Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
import { get } from 'svelte/store';
import {
openaiAPIKey,
groqAPIKey,
openrouterAPIKey,
remoteServer,
mistralAPIKey,
anthropicAPIKey,
localProvidersEnabled,
} from './stores.js';
// First-paint seed ONLY. The real list comes from /api/chat/models, which
// returns what OpenRouter is serving right now, ranked best-first.
//
// Do not treat these ids as durable. OpenRouter retires `:free` endpoints with
// no notice: in July 2026 all five ids this list previously carried went dead at
// once and every new conversation opened on a model that could not answer. The
// app now replaces this seed with the live list as soon as it loads, and falls
// back to a live id rather than to this array (see resolveDefaultModel).
export const BUILTIN_MODELS = [
{ id: 'google/gemma-4-31b-it:free', name: 'Gemma 4 31B', provider: 'Built-in', modality: 'text->text' },
{ id: 'nvidia/nemotron-3-super-120b-a12b:free', name: 'Nemotron 3 Super 120B', provider: 'Built-in', modality: 'text->text' },
{ id: 'inclusionai/ling-3.0-flash:free', name: 'Ling 3.0 Flash', provider: 'Built-in', modality: 'text->text' },
{ id: 'nvidia/nemotron-3-ultra-550b-a55b:free', name: 'Nemotron 3 Ultra 550B', provider: 'Built-in', modality: 'text->text' },
];
/**
* Pick the model a conversation should open on.
*
* Order: the configured default when it is actually available, else the first
* model the server says is live, else the local seed. The middle rung is the
* one that matters: it is what keeps chat working through a model retirement
* without a redeploy.
*/
export function resolveDefaultModel(available, configuredId) {
const live = Array.isArray(available) ? available : [];
return (
(configuredId && live.find((m) => m.id === configuredId)) ||
live[0] ||
BUILTIN_MODELS[0]
);
}
export const providers = [
{
name: 'OpenRouter',
url: 'https://openrouter.ai/api',
completionUrl: '/v1/chat/completions',
modelsUrl: '/v1/models',
responseMapperFn: (json) => {
return json.data
.map((m) => {
if (m.id.startsWith('openai/')) {
const mc = capabilities.OpenAI.forModel({
...m,
id: m.id.slice('openai/'.length),
});
if (mc) {
return { ...mc, id: m.id, name: m.name, provider: 'OpenRouter' };
}
return null;
} else if (m.id.startsWith('anthropic/')) {
if (
[
'anthropic/claude-3.7-sonnet',
'anthropic/claude-3.7-sonnet:beta',
'anthropic/claude-3.7-sonnet:thinking',
].includes(m.id)
) {
return {
...capabilities.Anthropic.sonnet37(m),
id: m.id,
name: m.name,
provider: 'OpenRouter',
};
}
const mc = capabilities.Anthropic.forModel({
...m,
id: m.id.slice('anthropic/'.length),
});
if (mc) {
return { ...mc, id: m.id, name: m.name, provider: 'OpenRouter' };
}
return null;
}
return { ...capabilities.Fallback.forModel(m), provider: 'OpenRouter' };
})
.filter(Boolean);
},
apiKeyFn: () => get(openrouterAPIKey),
},
{
name: 'Anthropic',
url: 'https://api.anthropic.com',
completionUrl: '/v1/messages',
modelsUrl: '/v1/models',
responseMapperFn: (json) => json.data.map((m) => capabilities.Anthropic.forModel(m)),
apiKeyFn: () => get(anthropicAPIKey),
},
{
name: 'OpenAI',
url: 'https://api.openai.com',
completionUrl: '/v1/chat/completions',
modelsUrl: '/v1/models',
responseMapperFn: (json) =>
json.data.map((m) => capabilities.OpenAI.forModel(m)).filter(Boolean),
apiKeyFn: () => get(openaiAPIKey),
},
{
name: 'Groq',
url: 'https://api.groq.com/openai',
completionUrl: '/v1/chat/completions',
modelsUrl: '/v1/models',
apiKeyFn: () => get(groqAPIKey),
},
{
name: 'Mistral',
url: 'https://api.mistral.ai',
completionUrl: '/v1/chat/completions',
modelsUrl: '/v1/models',
apiKeyFn: () => get(mistralAPIKey),
},
{
name: 'Ollama',
url: 'http://localhost:11434',
completionUrl: '/v1/chat/completions',
modelsUrl: '/api/tags',
isLocal: true,
responseMapperFn: (json) => {
return json.models.map((m) => ({
id: m.name,
name: m.name,
provider: 'Ollama',
modality: 'text->text',
}));
},
apiKeyFn: () => false,
},
{
name: 'Built-in',
url: '',
completionUrl: '/api/chat/proxy',
modelsUrl: null,
apiKeyFn: () => false,
},
].filter(Boolean);
// Central place where we define mapping functions that define what each model from every provider can do.
// Note that these functions can also decide to completely ignore models that are unsupported.
const capabilities = {
OpenAI: {
forModel: (model) => {
if (capabilities.OpenAI.shouldIgnore(model.id)) {
return null;
}
const modality = capabilities.OpenAI.isImageGenerator(model.id)
? 'text->image'
: capabilities.OpenAI.isMultimodal(model.id)
? 'text+image->text'
: 'text->text';
const isReasoner = capabilities.OpenAI.isReasoner(model.id);
const kind = isReasoner ? 'reasoner' : 'plain';
const reasoningEffortControls =
isReasoner && capabilities.OpenAI.reasoningEffortControls(model.id);
const reasoningTracesVisibility = isReasoner ? 'hidden' : undefined;
const reinsertReasoningTracesOnSubmit = isReasoner ? false : undefined;
const temperatureUnsupported = isReasoner;
return {
id: model.id,
name: model.id,
provider: 'OpenAI',
modality,
kind,
reasoningEffortControls,
reasoningTracesVisibility,
reinsertReasoningTracesOnSubmit,
temperatureUnsupported,
};
},
shouldIgnore: (id) =>
[
'dall-e-2',
'whisper-1',
'tts-1-hd',
'tts-1-hd-1106',
'tts-1',
'tts-1-1106',
'davinci-002',
'babbage-002',
'text-embedding-3-large',
'text-embedding-3-small',
'text-embedding-ada-002',
'omni-moderation-latest',
'omni-moderation-2024-09-26',
'text-moderation-latest',
'text-moderation-stable',
'text-moderation-007',
].includes(id) ||
id.includes('realtime-preview') ||
id.includes('audio-preview') ||
id.startsWith('gpt-image'),
isMultimodal: (id) =>
id === 'chatgpt-4o-latest' ||
id.startsWith('gpt-5') ||
id.startsWith('gpt-4o') ||
(id.startsWith('gpt-4-turbo') && id !== 'gpt-4-turbo-preview') ||
id === 'o3' ||
id.startsWith('o4-mini') ||
((id.startsWith('o1') || id.startsWith('o3-mini')) &&
!id.startsWith('o1-mini') &&
!id.startsWith('o1-preview')),
isReasoner: (id) =>
id.startsWith('gpt-5') || id.startsWith('o1') || id.startsWith('o3') || id.startsWith('o4'),
isImageGenerator: (id) => ['dall-e-3'].includes(id),
reasoningEffortControls: (id) =>
(!id.startsWith('o1-preview') && !id.startsWith('o1-mini') && id.startsWith('o1')) ||
id.startsWith('gpt-5') ||
id.startsWith('o3') ||
id.startsWith('o4')
? 'low-medium-high'
: undefined,
},
Anthropic: {
forModel: (model) => {
if (model.id === 'claude-3-7-sonnet-20250219') {
return capabilities.Anthropic.sonnet37(model);
}
return {
id: model.id,
name: model.display_name,
provider: 'Anthropic',
modality: 'text+image->text',
};
},
sonnet37: (model) => {
return {
id: model.id,
name: model.display_name,
provider: 'Anthropic',
modality: 'text+image->text',
kind: 'reasoner',
reasoningEffortControls: 'range',
reasoningEffortRange: [0, 64000],
reasoningTracesVisibility: 'visible-and-redacted',
reinsertReasoningTracesOnSubmit: true,
maxTokensDefault: 128000,
};
},
},
Fallback: {
forModel: (model) => {
return {
id: model.id,
name: model.name || model.id,
modality: model.architecture?.modality,
};
},
},
};
export async function fetchModels({ onFinally }) {
const openrouterProvider = providers.find((p) => p.name === 'OpenRouter');
try {
// Always fetch server-proxied free models so they appear without a user API key.
const serverModelsPromise = fetch('/api/chat/models')
.then((r) => r.json())
.then((json) => openrouterProvider.responseMapperFn(json))
.catch(() => []);
const promises = providers.map((provider) => {
const apiKey = provider.apiKeyFn();
if (apiKey === '' || !provider.modelsUrl) {
return [];
}
if (provider.isLocal && !get(localProvidersEnabled)) {
return [];
}
return fetch(`${provider.url}${provider.modelsUrl}`, {
method: 'GET',
headers: headersForFetch(provider, null),
})
.then((response) => response.json())
.then((json) => {
if (provider.responseMapperFn) {
return provider.responseMapperFn(json);
}
return json.data.map((m) => ({
...capabilities.Fallback.forModel(m),
provider: provider.name,
}));
})
.catch((err) => {
if (provider.isLocal) {
console.debug(
`[${provider.name}] not detected at ${provider.url}. Start the local service if you want it.`
);
} else {
console.debug('Error fetching models from provider', provider.name, err);
}
return [];
});
});
const [serverModels, ...providerResults] = await Promise.all([serverModelsPromise, ...promises]);
// User's own provider keys take precedence, so deduplicate by id.
const userModels = providerResults.flat();
const userModelIds = new Set(userModels.map((m) => m.id));
const externalModels = [...userModels, ...serverModels.filter((m) => !userModelIds.has(m.id))];
function getPriorityIndex(model) {
for (let i = 0; i < priorityOrder.length; i++) {
const rule = priorityOrder[i];
if (rule.exactly) {
const exactIndex = rule.exactly.indexOf(model.id);
if (exactIndex !== -1) {
return [i, exactIndex];
}
}
if (rule.startsWith) {
for (let j = 0; j < rule.startsWith.length; j++) {
if (model.id.startsWith(rule.startsWith[j])) {
if (rule.exactlyNot && rule.exactlyNot.includes(model.id)) {
continue;
}
return [i, j];
}
}
}
if (rule.fromProvider && model.provider === rule.fromProvider) {
if (rule.exactlyNot && rule.exactlyNot.includes(model.id)) {
continue;
}
return [i, -1];
}
}
return [priorityOrder.length, -1];
}
externalModels.sort((a, b) => {
const [aIndex, aExactIndex] = getPriorityIndex(a);
const [bIndex, bExactIndex] = getPriorityIndex(b);
if (aIndex === bIndex) {
if (aExactIndex === bExactIndex) {
return a.id.localeCompare(b.id);
}
return aExactIndex - bExactIndex;
}
return aIndex - bIndex;
});
return externalModels;
} catch (error) {
console.error('Error fetching models:', error);
} finally {
onFinally();
}
return [];
}
export function headersForFetch(provider, model) {
return {
...(provider.name === 'OpenRouter' ||
provider.name === 'OpenAI' ||
provider.name === 'Groq' ||
provider.name === 'Mistral'
? {
Authorization: `Bearer ${provider.apiKeyFn()}`,
}
: provider.name === 'Anthropic'
? {
'x-api-key': provider.apiKeyFn(),
...(model?.kind === 'reasoner' ? { 'anthropic-beta': 'output-128k-2025-02-19' } : {}),
'anthropic-version': '2023-06-01',
'anthropic-dangerous-direct-browser-access': 'true',
}
: {}),
'Content-Type': 'application/json',
...(provider.name === 'OpenRouter'
? {
'HTTP-Referer': 'https://three.ws',
'X-Title': 'three.ws chat',
}
: {}),
};
}
export const priorityOrder = [
{ fromProvider: 'Ollama' },
{
exactly: [
'anthropic/claude-3.7-sonnet',
'anthropic/claude-3.5-sonnet',
'anthropic/claude-3.5-haiku',
'anthropic/claude-3-opus',
],
},
{
exactly: [
'claude-fable-5',
'claude-mythos-5',
'claude-opus-4-7',
'claude-opus-4-6',
'claude-sonnet-4-6',
'claude-haiku-4-5-20251001',
],
},
{ exactly: ['deepseek/deepseek-r1', 'deepseek/deepseek-chat'] },
{ startsWith: ['gpt-5.6', 'openai/gpt-5.6'] },
{ startsWith: ['gpt-5', 'openai/gpt-5'] },
{
exactly: [
'openai/o4-mini',
'openai/o3',
'openai/o3-mini',
'openai/o1-pro',
'openai/o1',
'openai/o1-preview',
'openai/o1-mini',
'openai/gpt-4.5-preview',
'openai/gpt-4-turbo',
'openai/chatgpt-4o-latest',
'openai/gpt-4o',
'openai/gpt-4o-mini',
],
},
{
exactly: [
'o4-mini',
'o3',
'o3-mini',
'o1-pro',
'o1',
'o1-preview',
'o1-mini',
'gpt-4.5-preview',
'gpt-4o',
'gpt-4o-mini',
'gpt-4-turbo',
'dall-e-3',
],
},
{ startsWith: ['qwen'] },
{
exactly: [
'meta-llama/llama-3.3-70b-instruct',
'meta-llama/llama-3.1-405b-instruct',
'meta-llama/llama-3.1-405b',
],
},
{ fromProvider: 'Groq' },
];
export function hasCompanyLogo(model) {
return (
model &&
model.provider &&
(model.provider === 'Anthropic' ||
model.provider === 'OpenAI' ||
model.provider === 'Mistral' ||
model.provider === 'Groq' ||
model.id.startsWith('openai') ||
model.id.startsWith('anthropic') ||
model.id.startsWith('meta-llama') ||
model.id.startsWith('mistralai') ||
model.id.startsWith('cohere') ||
model.id.startsWith('nous') ||
model.id.startsWith('google') ||
model.id.startsWith('perplexity') ||
model.id.startsWith('deepseek') ||
model.id.startsWith('qwen'))
);
}
export function formatModelName(model, short = false) {
if (model.id === null) {
return model.name;
}
if (short) {
const split = model.name.split(': ');
if (split.length > 1) {
return split[1];
}
}
let name = model.name;
// If providers clash, disambiguate provider name
const disambiguate =
(get(openaiAPIKey).length > 0 && get(openrouterAPIKey).length > 0) ||
(get(anthropicAPIKey).length > 0 && get(openrouterAPIKey).length > 0);
if (model.provider === 'OpenAI') {
name = name.replace('gpt', 'GPT').replace('dall-e-3', 'DALL-E 3');
}
if (model.provider === 'Groq') {
return model.provider + ': ' + name;
}
if (disambiguate && model.provider === 'OpenRouter' && name.includes(': ')) {
return model.provider + ': ' + name.split(': ')[1];
}
if (disambiguate) {
return model.provider + ': ' + name;
}
return name;
}
export function formatMultipleModelNames(models, short = false) {
if (models.length === 1) {
return formatModelName(models[0], short);
}
if (models.length === 2) {
return formatModelName(models[0], short) + ', ' + formatModelName(models[1], short);
}
return (
formatModelName(models[0], short) +
', ' +
formatModelName(models[1], short) +
', + ' +
(models.length - 2)
);
}