md Phase 12-13 — closed perceptual loop, EchoicMemory, PredictiveCoder, NeuroModulators, DreamEngine

Operations

Document What it covers
TROUBLESHOOTING.md Common problems and solutions — install, boot, LLM, self-modification, platform-specific
ROADMAP-v6.md Development roadmap — completed phases, open items, deferred proposals
AUDIT-BACKLOG.md Architectural health tracking — resolved issues, monitor items, fitness metrics
SELF-REVIEW-v5.0.0.md Cross-reference of Genesis's self-reflection against code-level evidence

Contributing & Security

Document What it covers
CONTRIBUTING.md How to contribute — conventions, security rules, PR process, service templates
SECURITY.md Security policy — 7-layer defense model, threat model, vulnerability reporting
CHANGELOG.md Version history with detailed per-release notes

Schemas & Configuration

Document What it covers
schemas/skill-manifest.schema.json JSON Schema for third-party skill/plugin manifests
typedoc.json TypeDoc config — run npx typedoc to generate API docs

Contributing

See CONTRIBUTING.md for the full guide — architecture, conventions, security rules, and PR process.

License

MIT © Daniel (Garrus800-stack)


Genesis doesn't just answer questions. It persists its goals, reasons over its own graph, feels the consequences of failure, and graduates its own autonomy.

About

Self-aware cognitive AI agent that reads, modifies & verifies its own code. Autonomous planning, episodic memory, emotional state & MCP integration. Runs on Claude, GPT-4 or Ollama. Electron desktop app for Windows, macOS & Linux.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages