Akeh sistem AI nindakake kesalahan sing padha.

Dheweke takon model basa dhisik lan ngarep-arep jawaban keprungu bener.

Watch 30 second demo

Alam wis mecahake masalah iki pirang-pirang yuta taun kepungkur.

Ing sarang lebah, panemu ora langsung dadi keputusan amarga siji aktor ngomong ngono. Lebah pengintai bali menyang sarang lan nari wolu-angka ing permukaan tegak madu — sudut bagian lurus nuduhake arah, dawane nuduhake jarak, kuwat nuduhake kualitas. Nanging tariane dudu monolog. Lebah-lebah sadulur ngetutake panari, nyenggol nganggo antenna, lan menehi umpan balik ing wektu nyata. Sinyal mandheg bisa mbatalake tari kanthi sakabehe. Mung yen pesen ngatasi pemeriksaan komunitas, rute sing kudu diikuti bakal muncul.

WaggleDance dibangun ing logika iki.

Iku ora langsung menehi masalah menyang LLM. Iku pisanan ngarahake marang solver sing bener, verifikasi asil liwat pirang-pirang agen, lan nggunakake model basa mung yen pancen mbantu. Saben langkah ninggalake jejak sing bisa diaudit. Saben solusi bisa dibenerake. Saben siklus mbangun keahlian sistem dhewe.

Tari wolu-angka dadi routing algoritmik. Madu dadi struktur memori MAGMA. Lan istirahat wengi lebah dadi Dream Mode — simulasi nalika sistem ngreviu kegagalan dina, nyoba ewonan rute alternatif, lan tangi tambah pinter.

Iki dudu metafora. Iki arsitektur kanggo kecerdasan mesin kolektif.

Solver-first, dudu chat-first
Offline / piyambak / perangkat dhewe
MAGMA: replay, asal-usul, kapercayan
Clone lan coba saiki

⮕ Clone & Run

Download, fork, lan jalanake lokal saiki. Kabeh repo kasedhiya ing GitHub tanpa registrasi.

License model: Apache 2.0 + BUSL 1.1 (open core + source-available protected modules). Check the terms on GitHub.
BUSL module change date: March 18, 2030.

Sebabe Iki Beda

AI sing ora nebak

Solver dhisik. Verifier ngecek. LLM mlebu mung yen solver sing tepat ora cukup.

AI sing kelingan kabeh

MAGMA nyathet keputusan, sumber, replay, lan skor kepercayaan. Delengen apa sing kedadeyan, kenapa, lan urutane kepiye.

AI sing sinau sewengi

Dream Mode mriksa kegagalan, nyimulasi rute sing luwih apik, lan mbangun model sing luwih apik kanggo dina candhake.

AI sing nuduhake kahanane

Hologram Brain ndadekake kahanan 32 node katon ing wektu nyata. Sampeyan ora ndeleng kothak ireng — sampeyan ndeleng sistem sing mlaku.

AI sing tetep ing jaringan sampeyan

Kabeh mlaku ing lingkungan sampeyan dhewe. Ora ana mega wajib, ora ana data prompt sing metu, ora ana ketergantungan SaaS.

AI sing bisa scale

Codebase sing padha bisa mlaku saka Raspberry Pi nganti profil pabrik. Ora mung demo, ora mung kerangka.

Profil Deploy — runtime sing padha, papat profil

Piranti GADGET

RPi, edge, sensor

Pondhok COTTAGE

Offline, sambungan ora tetep

Omah HOME

Otomatisasi lokal

Pabrik FACTORY

Pemantauan, anomali, audit

Dashboard lan Hologram Brain kasedhiya langsung sawise startup. Kacepetan respon pisanan gumantung marang profil, hardware, lan apa full utawa stub mode sing digunakake.

Apa sing kedadeyan nalika klik “Analisis Repositori”?

Prompt dikirim langsung menyang Grok — lan disalin menyang clipboard minangka cadangan

Grok dibukak ing tab anyar kanthi prompt siap

Sampeyan entuk analisis komprehensif repo, perbandingan pesaing kanthi nilai, lan penilaian kesiapan pabrik.

Yen pre-fill ora bisa, tempel manual — prompt wis ana ing clipboard sampeyan.

Sampeyan uga bisa nggunakake prompt sing padha ing Claude, ChatGPT, utawa LLM liyane. Grok minangka pilihan standar ing kaca iki.

Apa sing Dianalisis Grok

Nalika sampeyan klik “Analisis Repositori”, AI nindakake analisis jero sing nyakup:

Kahanan codebase saiki

Cabang utama, arsitektur, modul, lan commit paling anyar

README vs kenyataan

Apa sing wis diimplementasikake vs apa sing direncanakake utawa aspirasional

Tes lan kematangan

Cakupan tes, kematangan praktis, lan kesiapan produksi

Hologram Brain lan MAGMA

Model memori, arsitektur audit, asal-usul, lan mekanisme kapercayan

Perbandingan pesaing

Dinilai 1-10 ing enem sumbu vs Omah Assistant, Node-RED, n8n, Open WebUI, LangGraph, AutoGen, CrewAI, Ollama

Penilaian deploy pabrik

Kasus penggunaan industri, risiko, integrasi sing kurang, panghalang deployment

Prompt analisis:

Prompt Grok kelanjutan

Klik prompt kanggo nyalin. Tempel ing sesi Grok sampeyan kanggo eksplorasi sing luwih jero.

Kepiye Aku Nyambungake WaggleDance?

Pilih profil lan entuk pandhuan deploy sing disesuaikan saka Grok.

Kepiye WaggleDance Dibandhingake

Saben alat ing ngisor iki apik kanggo tugase dhewe. Perbandingan iki nuduhake carane arsitektur solver-first WaggleDance beda — dudu kanggo ngomong yen liyane elek.

vs. Omah Assistant

HA: Deterministic rules and automations, but no solver-based routing before the LLM.
WD: Solver-first routing (7+ deterministic solvers) → verifier → LLM only as fallback. Every decision auditable via MAGMA trail.
HA: No autonomous model training, no overnight learning.
WD: 8 sklearn specialist models + Dream Mode overnight learning with canary lifecycle.
HA's advantage: excellent integration ecosystem (2000+ integrations).

vs. LangGraph

LG: Graph-based multi-agent, but LLM-centric — everything goes through the LLM.
WD: Solver-first. LLM is Layer 1 (last), not Layer 3 (first).
LG: No append-only auditing, no canary model training, no dream mode simulation.
WD: MAGMA 5-layer provenance + 8 specialist models + counterfactual simulations.
LG's advantage: stronger cloud ecosystem and documentation.

vs. AutoGen / CrewAI

AG/CA: Multi-agent frameworks, but without deterministic solvers.
WD: 7+ deterministic solvers are routed BEFORE any LLM call.
AG/CA: No edge/factory profiles, no offline-first architecture.
WD: 4 profiles (GADGET → FACTORY), fully offline, from ESP32 to DGX.
AG/CA: No autonomous overnight learning or canary promotion.

vs. Ollama / LocalAI

Ollama: Local LLM engine, no decision-making architecture.
WD: Uses Ollama as one component (Layer 1 fallback), but builds solver routing, MAGMA auditing, specialist models, and Dream Mode on top.
Ollama is the engine. WaggleDance is the whole car.

vs. n8n / Node-RED

n8n/NR: Visual workflow automation tools, excellent flow editors.
WD: Not a visual flow editor but an autonomous multi-agent runtime that learns and improves.
n8n/NR: No sklearn models, no append-only provenance, no counterfactual simulation.
WD: 8 models + 9 SQLite databases + ChromaDB/FAISS + Dream Mode.

Deploy — WD's Advantage

Docker: clone → docker compose up -d — Ollama, Voikko (Finnish NLP), and the app all in one.
No separate manual installations in Docker mode.
4 profiles with automatic hardware detection (GADGET / COTTAGE / HOME / FACTORY).

Time Evolution — WD's Decisive Advantage Over ALL Competitors

No competitor improves autonomously over time. WaggleDance is the only one that builds cumulative expertise.

Time	WaggleDance	Omah Assistant	LangGraph	AutoGen/CrewAI	Node-RED/n8n	Ollama
Day 1	LLM fallback ~30-50%, solvers learning	Same as always	Same as always	Same as always	Same as always	Same as always
Month 1	HotCache fills, LLM ~20-30%, first canary promotions	No change	No change	No change	No change	No change
Month 6	LLM ~10-15%, specialists maturing, ~180 nights of Dream Mode	No change	No change	No change	No change	No change
Year 1	LLM ~5-8%, MAGMA with thousands of audited paths	No change	No change	No change	No change	No change
Year 2	LLM <3-5%, >95% deterministic, TCO a fraction of day 1	No change	No change	No change	No change	No change

The competitors' column is empty everywhere except day 1. They don't learn. They don't improve. On day 730, they are exactly the same as on day 1.

Pitakonan sing Asring Ditakokake

Is WaggleDance Swarm AI free?

Yes. Download and run immediately. Apache 2.0 parts are freely usable. Non-commercial personal use of BUSL-protected modules is permitted. For commercial use, check the license terms on GitHub.

Does it require an internet connection?

No. WaggleDance is designed to work fully offline on local hardware. Internet is only needed for initial setup and updates.

What hardware is needed?

Minimum: Raspberry Pi 4 or equivalent (GADGET profile). Recommended: modern x86 server for multi-agent orchestration (FACTORY profile).

Why Grok for analysis?

You get a quick second technical opinion on the public repo, documentation, and competitive landscape. You can use the same prompt in Claude, ChatGPT, or any other LLM.

What is MAGMA?

An auditing and provenance framework. Every agent decision is recorded so you get traceability, replay, and trust assessment visibility.

What is Dream Mode?

An overnight learning mode where the system reviews the day's failures, simulates better routes, and builds better models for the next day — automatically without user action.

What happens after first startup?

Dashboard and Hologram Brain are available immediately. First response speed depends on profile and hardware.