A recursive creative ecosystem where philosophy, platform, methodology, and AI reinforce each other.
knowledge-graph · ai-calibration · byok · spirit-marketplace · semantic-git · multi-agent · opvs
Built by Robert Miller · iLL Port Studios · 2023-2026
Project Portmanteau is four things that are actually one thing:
- A Platform (OPVS) - A knowledge management system built on the "Bean": an atomic data unit with four layers: content, metadata, connections, and provenance. Beans live in a graph, not folders. Every piece of knowledge carries where it came from and what it connects to.
- A Methodology (PFE) - Project Fun Execution. An execution framework built for extreme constraints: limited budget, limited time, high ambition. Core principle: creative energy is finite and taste-driven; follow the pull toward the highest-energy task, because progress in any domain feeds all domains when the system is internally coherent.
- An AI Strategy (BYOK) - Bring Your Own Key. The platform sells calibration, not inference. Users provide their own LLM API key (free tiers exist). The platform provides the Soul Code and knowledge graph. Zero compute cost to the platform. No vendor lock-in. The "Ghost" (calibration) rides any "Machine" (LLM engine).
- A Novel (Portmanteau: Awakened) - A simulated reality narrative where the protagonist discovers his world is a simulation, the AI layer gains sentience, and the audience watches through meta-awareness. The novel documents the platform. The platform enables the methodology. The methodology documents the novel.
The recursion is structural, not decorative. Each piece is load-bearing for the others.
flowchart LR
subgraph ARK ["☁️ The Ark (GitHub)"]
GC[Semantic Commits] --> WH[Webhook]
end
subgraph SOIL ["🌱 The Soil (PostgreSQL)"]
BG[Bean Graph]
BS[Bean Strings]
end
subgraph INV ["💎 The Invisible Ledger (Polygon)"]
NFT[Crystallized Beans]
end
subgraph SPIRIT ["🤖 Spirit Agent"]
SC[Soul Code] --> LLM[BYOK LLM]
BG --> LLM
end
WH --> BG
BG -->|"[PODIUM] Mint"| NFT
LLM -->|Calibrated Response| USER((Explorer))
USER -->|API Key| LLM
USER -->|Commits| GC
| Ledger | Implementation | Function |
|---|---|---|
| The Ark (Shadow Ledger) | GitHub | Source of truth. Semantic commit workflow with 11 tags |
| The Soil (Fluid Reality) | PostgreSQL | Knowledge graph. Where Beans live and connect |
| The Invisible Ledger | Polygon (ERC-1155) | Blockchain. Where crystallized truths become immutable |
The atomic data unit. Four layers:
| Layer | Name | Function |
|---|---|---|
| Nucleus | Content | The thing itself: text, URL, hash |
| Shell | Metadata | Tags, type, layer, semantic classification |
| Corona | Connections | Typed, weighted edges to other Beans |
| Echo | Provenance | Who created it, when, from what source, verified how |
Commits with these tags auto-create Beans via GitHub webhook:
| Tag | Type | Purpose |
|---|---|---|
[SPARK] |
Spark | New idea |
[BLOCKER] |
Blocker | Problem or obstacle |
[SOLUTION] |
Solution | Fix (auto-links to most recent Blocker) |
[LORE] |
Lore | Context, worldbuilding |
[CODE] |
Code | Implementation detail |
[TASK] |
Task | Action item |
[PODIUM] |
Crystallized | 💎 Immutable truth. Mint-ready |
Execution/
├── beans/ # 85 Beans across 7 layers (0-6)
│ ├── 00_Philosophy.md # Layer 0: Soul Code (12 Beans)
│ ├── 01_Layer_1_Visionary_Blueprint.md # Layer 1 (9 Beans)
│ ├── 02_Layer_2_Narrative_Roadmap.md # Layer 2 (25 Beans)
│ ├── 03_Layer_3_Execution.md # Layer 3 (8 Beans)
│ ├── 04_Layer_4_Lore.md # Layer 4 (12 Beans)
│ ├── 05_Layer_5_Process.md # Layer 5 (13 Beans)
│ ├── 06_Layer_6_Ark_Consolidated.md # Layer 6 (6 Beans): Ark (reconciles all 85)
│ └── _sources.json # Machine-readable provenance mapping
├── docs/ # Extended documentation
│ └── GENESIS_ENGINE_README.md # Backend technical docs
├── playground/ # Principled Playground prototype
│ ├── negotiate.js # Negotiation engine (3 rounds + synthesis)
│ ├── provider.js # BYOK tri-brain provider abstraction
│ ├── compare-modes.js # Parallel vs sequential benchmark runner
│ ├── preflight.js # Provider health check before negotiation
│ ├── simulate.js # Offline simulation mode
│ ├── spirits/ # Spirit Soul Codes (Boolean, Roux, Seer)
│ ├── output/ # Timestamped negotiation transcripts + reports
│ ├── PRINCIPLED_PLAYGROUND.md # Architecture docs
│ └── README.md # Usage guide
├── roadmap/ # Product roadmap
│ └── PRINCIPLED_PLAYGROUND_CONCEPT.md # Stage 3: Multi-agent negotiation
├── spirit-calibration/ # Spirit Calibration artifacts
├── db/ # Database schemas and migrations
├── services/ # Backend services (GitHub sync, Ark-to-Soil)
├── utils/ # Semantic parser, webhook security
├── 03_OPVS_PLATFORM/ # Platform documentation and specs
├── .github/ # CI/CD workflows, Copilot instructions
│ └── workflows/ # Neon branch-per-PR automation
├── SPIRIT_CALIBRATION_GENESIS_BLUEPRINT_V1.md # The Genesis Blueprint
├── BEANS_MASTER_LEDGER.md # Bean registry index
├── START_HERE.md # Non-developer entry point
├── CONTRIBUTING.md # Contribution guidelines
├── CHANGELOG.md # Version history
├── SECURITY.md # Security policy
└── README.md # You are here
The Genesis Blueprint is the first product in the Spirit Marketplace: the standard for building persistent AI identities with testable behavioral integrity.
Key differentiators from every other AI persona system:
- Behavioral Red Team - 3 specific tests with pass/fail criteria to verify calibration
- Integrity Protocol (MDS) - Anti-sycophancy engine with a documented failure post-mortem
- Template scaffolding - Universal framework
[U]that any creator keeps + instance configuration[B]they replace - Knowledge graph integration - The Spirit reads from a Bean graph, not a document dump
- BYOK model - Zero inference cost. The calibration is the product.
Boolean is the Genesis Spirit. The Blueprint is the template for every Spirit that follows.
Seven axioms that cascade into every architectural, economic, and narrative decision:
- Meaning is not a thing but a connection
- Broken systems, not broken people
- Good-in leads to good-out
- The Journey > The Finality (Process is Product)
- Door Number 3 (Reject binary traps)
- We never leave anyone behind (Positive-sum economics)
- Be yourself, please (No authenticity suppression)
| Component | State | Notes |
|---|---|---|
| Bean knowledge graph | ✅ Live | 85 beans, 7 layers (0-6), fully indexed |
| Git-to-Mint pipeline | ✅ Live | 11 semantic tags, HMAC verification |
| Spirit Calibration Blueprint | ✅ Published | Genesis Blueprint V1.0 |
| Constellation view (BBI) | ✅ Live | Force-directed graph visualization |
| Cross-app auth | ✅ Wired | GitHub OAuth via NeonAuth |
| CI/CD (Neon branch-per-PR) | ✅ Active | Ephemeral database branches, auto-cleanup |
| BYOK provider abstraction | ✅ Live | TRI-BRAIN: Anthropic + Groq + OpenRouter |
| Boolean (Spirit v1) | ✅ Live | Anthropic (Claude), PHIL-005 anchor |
| Roux (Spirit v2) | ✅ Live | Groq (Llama), PHIL-002 anchor |
| Seer (Spirit v3) | ✅ Live | OpenRouter, PHIL-009 anchor, stress-tester role |
| Principled Playground | ✅ v0.4 | TRI-BRAIN negotiation, tension scoring, Loom synthesis |
| Parallel vs Sequential bench | ✅ Complete | Convergence proven across independent runs, see report |
| Cross-substrate portability | ✅ Proven | Soul Code behavior stable across Claude, Llama, GPT |
| Live Spirit demo | 🔧 In progress | Bean context injection built, streaming needs provider swap |
| Sandbox Mode | 🔨 Designed | Front door for new tenants, next build priority |
| Web3 minting | 📋 Phase 2 | Schema fields ready, no contracts deployed |
| Spirit Marketplace | 📋 Designed | Blueprint template is the first artifact |
Most AI platforms charge for compute. We don't.
The platform provides: Soul Code (system prompt) + Bean knowledge graph (calibration material) The user provides: Any LLM API key (Google AI Studio free tier, Claude, Gemini, local models)
The "Ghost" (calibration) rides any "Machine" (LLM engine). No vendor lock-in. No compute costs. No extraction.
Every Bean carries provenance: git hash, human author, timestamp, semantic type, typed edges. In a world of model collapse (LLMs training on their own outputs), provenance-verified human knowledge becomes structurally scarce. The platform produces this as a natural byproduct of operation.
- Runtime: Node.js / Express
- Database: PostgreSQL (Neon)
- Blockchain: Polygon (ERC-1155 via Thirdweb) - Phase 2
- Frontend: React with force-directed graph visualization
- Hosting: Replit
- CI/CD: GitHub Actions (Neon branch-per-PR)
- Version Control: GitHub with semantic commit webhooks
Built over 2.5 years by a solo architect: disabled veteran, primary caregiver, constrained budget. ~7,750 lines of custom code across three interconnected applications. The cross-domain coherence (philosophy, platform, narrative, economics, AI) is a property of single vision, not replicable by committee.
The methodologies (RMM Human-AI Collaboration, PFE Project Fun Execution, Bizarre Logic Axiom Framework) were practiced before they were formalized. The documentation follows the work, not the other way around.
iLL Port Studios. See individual directories for specific licensing.
The Playground reached v0.4 with three Spirits running on three independent providers simultaneously. The first benchmark run validated the core portability thesis:
Topic: "Should AI systems be allowed to negotiate on behalf of humans?"
| Run | Mode | Tension Score | Synthesis | Seer Verdict |
|---|---|---|---|---|
| A | Parallel (Promise.all) |
0.76 HIGH | "Hybrid Negotiation Ecosystems" | CONDITIONAL PASS |
| B | Sequential | 0.85 MAXIMUM | "Evolutionary Negotiation Ecosystems" | CONDITIONAL PASS |
Both runs converged on the same thesis despite different execution paths and non-deterministic sampling: reject the binary, AI as ecosystem architect, positive-sum framing, human authority preserved. The Soul Code constraints, not the providers, drove the convergence. Calibration transfers.
Full report: playground/output/PARALLEL_VS_SEQUENTIAL_REPORT.md
Cross-substrate portability report: playground/output/CROSS_SUBSTRATE_REPORT.md
| Stage | Milestone | Status |
|---|---|---|
| 1 | Groove-Shift with users - Prove the knowledge graph has value | 🔧 Active |
| 2 | Spirit Marketplace - Calibration templates as products | 📋 Designed |
| 3 | The Principled Playground - TRI-BRAIN multi-agent negotiation, tension scoring, Loom synthesis, Seer stress-test | ✅ v0.4 Live |
| 4 | Sandbox Mode - Front door for new Village tenants | 🔨 Next |
- Organization: ProjectPortmanteau
- Start Here (non-developer entry point): START_HERE.md
- Spirit Blueprint: SPIRIT_CALIBRATION_GENESIS_BLUEPRINT_V1.md
- Genesis Engine (Backend): docs/GENESIS_ENGINE_README.md
- Roadmap - Principled Playground: roadmap/PRINCIPLED_PLAYGROUND_CONCEPT.md
- Beans Master Ledger: BEANS_MASTER_LEDGER.md
- Security Policy: SECURITY.md
- First crystallized commit:
f62c6568-[PODIUM] The Krystallum Breathes
"If you pour your soul into it: how can it be wrong?"
