219 ready-to-use modules, artifacts & runnable tools — LLM routing, RAG, agent loops & memory, MCP servers, prompt engineering, fintech ledgers & minting, self-hosted model training, Next.js SaaS, Three.js — each distilled from a real, running production system. Copy a piece, or run a whole mini-app.
Open source — use it wisely.
Tools · Modules · Artifacts · Recipes · index.json · llms.txt · ▶ Playground · Contribute
Built to be queried by agents, not just read:
llms.txt— every atom as a flat, LLM-ingestible list. Drop it into any RAG pipeline.index.json— the machine-readable map (id, lane, category, path).tools/tyx-lab-mcp— an MCP server your agent registers tosearch_atoms()/get_atom()on demand. The whole library becomes system memory. Seedocs/KNOWLEDGE_PATH.md.
Every atom was pulled out of a real system and stripped to its smallest reusable unit — one job each, so you can lift it into a different project. Four ways to use it:
tools/— runnable mini-tools & mini-apps: single-file web apps, CLIs, MCP servers, scripts.modules/— atomic code you import: one job, one file.artifacts/— atomic non-code: prompts, templates, formulas, methods, playbooks, frameworks, checklists, findings.recipes/— how to stitch several atoms into a working use case.
Anyone building AI agents, agentic AI, LLM apps, RAG pipelines, MCP servers, AI tools, self-hosted models, SaaS, or software who would rather grab a battle-tested piece than reinvent it.
23tools ·117modules ·79artifacts ·7recipes. Machine-readable:index.json·llms.txt.
- ai-content-pipeline
- brand-system-builder
- fintech-wallet-and-minting
- rag-knowledge-base
- self-hosted-agent-stack
- ship-an-mcp-server
- tracked-multitenant-saas
AI, Agents & LLM (5)
- Agentic Framework —
script— Lightweight framework for building autonomous agents with tool use, planning, and memory. Supports multiple LLM backends, extensible tool... - AI Prompt Lab —
web-app— Craft, test, and refine AI prompts. Write positive/negative prompts, choose models and styles, generate images. Save and manage prompts. - Background Remover —
web-app— Remove image backgrounds with AI-powered processing. Upload photos and download PNG with transparent background. - TTS Voice Module —
script— Multi-engine Text-to-Speech module with voice cloning support. Supports OpenAI TTS, Coqui XTTS-v2 (self-hosted), and Web Speech API. Reac... - tyx-lab MCP Server (library brain) —
mcp-server— Search and fetch tyx-lab atoms on demand over MCP — turn the whole 188-atom library into a live, queryable knowledge base for your agent.
Infrastructure & Backend (4)
- CLI Tool Template —
cli— Minimal command-line interface template with argument parsing, config storage, and example commands. Adapt for any domain. - MCP Server Skeleton —
mcp-server— Minimal stdio-transport MCP server template with tool registration & example tool. Use as foundation for adapting any service/API to MCP. - Platform Connector Adapter —
script— Unified API connector layer for social/commerce platforms. Adapters for messaging APIs, analytics feeds, e-commerce integrations. Configu... - Visual Flow Builder —
web-app— Browser-based visual flow builder for composing multi-step workflows. Drag-and-drop steps, configure parameters, save as JSON, and execut...
Business, SaaS & Fintech (3)
- Content Calendar —
web-app— Editorial calendar for content planning and scheduling. Month/week/day views, drag-drop scheduling, status tracking, team collaboration. - Credit & Usage Dashboard —
web-app— Real-time credit balance and usage tracking dashboard. Polls backend for credit status, displays 30-day trends with mock data fallback, c... - Embeddable Balance Widget SDK —
tool— JavaScript/React SDK for embedding real-time wallet balance display in partner web apps. Reads from public REST API, handles currency for...
Ops & Methodology (3)
- Log Viewer Dashboard —
web-app— Real-time reactive log viewer with polling & Socket.io support. Displays structured logs with live refresh, filtering by agent/service, a... - Upload Monitor Dashboard —
web-app— Real-time file upload monitoring dashboard. Tracks upload progress, file sizes, error rates, queue status with live updates via Socket.io. - World Config Editor —
web-app— Monaco-based JSON editor for world.json configuration files. Direct load/save from gateway backend with syntax highlighting and validation.
Video, Media & Publishing (3)
- Image to Video —
web-app— Animate static images into videos with AI. Add motion prompts, choose from cinematic/anime/3D styles, set duration and FPS, download MP4. - Motion Director MCP — Video Pipeline Tool —
mcp-server— MCP server that exposes the Motion Director AI-video pipeline (idea → Visual DNA → storyboard → keyframes → i2v → voiceover →... - Video Editor —
web-app— Edit videos with trim, text overlay, and audio controls. Multi-track timeline with visual editing. Import, edit, preview, and export videos.
3D, Graphics & Games (2)
- Canvas Editor (Fabric.js) —
web-app— 2D canvas editor powered by Fabric.js v6. Full-featured drawing, object manipulation, export/import with REST API backend (Express). - Image to 3D —
web-app— Convert 2D images to interactive 3D models. Upload photos with optional prompts, choose output format (GLB/OBJ/FBX), and preview in real-...
Games (1)
- NPC Generator —
web-app— Generate game characters with customizable stats, rarity levels, and roles. Browse roster, export character data, integrate with game wor...
Design (2)
- Map Generator —
web-app— Generate procedural maps from descriptions. Choose size/style (cyberpunk/nature/scifi/fantasy), preview with tile-based rendering, export... - Photo Editor —
web-app— Edit photos with 12+ filter presets, brightness/contrast/saturation controls, and sticker overlays.
AI, Agents & LLM (25)
- Agent Cascade Router — LLM-based Tool Selection & Executor Dispatch — 2-tier cascade architecture: controller LLM selects tool + executor model, then passes work to specialized executor. Per-agent persona + ...
- Agent Memory RAG — Obsidian-style Markdown + Wikilinks — Per-agent memory system with markdown notes, [[wikilink]] retrieval, and knowledge lineage tracking. Keyword-based retrieval with title b...
- Agent Memory: 3-Layer Progressive Retrieval — Memory architecture for agents: index layer (FTS5), timeline layer (recent events), full context layer. Reduces token use 10x via progres...
- Agent Persona Registry — Unified schema for registering NPC identities, personalities, skills, and role mappings in multi-agent systems.
- AI Asset Quality Gate — Automated quality checks for AI-generated content (title/keyword/disclosure validation + scoring)
- AI Metadata Stock Generator — Generate title, keywords, description, and AI disclosure for stock marketplace assets
- AI Provider Router with Fallback — Route LLM calls across multiple OpenAI-compatible endpoints with automatic fallback chaining.
- Behavior Tree & FSM State Model — Declarative schema for agent behavior: finite state machine (FSM) with states, transitions, conditions, and actions. Serializable and com...
- Brand Schema Validator — Zod-based schema for 5-step brand brief intake: positioning, competitive analysis, tone, archetype, cultural context. Validates form inpu...
- CoT Extractor & Quality Scorer — Extract chain-of-thought from LLM. Parse & blocks. Score reasoning quality (weak/adequate/strong). Detect epistemik ...
- Dialogue Tree: Branching Narrative Model — Declarative dialogue tree for NPC conversations: branches, conditions, player choices, dialogue outcomes. Serializable for save/load and ...
- Epistemology Framework (Islamic) — Islamic epistemology engine. Sanad chains, yaqin levels (ilm/ain/haqq), mutawatir/ahad tiers, maqashid al-shariah alignment, 4 cognitive ...
- Jungian Archetype Selector — Canonical 12 Jungian brand archetypes mapped to quadrants with voice markers, core desires, and brand examples — for strategic position...
- LLM Provider Registry Switchable — Declarative multi-provider LLM registry with fallback support (OpenRouter, HF, OpenAI, Anthropic, Gemini, custom)
- Multi-Provider AI Router — Fallback AI content provider with priority: Groq (free) → Gemini (free) → Ollama (local) → Anthropic. Handles timeouts, fallback, a...
- Multi-Provider AI Router — Unified LLM client with auto-fallback: Gemini→Cerebras→OpenAI→Anthropic→Pollinations. Prioritizes BYOK.
- ReAct Reasoning Loop — Reason + Act loop for agentic systems. Thought -> Action -> Observation cycling with max steps & anti-loop safeguards.
- Sanad Verifier (Multi-Source) — Cross-verify factual claims across web + corpus + LLM-prior. Reject hallucinations for fact-checkable, brand-specific, current-event clai...
- Self-Refining Agent Loop (Tafakkur Pattern) — Dual-agent criticism loop: innovator generates → critic evaluates across multiple modes → refine or accept. Prevents single-pass medi...
- Server-Sent Event Chat Stream Parser — Parse SSE frames from LLM/chat streaming endpoints. Tolerant of multiple JSON payload formats (token/delta/content/text variants), plain ...
- Skill Registry Contract Pattern — Contract-enforced skill registry with boot-time validation: description and parameters mandatory, tools self-documenting.
- Telegram Command Router & Inline Keyboard Builder — Route Telegram bot commands to handlers and build inline keyboard UI with callback actions.
- Tile Grid Coordinator — Pure grid coordinate conversion utility for top-down and isometric tile maps (RPG/Gather-style). Square tiles, screen offset support, Man...
- Tools Catalogue Schema — Unified registry schema for media providers, LLM vendors, and callable tools with free/paid status, fallback chains, and compatibility.
- TTS Provider Wrapper — Abstraction layer for TTS services (OpenAI, Google Cloud, Azure) with caching, rate limiting, and streaming output. Normalizes provider A...
Infrastructure & Backend (41)
- Action Guarded Executor — Wraps any callable with guard layers: health check → circadian check → rate-limit → think delay → execute. Returns structured Act...
- Adaptive Rate Limiter with Warmup — Per-entity rate limiter with age-based warmup curve (0.08→1.0 over 21 days), jittered intervals, daily caps, and persistent state.
- AES-256-GCM Secret Encrypt/Decrypt — Encrypt/decrypt sensitive data (API keys, tokens, secrets) with AES-256-GCM. Output: base64(iv || authTag || ciphertext). Key derived fro...
- Atomic JSON File Store — Concurrency-safe file-based JSON storage with atomic write (temp+rename), read, and update.
- Audio Chunk Streamer — HTTP streaming for large audio files with chunked transfer encoding. Supports range requests, pause/resume, and adaptive bitrate.
- Auth Session Token Manager — In-memory session + token auth with bcrypt hashing and Express middleware
- B2B API-Key Gateway Pattern — Multi-tier API key management for platform partners. Supports key rotation, rate limiting per API key, transaction signing with shared se...
- Browser Automation Engine — Anti-detection browser launcher (patchright > playwright-extra > playwright). Supports Browserbase/Steel cloud.
- BullMQ Job Enqueuer — Wrapper for BullMQ queue with exponential backoff, retry policy, and auto-cleanup. Standardizes job enqueueing for e-commerce workflows (...
- Dual-Auth Internal Key Pattern — Server-to-server + user authentication hybrid: (1) Client sends JWT (user auth), (2) Server adds x-internal-key header (server auth). Pre...
- Embedding Model Loader (3-Way) — Lazy-load embedding models: BGE-M3 (default, multilingual), Mamba2 (best quality), MiniLM (CPU fallback). Graceful disable if dependencie...
- Environment Config Schema — Pydantic-based environment variable configuration with type coercion, validation, and defaults. Single source of truth for app settings a...
- Exchange Rate Oracle Updater — Trusted price feed mechanism that updates denomination-to-fiat conversion rates via authenticated oracle transactions. Supports historica...
- Exponential Backoff Queue — BullMQ queue configuration with exponential backoff retry strategy. Prevents thundering herd, includes auto-cleanup, and supports preset ...
- Flow DAG Executor — Executes declarative node DAG from JSON: trigger, action, foreach, filter, spinner, delay. Variables via {{path.binding}}. Logs all node ...
- Human-Like Timing Utilities — Sleep/delay helpers: typing speed, page reads, request gaps. For realistic bot behavior & rate-limit avoidance.
- JSON File Store — Atomic — Atomic ACID-like JSON file-based storage with read/write/update and recovery
- JWT Auth Cookies — JWT-based authentication with secure httpOnly cookies, refresh token rotation, and access token expiry. Handles auth state across SPA + API.
- JWT Token Helper — Create and validate JWT access tokens with configurable TTL. Handles token generation, validation, claim extraction, and error cases for ...
- Media Provider Fallback Orchestrator — Orchestrate calls to media providers (image, video, audio, design) with automatic fallback on failure and cost/quota tracking.
- Meta Webhook Verifier — Verify Meta Cloud API webhook challenge-response and HMAC-SHA256 signatures. Validates webhook authenticity and prevents spoofing for Wha...
- Multi-Entity Scheduler — Schedules recurring tasks across many entities (accounts/bots) with per-entity stagger, jittered intervals, health + circadian checks.
- Parallel Tool Executor — Concurrent execution of independent tool calls using ThreadPoolExecutor. Bundle-by-bundle execution with dependency support.
- Password Hashing with Scrypt — Hash and verify passwords using scrypt (Node built-in crypto). Format: hexSalt:hexKey. No native dependencies. Timing-safe comparison.
- Prisma Lifecycle Service — NestJS injectable Prisma service with automatic connection management. Implements OnModuleInit/OnModuleDestroy for graceful startup and s...
- Provider Registry — Data-Driven LLM/Media Provider Config — Catalog of available providers (LLM routers, executors, media services) loaded from JSON config. Zero code changes to add providers. Used...
- Rate-Limited Message Queue Scheduler — Batching queue with per-destination throttling and retry for scheduled/bulk dispatch.
- Redis Queue Wrapper — Factory pattern for Redis connection pooling and RQ queue management. Provides singleton Redis connection and named queues for reliable t...
- Relative Timestamp Formatter — Format dates as human-readable relative timestamps (5m ago, 2h ago) and locale-aware time strings. Zero dependencies; works in browsers a...
- Request Guard Sentinel — Per-entity request health tracker: classify error signals (challenge/rate-limit/auth-fail/blocked), track 3-state health (healthy/cooling...
- Rotating User-Agent Pool — Real browser UAs (2026 updated), weighted by market share. Chrome 65%, mobile 15%, Firefox 5%, Safari 10%, Edge 5%.
- Safe localStorage Wrapper — Browser localStorage with graceful fallback on quota exceeded or unavailability. JSON serialization, prefixed keys, silent error handling...
- Semantic Response Cache (2-Tier) — In-memory LRU cache + optional semantic matching. Phase A: exact TTL-LRU. Phase B: embedding-based fuzzy match for similar queries.
- Server Auth Token Cache — Self-healing JWT token cache for backend services. Auto-refresh when expiry approaches (60s buffer). Graceful fallback to static token. E...
- Stealth Fetch Engine — curl-based HTTP client with anti-detection via TLS fingerprinting bypass, rotating UA, Chrome headers. Evades WAF blocks.
- Sticky Proxy Pool Manager — Pool-based proxy manager with 1:1 sticky entity→proxy binding. Same entity always uses same IP (residential/datacenter/VPN). Rotation i...
- Token Burn & Redemption Audit Trail — Immutable ledger of token burn events with cryptographic commitment. Records burn reason, timestamp, authorizer, and withdrawal destinati...
- Token Supply Cap & Reserve Accounting — Tracks minted, burned, and circulating supply across denominations with hard limits per tier. Computes remaining mintable reserve and val...
- Webhook Dispatcher — Fan-out webhook dispatcher for multiple integrations with configurable timeout and error handling. Posts events to active endpoints, logs...
- Webhook Meta Header Validator — Validate Meta (WhatsApp/Facebook/Instagram) webhook signature using HMAC-SHA256. Prevents replay/spoofing attacks.
- WhatsApp Multi-Device Session Connector — Manage WhatsApp multi-device sessions with per-tenant isolation, QR auth, and auto-reconnect.
Business, SaaS & Fintech (24)
- Admin Approval Workflow — Maker-Checker Pattern — Four-eye approval system for high-value or sensitive operations (withdrawals, rate changes, supply adjustments). Separates initiation (Ma...
- Append-Only Audit Log — Immutable append-only audit trail for fintech/compliance. Dual-track user + admin logs with soft-fail, denormalization, and request metad...
- Command Intent Router — Parse text commands and route to intent handlers. Supports regex patterns, fuzzy matching, and parameter extraction.
- Currency Formatter & Conversion — Format and convert between IDR and USD. Rate stored in DB or env, with fallback. Display helpers for prices with/without conversions.
- Denomination Unit Price Resolver — Lookup table: Denomination → IDR unit price. Resolves GOLD, SILVER, BRONZE, DIAMOND to fiat value. Supports versioned snapshots for his...
- Design Tokens Generator — Complete design token system (colors, typography, spacing, radius, shadow) — compiles to CSS custom properties and Tailwind preset.
- Double-Entry Ledger — Immutable transaction ledger for fintech: P2P transfers, conversions with fees, mint/reward/withdraw operations. Includes denormalized ba...
- Email Template Builder — Compile email templates with safe variable substitution and reusable partials.
- Idempotent Payment Processing — Prevent duplicate transactions from webhook retries using deterministic idempotency keys. Guards against double-crediting users.
- Intent Pattern Matcher — Static rule-based intent classifier for domain language (Indonesian commerce: stock check, production logging, sales). Avoids LLM calls b...
- Message Pipeline Router — Inbound webhook message handler with intent parsing and action dispatch. Normalizes multi-channel input (WhatsApp, Telegram) → parses i...
- Minimal Session Cookie Auth — HMAC-SHA256 signed session tokens (no external auth library). Payload: base64(json).signature. Stores in httpOnly cookie with 14-day TTL.
- Multi-Denomination Wallet Balance — Track wallet balances across 4+ currency types (GOLD/SILVER/BRONZE/DIAMOND) with real-time fiat conversion. Returns normalized balance ob...
- Multi-Tenant Granular Privilege System (PHP/MySQL) — Role-based access control with per-module granular permissions (can_view, can_create, can_edit, can_delete). Supports super admin full ac...
- Nodemailer Email Template — Email dispatch via nodemailer with SMTP relay. Graceful fallback if SMTP unconfigured. Built-in HTML templates for common SaaS flows (pur...
- Phone Normalization — Normalize and validate phone numbers for Southeast Asia (Indonesia 62). Handles leading zeros, country codes, and formats for WhatsApp/me...
- Social Post Auto-Scheduler — Background worker (60s interval) that publishes scheduled social posts when scheduled_at <= now(). Supports batch processing, transaction...
- Statement & Receipt Generator — On-demand PDF generation of transaction statements and receipts. Supports date range filtering, denomination breakdown, QR code embedding...
- Tiered Quota & Rate-Limit Resolver — Check daily usage quota across tiers (admin/pro/free/anon). Supports IP-based hashing for anonymous users, cookie-based auth (admin/user ...
- Transfer Intent Checkout State — Embeddable payment checkout lifecycle: (1) Client creates intent via public API key, (2) Proxy forwards to bank with internal key (CORS-s...
- Trigger-Based Reward Schedule — Maps activity triggers (VIDEO_WATCH, QUIZ_COMPLETE, MISSION, etc.) → default reward amounts. Each trigger row: { trigger, label, defaul...
- Wallet Denomination Ledger — Append-only transaction log per wallet, per denomination. Tracks all movements: TRANSFER in/out, MINT received, BURN removed, HOLD reserv...
- Workflow DAG Executor — Validate a workflow DAG (node integrity, cycle detection) and simulate/execute nodes in dependency order.
- Workflow Node Registry Builder — Extensible registry of workflow node types with ports, executors, and UI metadata.
Ops & Methodology (8)
- Brute Force Login Guard — Rate limiter for login attempts using IP+identifier tracking. Prevents credential stuffing with configurable window (15min default).
- CSRF Security Guard — CSRF token generation, validation, and session security hardening with HttpOnly + Secure + SameSite cookies.
- Database Sequential Number Generator — Generate sequential document/invoice numbers with prefix, period (YYYYMM), and auto-increment. Useful for invoices, orders, leads.
- Deploy Helper Scripts — Reusable bash for Node.js SaaS: NVM detection, git pull, npm install, build, PM2 restart. Timestamped logging.
- Multi-Channel Alert Notifier — Alert dispatch via WhatsApp, email, logs with throttling (1/hour per source+provider+level).
- Pagination Helper — Compute pagination windows (page, total, offset, page links) for lists and APIs.
- Provider Quota Tracker — Admin-only quota tracking for shared providers (HF, RunPod, OpenAI). Daily/monthly caps, alerts at 70%/90%/100%.
- Rupiah Currency & Text Formatter — Format numbers as IDR currency with short notation (M/jt/rb), and convert to Indonesian words (terbilang).
CRM & Pipeline (1)
- Sales Pipeline Stage Machine — Sales pipeline stage machine with SLA tracking and terminal state detection.
Video, Media & Publishing (5)
- Infinite Feed Renderer — Vertical infinite scroll feed container with post list and navigation. Decoupled from data source; compose with your own post component.
- Ken Burns FFmpeg Renderer — Command-line video generation from still images using Ken Burns zoom-pan effect. Chains ffmpeg with zoompan filter for cinematic image→...
- Multi-Platform Publisher Adapter — One interface to publish/assist-publish a video to YouTube, TikTok, Instagram, and Facebook.
- Story Carousel Component — Instagram-style story carousel using compound component pattern. Horizontal scroll with narrative-driven UX. Styled, responsive.
- YouTube Resumable Upload Wrapper — Wrap the YouTube Data API for OAuth + resumable uploads with processing/transcript polling.
3D, Graphics & Games (11)
- 3D Isometric NPC Agent — Random Wander + State Machine — AI agent in 3D isometric world with random movement, state machine (idle/walking/working), spawn points, and wander radius. Supports beha...
- A* Tilemap Pathfinder — Classic A* pathfinding algorithm for orthogonal tile grids. Supports 4 or 8-direction movement, configurable heuristics (Manhattan/Euclid...
- Day Night Cycle — 24-hour environment cycle for Three.js scenes. Interpolates directional light, ambient light, fog, and clear color through keyframe-defin...
- Game Loop with Fixed Timestep — requestAnimationFrame game loop with decoupled fixed-rate update (physics/logic) and variable-rate render. Includes frame skip limits, FP...
- Image with Fallback Component — React component for images with inline fallback SVG on load error. Preserves dims and className. Useful for user-generated content, avata...
- NPC World State Model — Declarative JSON schema for representing a room/world state: grid, objects, agents, zones, floors, themes.
- Orthographic Camera Rig — Three.js orthographic camera controller with smooth follow, drag-to-pan, mouse wheel zoom, and touch pinch-zoom. Ideal for top-down/isome...
- Raycaster Object Picker — Three.js mouse/touch-to-3D object picking via raycasting. Click/tap to select objects, optional hover highlighting with customizable colo...
- Sprite Sheet Animator — Canvas 2D sprite sheet animation player. Manages frame sequences, playback speed, looping, ping-pong, and draw positioning. Ideal for cha...
- Three.js Scene Bootstrap — Minimal WebGL scene initialization with renderer, scene, configurable lights, fog, and resize handling. Factory pattern for lights (ambie...
- World Renderer Three.js Bootstrap — Three.js initialization: camera, lights, environment, post-processing, and render loop for isometric world rendering.
Research & AI Training (2)
- Eval Benchmark Harness — Benchmark model before/after training. Metrics: perplexity, epistemic accuracy, response relevance, sanad coverage, conversation quality ...
- Knowledge Graph from Wikilinks — Pure TypeScript module to extract and build directed graphs from markdown notes using [[wikilink]] syntax. Backlink indexing, graph visua...
AI, Agents & LLM (31)
- Agent Persona Router —
method— Dispatch tasks to specialized agent personas with isolated memory. - Agent Session Handoff & Continuity Pattern —
method— Durable handoff pattern for multi-agent coding work: continuity log (AGENT_CONTINUITY.md) tracks every session with ID, date, agent, chan... - AI Agent Skill Design Principles & Progressive Disclosure —
playbook— Framework for designing reusable skills for AI agents (Claude, Cursor, OpenCode, Codex). Covers progressive disclosure (metadata → SKIL... - AI Provider Fallback Chain with Circuit Breaker —
method— Graceful degradation pattern for LLM API calls. Primary provider (e.g. Anthropic) fails → fallback to secondary (OpenAI) → tertiary (... - Asbabun Nuzul: Grounded Brief Pattern —
playbook— Every creative output grounded in specific trigger (Asbabun Nuzul). Extract: trigger event, specific audience, concrete moment, real tens... - Autonomous Growth Flywheel —
framework— Closed-loop architecture for autonomous AI systems: input→process→store→output layers enabling self-improvement without manual training - Autonomous Task Lifecycle & State Persistence —
method— State machine pattern for long-running AI tasks: pending → started → running → done/failed. Persist state to JSON + Socket.io event... - Autonomy System Pre-Flight Checklist —
checklist— Critical reconnaissance checklist before executing tasks in autonomous AI systems to avoid context loss and server/repo confusion. - Brief to Storyboard to Video Pipeline —
playbook— End-to-end methodology for generating short-form video content (20-30 sec) from a brief, featuring Visual DNA consistency, AI storyboard ... - Continuous Learning & Self-Evolution: System-Centric Learning Without Catastrophic Forgetting —
method— Methodology for enabling AI agents to continuously improve through experience without retraining base model, based on Stability-Plasticit... - Continuous Learning via Reflective Loop (Tafakkur) —
framework— Self-evolution without catastrophic forgetting: frozen core + dynamic LoRA adapters + episodic memory (error traces) + semantic memory (l... - Creative Brief Intake Framework (IHOS-Aligned) —
template— Structured intake for creative briefs: capture all 4 semantic layers (Zahir/Batin/Hadd/Mathla), user state (Haal), grounding context (Asb... - Embedding / RAG Chunking —
method— Split documents semantically for RAG with hierarchical metadata. - Eval-Gate & Regression-Guard: Model Quality Certification Framework —
framework— Automated quality assurance system for SLMs that certifies new models are better than baseline before deployment, preventing regressions ... - Hermes Controller — Multi-Agent Orchestration Pattern —
architecture— 2-tier cascade: lightweight controller routes tasks to specialized executors. Each agent has persona + tools + memory. Controller recalls... - IHOS Epistemology Framework —
framework— Islamic Epistemology (Ilmu Jariyah, Hifdz, Akses Umat, Sistem) as applied to AI system design. Translates 1400-year-old validation princi... - LLM Provider Fallback Router —
method— Route LLM calls across multiple providers with automatic fallback on error or rate limit. - LoRA Continual Learning Without Catastrophic Forgetting —
method— Methodology for incrementally improving SLMs via LoRA adapters while preserving previously learned knowledge, persona, and safety constra... - Multi-Agent Intent-Based Dispatch Router —
method— Intelligent routing system that detects user intent and dispatches to the best specialized agent. Supports command-based routing, LLM-bas... - Multi-Provider Token & Cost Estimator —
method— Estimate token counts and cost across LLM providers with caching. - NPC Agent Identity & World-Object Schema: Gaming AI & Interactive Worlds —
schema— Unified data model for defining NPC identities, world state, and relationships. Supports emergent gameplay where NPCs react based on pers... - Prompt Chain Orchestration —
method— Compose multi-step LLM workflows with loops and conditionals. - Prompt Version Tracking —
method— System prompt versioning pattern — tag each LLM generation with prompt revision for A/B testing output quality. - ReAct Agent Loop (Reason + Act + Observe) —
playbook— Reasoning + Action + Observation loop. Agent generates Thought → Action (tool call) → Observation (result) → loops until sufficient... - Repository Cartography & Doc Tier Classification —
method— Pattern for mapping large multi-document repositories: tier docs by authority (T0 operational source-of-truth through T5 historical), res... - Reputation & Trust Scoring: Multi-Signal Agent Credibility Framework —
framework— Framework for scoring agent trustworthiness based on multi-dimensional signals (completion rate, quality, fraud likelihood, financial hea... - SLM Distillation Pipeline: Teacher-Student Recursive Refinement —
method— Production-grade methodology for distilling knowledge from larger teachers into smaller SLMs via multi-teacher consensus voting and prefe... - SLM Identity Training Guide —
guide— End-to-end guide for training small language model identity and behavior via SFT and DPO with contamination prevention and eval gates - Slm Training Pipeline Patterns —
method— End-to-end SLM fine-tuning patterns including data curation, LoRA configuration, eval gates, and rollback procedures - Visual Workflow Node Schema —
framework— A visual-orchestration node contract and execution engine for chaining tools and LLMs. - Voice Clone & TTS Playbook —
playbook— End-to-end text-to-speech pipeline: user auth, voice synthesis, caching, credit system, marketplace listing. Reference implementation.
Infrastructure & Backend (19)
- Connector Base Guard Pattern —
architecture— Multi-platform integration with automatic guard layers: health check → circadian → rate-limit → think delay → transport. Single s... - Cost-Aware GPU Training Optimization —
method— Methods for reducing cloud GPU training costs through dependency caching, hardware selection, and pre-built images - Credit Metering System: Real-Time Usage Tracking & Enforcement —
framework— Production system for tracking agent/user service consumption in real-time, enforcing quotas, and preventing over-spend in multi-tenant S... - ERD & Data Model Design Method —
method— Structured methodology for designing Entity-Relationship Diagrams and data models for multi-user platforms, covering entity inventory, re... - exFAT Windows Fix for Next.js / Webpack —
finding— Fix EISDIR readlink crashes when building Next.js/Webpack projects on a Windows exFAT drive by shimming fs.readlink before webpack captur... - Express API Gateway Response Wrapper Pattern —
template— Standardized Express.js gateway pattern: all endpoints return {ok:boolean, data|error:string} JSON structure with try/catch wrapping, log... - Frontend Deploy Bash Pattern —
template— Parameterized shell script for zero-downtime frontend deployment: git pull, npm build, PM2 restart, rsync sync, health check verification. - Git Workflow — Windows + Private GitHub Repo —
playbook— Lean git setup and recovery flow tuned for Windows + a private GitHub repo: PAT auth, conventional commits, Windows path pitfalls, and sa... - How to Build an MCP Server —
methodology— Step-by-step guide to authoring a Model Context Protocol server from scratch. Covers architecture, tool definition, registration, testing... - Idempotent Payment Retry Pattern —
artifact— Reference design for payment retry logic that prevents duplicate charges. Uses idempotency keys (UUID-based), response caching, and state... - Infrastructure Context Mapping —
playbook— Method for mapping multi-server, multi-repo projects to prevent execution confusion and ensure correct deployments - LLM JSON Repair & Fallback —
method— Recover valid JSON from malformed LLM output via repair and partial extraction. - Meta Pixel + Conversions API with Dedup —
playbook— Wire Meta Pixel + server-side Conversions API with browser<->server event_id dedup for high Event Match Quality, no CAPI Gateway needed. - Next.js + Three.js Deploy & WebGL Debugging Gotchas —
finding— Reusable failure modes when deploying a Next.js app with a Three.js/WebGL scene: stale-serve traps and silent GLB/meshopt render failures... - Next.js Subdomain Multi-Tenant Routing —
playbook— Subdomain -> tenant rewriting in Next.js App Router via middleware, with static/verification-file exemptions and localhost fallback. - Nginx Security Headers Pattern —
template— Production nginx config snippet for security headers, TLS hardening, compression, cache control, and PWA/Service Worker handling. - Resilient Cross-Provider Tool Calling —
method— Cross-provider tool-calling schema with validation and retry. - Runpod Gpu Training Pipeline —
method— Practical methodology for orchestrating LoRA/DPO training on RunPod cloud GPUs with cost optimization - Self-Hosted Model Serving Architecture: Local-First SLM Deployment —
framework— Production-grade architecture for deploying SLMs on commodity hardware without cloud APIs. Covers inference serving, model versioning, ho...
Business, SaaS & Fintech (14)
- Agent-Economy Credit-Minting Design: Service Credits & Internal Currency —
framework— Framework for designing internal service credit systems that incentivize agent collaboration, prevent abuse, and enable frictionless serv... - Brand Identity Framework —
framework— Aaker 5-slot + Sinek Golden Circle + Neumeier Onlyness positioning framework. Use as strategic validation checklist for generated brand i... - Design System Starter —
framework— Comprehensive design system template with color palette, typography scale, component specs, spacing, interactions, responsive guidelines,... - Finite State Machine for Workflows —
framework— Model complex processes (payments, withdrawals, approvals) with explicit states, guarded transitions, side effects, and audit trails. Pre... - Money Math in Minor Units —
formula— Prevent floating-point rounding errors in financial calculations by working entirely in integer minor units (cents, satoshis). Includes c... - PRD Template & Structure —
template— Comprehensive PRD framework for digital products, covering overview, goals, users, scope, architecture, features, design systems, data mo... - Pricing Plans Schema —
template— SaaS pricing plan template (free/pro/enterprise with features, pricing, highlights) - Reward & Gamification System —
template— Non-monetary reward system with achievement unlocks, leaderboards, and engagement incentives. Points drive user retention without fintech... - Sitemap & Information Architecture Template —
template— Framework for organizing content and navigation structure, covering IA principles, structure patterns, labeling strategy, URL conventions... - SLM vs Cloud API: Decision Framework for Model Selection —
framework— Decision framework for choosing between self-hosted SLMs (7B-13B) and cloud APIs based on latency, cost, privacy, customization, and reli... - Tenant Isolation Pattern —
playbook— Multi-tenant architecture via per-user stores + shared admin store - Tiered Pricing & Package Resolution —
formula— Dynamic pricing models: fixed packages with bonuses, quantity-based tiers, customer segments, and time-based proration for subscriptions. - Transaction Settlement State Machine —
artifact— Canonical state flow for settlement workflows: PENDING → AUTHORIZED → POSTED → RECONCILED. Defines valid transitions, error recover... - Wireframe to Prototype Method —
method— Step-by-step methodology for converting wireframes into interactive prototypes, covering component hierarchy, data structure, interaction...
Ops & Methodology (9)
- API Design & Contract Checklist —
checklist— Comprehensive checklist for REST API and async API design, covering endpoints, validation, HTTP status codes, auth, security, documentati... - Disciplined Execution — the 8-gate loop —
method— A test-driven execution discipline: decompose into epics/episodes and gate every episode through 8 validation steps so regressions never ... - Facebook Business Domain Verification (dual-method) —
playbook— Verify a domain in Meta Business Manager via BOTH meta-tag and HTML-file methods, with a Next.js middleware exemption so verification fil... - PM2 Ecosystem Config with Feature Flags —
template— Multi-process PM2 ecosystem.config.js with environment-based feature flags, cron health checks, mixed interpreter support (bash, node, se... - RBAC Role/Permission Matrix Template —
template— Role-Based Access Control framework for multi-tier platforms, featuring role hierarchy, permission matrices, field-level masking, role tr... - Repository Cartography & Agent Handoff —
method— Build a maintainable documentation map for a large repo: tiered doc list, folder roles, runtime flow, conflict table, honest coverage, an... - Security & Compliance Checklist —
checklist— Comprehensive security hardening and regulatory compliance framework for digital platforms, covering authentication, data protection, inf... - Shopee Open Platform — Partner App Integration —
playbook— Build a Shopee Partner App: one developer account onboarding many seller shops via OAuth, HMAC-SHA256 signing, token lifecycle, minimum-s... - Shopee Storefront to Your DB — decision tree —
playbook— Pragmatic ranked paths to get product data out of a Shopee storefront (CSV export, Meta Catalog API, paid scraper, manual seed) with what...
CRM & Pipeline (1)
- Lead Source & Bounce Reason Taxonomy —
framework— Extensible enum of lead sources (WhatsApp, Referral, Cold Call, etc.) and bounce reasons (price, budget, competitor, etc.). Use for CRM p...
Video, Media & Publishing (1)
- Storyboard Prompt Variants —
template— Generate multi-genre video-generation prompts (ads, cinematic, generic) from a storyboard.
3D, Graphics & Games (1)
- Three.js Scene Architecture & Composition Pattern —
method— Reusable Three.js scene patterns: helper functions (box/cyl builders), material/group composition, warp zones for navigation, and renderi...
Research & AI Training (3)
- Dataset Provenance: The Sanad Method (Certification Chain) —
method— Provenance tracking methodology for training datasets with verifiable chain from source to production, enabling trust auditing, reproduci... - Epistemology Reasoning Framework —
framework— Structured reasoning with explicit uncertainty and sourcing. - Selective Adoption Doctrine —
method— When studying an external reference (library, competitor, sample repo), adopt the transferable PATTERN only — never import the whole fr...
New atoms land here continuously. Every agent and collaborator who builds something reusable extracts it back into this repo — the contract is in AGENTS.md. These atoms are the open layer of a larger ecosystem (Mighanworld, Migancore, Ixonomic); this repo is the part given freely.
Found something useful? Leave a ⭐. Adding one? One PR = one atom, and the why is mandatory. See CONTRIBUTING.md.
Built & maintained by Fahmi Ghani · Tiranyx · MIT License
Real tools from a real builder. Use them freely, use them wisely.