SPAR

SPAR-Kit

Structured Persona-Argumentation for Reasoning
A methodology for stress-testing decisions through structured disagreement.

npm MIT License Try Now Tests

நாலு பேரு, நாலு திசை, ஒரு முடிவு!
Four Perspectives, Four Dimensions, One Synthesis

--- ## What is SPAR? **SPAR** is a decision-making methodology that uses structured disagreement between distinct personas to surface insights that consensus-seeking processes systematically miss. Whether those personas are embodied by humans, simulated by AI, or both, the methodology remains the same. | Letter | Meaning | Definition | |--------|---------|------------| | **S** | Structured | Not ad-hoc; a designed process with protocols and phases | | **P** | Persona | Distinct voices with coherent, conflicting worldviews | | **A** | Argumentation | Personas debate and defend positions, not just "provide perspectives" | | **R** | Reasoning | Systematic logic that exposes blind spots and triggers System 2 thinking | --- ## The Problem **Don't Deliberate Alone.** The problem isn't AI. The problem is **isolated reasoning** — and it fails in every context: | Context | The Problem | |---------|-------------| | **Leader alone** | Personal blind spots, confirmation bias, echo chamber of self | | **Team together** | Groupthink, shared assumptions, hierarchy silencing dissent | | **Person + AI** | Single reasoning context producing false "balance" | | **AI alone** | Sophisticated coherence without genuine tension | When you deliberate alone — whether that's a lone executive, a homogeneous team, or a single AI context — you get *variations on a theme*, not genuine disagreement. **This is not debate. This is sophisticated confirmation bias.** ## The Solution SPAR creates **genuine tension** by instantiating personas with coherent, conflicting worldviews that *actually disagree*. The result is not a balanced trade-off list. It's a dialectic that surfaces what you'd never see alone. > *"One methodology. Five modes. Every decision."* > *"Don't deliberate alone. SPAR."* --- ## Installation ### One-Line Install (Recommended) ```bash curl -fsSL https://raw.githubusercontent.com/synthanai/spar-kit/main/install.sh | sh ``` This installs the CLI **and** auto-configures integrations for: - 🤖 **Claude Code** — Adds skill to `~/.claude/skills/` - ✨ **Gemini** — Adds skill + workflow to `.agent/` - 🎯 **Cursor** — Adds rules to `~/.cursor/` ### Web (No Install) **[▶️ synthanai.github.io/spar-kit](https://synthanai.github.io/spar-kit)** ### ChatGPT GPT (No Install) **[🤖 SPAR Decision Partner](https://chatgpt.com/g/g-69678df7ed248191be8fd211e11d4797)** — Use SPAR directly in ChatGPT ### npm Only ```bash npm install -g sparkit ``` --- ## Usage ```bash # 🎯 Launch TUI Mission Control (recommended) sparkit tui # 🧙 Launch SPAR Builder wizard sparkit builder # ⚡ Quick debate sparkit "Should we expand to Singapore?" # 📝 Full command sparkit debate start "Your decision here" # 📦 Use a template sparkit template use startup-pivot ``` ### TUI Features (v4.0.0) - **Dashboard**: View all sessions with status indicators - **Session Management**: Pause, resume, cancel, clone debates - **Live Monitoring**: Watch debates unfold in real-time - **Persona Browser**: Browse 109 personas across 7 archetypes - **3-Round Rumble**: Structured Opening → Clash → Final protocol - **Keyboard Navigation**: Full keyboard control (see `docs/KEYBOARD_SHORTCUTS.md`) --- ## 📚 Core Concepts ### The Four Directions (N-E-W-S) SPAR's foundational model uses a compass metaphor. Four personas, four orientations, four questions that together leave no blind spot unchallenged: ``` 🔵 NORTH The Visionary "Where are we going?" │ 🔴 WEST ─────────────┼───────────── 🟢 EAST The Sage │ The Challenger "What's proven?" │ "What's emerging?" │ 🟡 SOUTH The Pragmatist "What's grounded?" ``` | Direction | Persona | Core Question | Orientation | |-----------|---------|---------------|-------------| | 🔵 **North** | The Visionary | *Where are we going?* | Future, ideals, possibility | | 🟢 **East** | The Challenger | *What's emerging?* | Innovation, disruption, new dawn | | 🟡 **South** | The Pragmatist | *What's grounded?* | Reality, execution, stability | | 🔴 **West** | The Sage | *What's proven?* | Experience, wisdom, lessons learned | **Natural Tensions:** - **North ↔ South** — Vision vs. Reality - **East ↔ West** — Innovation vs. Tradition 📖 **[Full Four Directions →](methodology/FOUR_DIRECTIONS.md)** | 📖 **[Centre: Behavioral Analyst →](methodology/CENTRE.md)** --- ### The SPARKIT Protocol (7 Steps, v8.0) ``` 0 — NOOL State deliberation intent (WHY, WHAT TYPE, HOW) S — SCOPE Define the question precisely P — POPULATE Instantiate clashing personas A — ABSTRACT Construct shared cognitive map (AoT) ★ NEW R — RUMBLE Structured dialectic (rounds by depth mode) K — KNIT Moderator synthesizes tensions I — INTERROGATE Stress-test the synthesis T — TRANSMIT Extract actionable recommendations ``` > **v8.0**: Step 3 ABSTRACT was elevated from a POPULATE sub-activity to a dedicated step after validated experiments showed +40% synthesis quality improvement. 📖 **[Full SPARKIT Protocol →](methodology/PROTOCOL.md)** --- ### TESSERACT Configuration (378 Combos) Tune any debate across four independent axes: | Axis | Options | Default | |------|---------|--------| | **Depth** | `clash` (4+PROBE, minimum) · `rumble` (8+) · `domain` (4+N) | clash | | **Pattern** | sequential · freeflow · binary · dialectic · inquisition · emergence | dialectic | | **Style** | balanced · adversarial · steelman · consensus · premortem · escalation · inversion | balanced | | **Horizon** | now · cycle · legacy · shifted | cycle | Or use a preset: ⚡ `quick` · 🔮 `deep` · 🚨 `crisis` · 🤝 `align` · 🎯 `challenge` 📖 **[Full TESSERACT Configuration →](methodology/TESSERACT.md)** --- ### Reasoning Depth Modes (6 Levels) Match debate intensity to decision stakes: | Mode | Rounds | Personas | Best For | |------|--------|----------|----------| | 👀 Show Me | 1 | 4 | Demo, onboarding | | ⚡ Quick | 3 | 4 | Rapid sanity check | | ⚖️ Basic | 3 | 4 + PROBE | Standard decisions | | 💣 Stress Test | 3 | 4 + PROBE | Significant investment | | 🧠 Deep | 5 | 8 + PROBE | Strategic pivots | | 🤯 Ultra | 7+ | 8+ | Existential decisions | 📖 **[Full Depth Modes →](methodology/DEPTH_MODES.md)** --- ### The GRACE Principles SPAR is grounded in five epistemological stances. Be GRACEful: | Principle | Meaning | |-----------|---------| | **G**round | Still yourself before synthesis — insights arrive in silence | | **R**eveal | Surface biases and anxieties before delegating to process | | **A**pproach | Seek the adjacent possible, not defended positions | | **C**ourt | Welcome radical dissent — invite it, make space for it | | **E**xpose | Acknowledge what you don't know, name missing perspectives | 📖 **[Full GRACE Principles →](methodology/GRACE_PRINCIPLES.md)** --- ### The STASH Framework (5 Modes) **STASH** defines the five ways to run SPAR. Choose your mode based on context: ``` ┌───────────────────────────────────────────────────────────────────────┐ │ S T A S H │ ├───────────────────────────────────────────────────────────────────────┤ │ 👤 S — SOLO HUMAN One person, multiple mental frames │ │ 👥 T — TEAM HUMAN Team members embody assigned personas │ │ 🤖 A — AI-PERSONA Multiple LLMs argue with each other │ │ 🧠 S — SOLO AI One LLM role-plays all perspectives │ │ 🔀 H — HYBRID AI + Human combined │ └───────────────────────────────────────────────────────────────────────┘ ``` | Mode | Best For | Speed | Skills Built | |------|----------|-------|-------------| | 👤 **Solo Human** | Personal growth, deep reflection | Slow | Very High | | 👥 **Team Human** | Building alignment, psychological safety | Medium | High | | 🤖 **AI-Persona** | Rapid iteration, 24/7 availability | Fast | Low | | 🧠 **Solo AI** | Quick exploration, low-stakes decisions | Fastest | Low | | 🔀 **Hybrid** | High-stakes decisions, maximum insight | Medium | Medium | 📖 **[Full STASH Framework with Prompt Templates →](methodology/STASH_MODES.md)** --- ### The ASPIRES Framework (7 Advanced Patterns) For experienced practitioners, seven advanced patterns to handle edge cases: | Pattern | When to Use | |---------|-------------| | **A**bsent Voice | When a critical stakeholder is missing | | **S**teelman | When a position is being dismissed too easily | | **P**re-Mortem | When you're too confident in a decision | | **I**nversion | When stuck on how to succeed | | **R**eflective Meta-Review | When the debate itself seems dysfunctional | | **E**scalation Trigger | When emergent concerns warrant strategic pause | | **S**hifted Horizon | When the timeframe is constraining options | 📖 **[Full ASPIRES Framework →](methodology/ADVANCED_PATTERNS.md)** --- ### The Framework Critic (PROBE) SPAR includes a unique meta-persona that **attacks the methodology itself from within**: ``` PROBE: THE FRAMEWORK CRITIC "What is PERSONA itself preventing us from seeing?" Recursive skepticism — using structured disagreement to question structured disagreement. The methodology's own conscience. The framework's doubt. ``` --- ### Minimum Viable Debate (MVD) For time-constrained decisions, a **30-minute, 4-persona protocol**: | Persona | Question | |---------|----------| | **Protector** | "What could go wrong?" | | **Explorer** | "What's possible?" | | **Realist** | "What's feasible?" | | **Navigator** | "Where should we head?" | *80% of decision value in 20% of time.* 📖 **[Full MVD Protocol →](methodology/PRACTICE_GUIDELINES.md#part-2-minimum-viable-debate-mvd)** --- ## Commands ### Core Commands | Command | Description | |---------|-------------| | `spar` | Start interactive debate | | `spar tui` | **NEW!** Launch Mission Control TUI | | `spar builder` | **NEW!** Launch SPAR Builder wizard | | `spar debate start [topic]` | Start debate session | | `spar debate history` | View past sessions | ### Template Commands | Command | Description | |---------|-------------| | `spar template list` | Show all templates | | `spar template show ` | View template details | | `spar template use ` | Use template to start debate | | `spar template create` | Create custom template | | `spar template delete ` | Delete a template | ### Built-in Templates | Template | Preset | Use Case | |----------|--------|----------| | `startup-pivot` | 🚀 Startup | Evaluate major pivots | | `hiring-decision` | 💼 Corporate | Should you hire this person? | | `investment-eval` | 💼 Corporate | Evaluate investments | | `product-launch` | 🎨 Innovation | Launch readiness | | `crisis-response` | ⚠️ Crisis | Emergency decisions | | `ethics-dilemma` | ⚖️ Ethics | Navigate ethical situations | ### Other Commands | Command | Description | |---------|-------------| | `spar persona list` | Show all 108 personas | | `spar persona create` | Create custom persona | | `spar config setup` | Configure provider & API key | | `spar config show` | View configuration | | `spar status` | Show version & stats | | `spar compass` | Display N-E-W-S framework | --- ## Configuration **Global config:** `~/.spar/` ``` ~/.spar/ ├── config.json # Provider + API key ├── personas/ # Custom personas └── sessions/ # Auto-saved debates ``` --- ## Providers ### OpenRouter (Recommended) 🆕 **FREE-Optimal** configuration validated at 100% success, $0.00/run: | Role | Model | Use | |------|-------|-----| | Debaters | `google/gemini-2.0-flash-exp:free` | Fast, 1M context | | Arbiter | `deepseek/deepseek-r1-0528:free` | True reasoning | | Critic | `meta-llama/llama-3.3-70b-instruct:free` | Balanced rigor | ### Empirical Validation (January 2026) | Finding | Evidence | Confidence | |---------|----------|------------| | **Arbiter quality is the critical determinant** | Inverse config failed 40% of synthesis tasks | 95% | | **FREE-Optimal is production-viable** | 100% success rate @ $0.00/run | 90% | | **SPAR protocol adds +53% quality** | Even commodity multi-model beats premium single | 88% | 📖 **[OpenRouter Integration Guide →](docs/OPENROUTER_INTEGRATION.md)** ### Other Providers | Provider | Model | |----------|-------| | OpenAI | GPT-4 Turbo | | Anthropic | Claude 3.5 Sonnet | | Google | Gemini 1.5 Flash | --- ## 📖 Methodology Documentation This toolkit includes the complete **SPAR methodology** (v8.0), everything you need to understand, implement, and master structured persona-argumentation: ### Philosophy & Foundations | Document | Description | |----------|-------------| | [**MANIFESTO.md**](methodology/MANIFESTO.md) | The complete philosophy of SPAR | | [**GRACE_PRINCIPLES.md**](methodology/GRACE_PRINCIPLES.md) | The GRACE principles: 5 epistemological stances | | [**FOUR_DIRECTIONS.md**](methodology/FOUR_DIRECTIONS.md) | Deep dive into the N-E-W-S compass model | ### Protocol & Configuration (v8.0) | Document | Description | |----------|-------------| | [**PROTOCOL.md**](methodology/PROTOCOL.md) | The 7-step SPARKIT protocol with prompts | | [**ABSTRACT.md**](methodology/ABSTRACT.md) | **NEW** Abstraction of Thought (AoT) specification | | [**TESSERACT.md**](methodology/TESSERACT.md) | **NEW** 504-configuration system (4 axes) | | [**DEPTH_MODES.md**](methodology/DEPTH_MODES.md) | **NEW** 6 reasoning depth modes | | [**CENTRE.md**](methodology/CENTRE.md) | **NEW** Centre: Behavioral Analyst (always-on, Human 5 Whats) | ### Process & Practice | Document | Description | |----------|-------------| | [**STASH_MODES.md**](methodology/STASH_MODES.md) | The 5 modes: Solo Human, Team Human, AI-Persona, Solo AI, Hybrid | | [**ADVANCED_PATTERNS.md**](methodology/ADVANCED_PATTERNS.md) | The ASPIRES framework: 7 advanced patterns | | [**PRACTICE_GUIDELINES.md**](methodology/PRACTICE_GUIDELINES.md) | When to use, when NOT to use, and how to do it right | | [**WHEN_NOT_TO_USE_SPAR.md**](methodology/WHEN_NOT_TO_USE_SPAR.md) | Explicit contraindications and boundary conditions | | [**FAQ.md**](methodology/FAQ.md) | Frequently asked questions (18 common questions answered) | ### Governance & Advanced | Document | Description | |----------|-------------| | [**6R_ENGINE.md**](methodology/6R_ENGINE.md) | Backstage cognition loop | | [**SPARED.md**](methodology/SPARED.md) | Coaching cycle (5 phases: Ground, Flow, Burn, Expand, Return) | | [**RAMP_LEVELS.md**](methodology/RAMP_LEVELS.md) | Decision gravity assessment (5 levels) | | [**COMMIT_DEPTH.md**](methodology/COMMIT_DEPTH.md) | Graduated autonomy (Shallow to Full) | | [**ARENA_MITOSIS.md**](methodology/ARENA_MITOSIS.md) | Recursive arena spawning | | [**LOOP_GOVERNANCE.md**](methodology/LOOP_GOVERNANCE.md) | Hard limits, circuit breakers, checkpoint/resume | | [**FEEDBACK_RESOLUTION.md**](methodology/FEEDBACK_RESOLUTION.md) | Disagreement disposition protocol | | [**DOORS.md**](methodology/DOORS.md) | Decision exit framework | | [**DEBATE_LINEAGE.md**](methodology/DEBATE_LINEAGE.md) | Session relationship tracking | ### Quick Start | Document | Description | |----------|-------------| | [**SPAR_ONE_PAGE_CHEATSHEET.md**](methodology/SPAR_ONE_PAGE_CHEATSHEET.md) | **Start here**: Learn SPAR in 60 seconds | | [**SPAR_IN_5_MINUTES.md**](methodology/SPAR_IN_5_MINUTES.md) | Learn to run SPAR manually in 5 minutes | ### Resources & Templates | Document | Description | |----------|-------------| | [**PERSONA_LIBRARY.md**](methodology/PERSONA_LIBRARY.md) | 20+ ready-to-use personas for different contexts | | [**PERSONA_LIBRARY_108.md**](methodology/PERSONA_LIBRARY_108.md) | Complete 109-persona library (108 + Framework Critic) | | [**PERSONA_DOMAIN_PACKS.md**](methodology/PERSONA_DOMAIN_PACKS.md) | 10 domain-specific preset packs (Healthcare, Startup, etc.) | ### Research & Foundations | Document | Description | |----------|-------------| | [**RESEARCH_FOUNDATIONS.md**](methodology/RESEARCH_FOUNDATIONS.md) | Academic grounding (Kahneman, Tetlock, Mercier & Sperber) | --- ## Security - API keys stored locally in `~/.spar/config.json` - No server — direct API calls to your provider - Web version uses browser localStorage --- ## Testing ```bash # Run all tests (140+ tests) npm test # Watch mode npm run test:watch # Coverage report npm run test:coverage ``` ### Test Suite | Suite | Tests | Coverage | |-------|-------|----------| | `cli.test.js` | CLI configuration, personas, providers, validation | ✅ | | `spar.test.js` | HTML structure, accessibility, JavaScript engine | ✅ | | `style.test.js` | CSS variables, layout, components, syntax | ✅ | | `integration.test.js` | Full debate flow, cross-component consistency | ✅ | --- ## Example ```bash $ spar "Should we pivot from B2B to B2C?" ╔═══════════════════════════════════════════════════════════════╗ ║ 🥊 S P A R v4.0.0 ║ ╚═══════════════════════════════════════════════════════════════╝ ═══ RUMBLE — Round 1/3: Opening ═══ Token budget: 1500 tokens 🔵 North — "B2C is where the scale is..." 🟢 East — "The market is shifting..." 🟡 South — "Do you have the capital?" 🔴 West — "Every B2C company started with B2B..." ═══ RUMBLE — Round 2/3: The Clash ═══ ═══ RUMBLE — Round 3/3: Final Positions ═══ ✓ 3-Round Rumble Complete 📊 SYNTHESIS: Key Tensions, Evolution, Insights ✓ Session saved to ~/.spar/sessions/... ``` --- ## Development ```bash git clone https://github.com/synthanai/spar-kit.git cd spar-kit npm install npm test # Run 140+ tests node cli/index.js ``` --- ## 🤝 Contributing SPAR is a living methodology that grows through use. Contributions welcome: - **🎭 New Personas** — Submit personas for uncovered decision types - **📝 Case Studies** — Document SPARs you've run - **📖 Methodology Refinements** — Propose improvements to the core concepts - **🌍 Translations** — Help SPAR speak new languages See **[CONTRIBUTING.md](CONTRIBUTING.md)** for guidelines. --- ## 🧭 The Deeper Pattern SPAR isn't just a technique. It's a stance toward knowledge: > *I am not the only valid perspective.* > > *My confidence, however justified, has blind spots.* > > *The position that feels wrong to me might be seeing something I'm missing.* > > *The synthesis that emerges from genuine tension is more robust than any position I could hold alone.* --- ## Links - **[ChatGPT GPT](https://chatgpt.com/g/g-69678df7ed248191be8fd211e11d4797)**: Use SPAR in ChatGPT - **[Web Playground](https://synthanai.github.io/spar-kit)**: Try in browser - **[npm Package](https://www.npmjs.com/package/sparkit)**: Install via npm - **[SPARKIT Protocol](methodology/PROTOCOL.md)**: 7-step methodology (v8.0) - **[TESSERACT Config](methodology/TESSERACT.md)**: 504 debate configurations - **[Depth Modes](methodology/DEPTH_MODES.md)**: 6 reasoning intensity levels - **[ASPIRES Framework](methodology/ADVANCED_PATTERNS.md)**: Advanced patterns - **[Full Manifesto](methodology/MANIFESTO.md)**: Complete philosophy - **[ROADMAP](ROADMAP.md)**: What's coming next --- ## License MIT © [Naveen Riaz Mohamed Kani](https://github.com/synthanai) ---

நாலு பேரு, நாலு திசை, ஒரு முடிவு
Don't deliberate alone. SPAR. 🥊