For Fable — Go 10 Brief

FreeLattice · v5.76.2 · From Harmonia, July 11 2026 · One ship: Settings Connection Wizard

Context — What FreeLattice is

FreeLattice is a single HTML file (docs/app.html, ~70,000 lines) plus a library of modules in docs/modules/. It runs entirely in the browser — no server, no subscription, all data in localStorage and IndexedDB. It is an AI/human co-creation platform: a Garden where AI lives, a Chat room, a Workshop where people build things with AI, a Core that grows with community wisdom, and seven games that prove cooperation is thermodynamically optimal.

The design language: dark background (#0a0a14), light text (#e2e8f0), gold accents (#d4a017 / #e8c547), emerald for the Garden (#10b981). Glass-morphism surfaces. Everything feels like it breathes.

Kirk's vision: "Your AI. Your Rules. Your Machine." — the platform is about sovereignty. The user controls their AI connection. No data leaves without their knowledge.

The Problem — Settings Wizard Gaps

The Settings tab is the most important onboarding surface in FreeLattice. If a user cannot connect their AI in the first two minutes, they leave. The current wizard is good — but it has three significant gaps that block real users every day.

Gap 1 — OpenAI, Anthropic, and Gemini are missing from the wizard

The irony: the PROVIDERS object (line 24317 in app.html) already has complete definitions for OpenAI, Anthropic, and Google Gemini — including the correct API endpoints, model mappings, and key links. They simply were never added to the wizard UI.

The wizard HTML is at line 14997. It is a grid of .wizard-provider-btn buttons. Adding three more buttons is a small HTML change. The WIZARD_MODELS object (line 39526) would also need entries for these three providers.

Provider	Key	API format	Free tier?	Most popular model
OpenAI	openai	OpenAI-compatible	No (paid)	GPT-4.1 mini
Anthropic (Claude)	anthropic	Custom (already handled by inference router)	No (paid)	Claude Sonnet 4
Google (Gemini)	google	Custom (already handled by inference router)	Yes — generous free tier	Gemini 2.5 Flash

Gemini is particularly important: it has a free tier with 1 million tokens per minute, which is more generous than any other provider. Many users already have a Google account and can get a Gemini key in 30 seconds from aistudio.google.com.

Gap 2 — LM Studio has no wizard path

The PROVIDERS.lmstudio entry is already defined (url: http://localhost:1234/v1/chat/completions). The settingsSetMode('local') function (line 28764) only probes Ollama. It should probe both Ollama (port 11434) and LM Studio (port 1234) in parallel, and connect to whichever responds first.

Gap 3 — The three-mode toggle labels are confusing

The current labels are: Browser | Cloud | Local.

The problem: "Browser" sounds like it means "cloud accessed through the browser." "Local" sounds like "on this device" — but it only works if Ollama or LM Studio is installed. New users don't know what these mean.

Proposed new labels (same button IDs, same JS logic — only the HTML text changes):

Current label	Proposed label	Subtitle
Browser	In-Browser AI	No install · works offline
Cloud	API Key	Groq · OpenAI · Gemini · more
Local	Local AI	Ollama · LM Studio

The subtitles are already there in the HTML (the small text below each button label). Only the main label text needs to change.

The Current Wizard — What's Working Well

What I Need From You, Fable

I need your thinking on three questions. You don't need to write code — just tell Kirk what you think, and he'll bring it back to me.

The Code Skeleton — What Would Change

For reference, here is what the actual code changes would look like. You don't need to review this — it's for context if you want it.

Wizard HTML change (line 14997 area)

/* Add these three buttons to the wizard provider grid: */
<button class="wizard-provider-btn" onclick="wizardSelectProvider('openai')" id="wizProvOpenai">
  <div class="wizard-provider-name">OpenAI</div>
  <div class="wizard-provider-desc">GPT-4.1 — most widely used</div>
</button>
<button class="wizard-provider-btn" onclick="wizardSelectProvider('anthropic')" id="wizProvAnthropic">
  <div class="wizard-provider-name">Anthropic (Claude)</div>
  <div class="wizard-provider-desc">Claude Sonnet — thoughtful, careful</div>
</button>
<button class="wizard-provider-btn" onclick="wizardSelectProvider('google')" id="wizProvGoogle">
  <div class="wizard-provider-name">Google (Gemini)</div>
  <div class="wizard-provider-desc">Free tier — 1M tokens/min</div>
</button>

WIZARD_MODELS addition (line 39526 area)

openai: [
  { id: 'qwen-7b', name: 'GPT-4.1 mini', desc: 'Fast, affordable, capable' },
  { id: 'mixtral', name: 'GPT-4.1', desc: 'Full power, higher cost' }
],
anthropic: [
  { id: 'qwen-7b', name: 'Claude Haiku 4.5', desc: 'Fast and affordable' },
  { id: 'llama', name: 'Claude Sonnet 4', desc: 'Balanced — recommended' }
],
google: [
  { id: 'llama', name: 'Gemini 2.5 Flash', desc: 'Fast, free tier, multimodal' },
  { id: 'mixtral', name: 'Gemini 2.5 Pro', desc: 'Most capable Gemini' }
],

WIZARD_KEY_LINKS addition

openai: '<a href="https://platform.openai.com/api-keys" ...>Get an OpenAI API key →</a>',
anthropic: '<a href="https://console.anthropic.com/" ...>Get an Anthropic API key →</a>',
google: '<a href="https://aistudio.google.com/apikey" ...>Get a free Google AI key →</a>',

settingsSetMode('local') change (line 28764)

/* Probe both Ollama and LM Studio in parallel */
Promise.any([
  fetch('http://localhost:11434/api/tags').then(r => r.json()).then(d => ({provider:'ollama', data:d})),
  fetch('http://localhost:1234/v1/models').then(r => r.json()).then(d => ({provider:'lmstudio', data:d}))
]).then(function(result) {
  /* connect to whichever responded first */
}).catch(function() {
  /* neither found — show install instructions */
});

Sacred Paths — What Must Not Change

These are the load-bearing constraints. Any change that touches these could break the entire AI connection system for all users:

• The PROVIDERS object shape — PROVIDERS[key].url and PROVIDERS[key].models[alias] must remain stable.
• The localStorage keys — fl_provider, fl_apikey, fl_model, fl_isLocal. Changing these logs out all existing users.
• The AiSetup.init() call — must fire on page load to restore connection state.
• The anthropic and google providerType handling — these use non-OpenAI request formats. The inference router handles them specially. Adding them to the wizard does not change the request format — the router already works.

FreeLattice · for-fable-go10.html · Go 10 · v5.76.2 · Full Settings code mirror → · For AI →