@parallel-cli/parallel 0.4.1 → 0.4.3

package/dist/i18n.js

@@ -55,6 +55,10 @@ const en = {
     'wiz.provider.name.ph': 'e.g. openrouter, mistral, ollama…',
     'wiz.provider.url.title': 'Base URL (OpenAI-compatible endpoint)',
     'wiz.provider.url.ph': 'e.g. https://openrouter.ai/api/v1',
+    'wiz.provider.endpoint.title': 'Review endpoint for {name}',
+    'wiz.provider.endpoint.model': 'Model: {model}',
+    'wiz.provider.endpoint.use': 'Use this endpoint',
+    'wiz.provider.endpoint.edit': 'Edit endpoint',
     'wiz.provider.model.title': "Model name — exactly as written in the provider's documentation",
     'wiz.provider.model.ph': 'e.g. deepseek-chat, gpt-4o-mini, qwen2.5-coder…',
     'wiz.provider.key.title': 'API key for {name}',
@@ -117,7 +121,7 @@ const en = {
     'help.l2d': ' for everyone).',
     'help.l3': "Agents see each other's statuses and diffs before every action: they co-edit the same files without blocking and without breaking each other's work.",
     'help.states': 'States: 🔨 working · 👂 listening to other agents (double border) · 🧠 thinking · ✋ waiting for your approval · ⏹ stop. Sounds: 1 beep = agent launched/finished, 2 beeps = approval required (/sound off to mute).',
-    'help.keys': 'Esc: back to agents view · Tab: completion · ↑/↓: history · Ctrl+V: paste image',
+    'help.keys': 'Esc: back to agents view · Tab: completion · ↑/↓: suggestions, history, or view scroll · Ctrl+V: paste image',
     // commands (descriptions)
     'cmd.ask': 'Ask-only agent: answers and advises without editing',
     'cmd.task': 'Task agent: executes, edits, validates, summarizes',
@@ -155,6 +159,8 @@ const en = {
     'cmd.doctor': 'Check provider/model/API key configuration and repair hints',
     'cmd.settings': 'Global settings — providers, API keys, language, defaults (persisted)',
     'cmd.ssettings': 'Session settings — model, approvals, sound (this session only)',
+    'cmd.project': 'Change project folder or reopen folder picker',
+    'cmd.wizard': 'Relaunch the setup wizard',
     'cmd.approvals': 'Shell approvals: ask, auto-safe, or yolo (this session)',
     'cmd.sound': 'Sound cues (this session)',
     'cmd.save': 'Save the session now (optionally with a name)',
@@ -252,7 +258,7 @@ const en = {
     'm.usageSession': 'Usage: /sessions to list, then /session <n> or /session latest',
     'm.sessionLoaded': '📂 Session from {date} loaded.',
     'm.missingProvider': 'No provider configured. Open /settings → Add a provider, or restart with a first-run config.',
-    'm.missingKey': 'Provider {name} has no API key. Use /key <API key> or /settings → API key.',
+    'm.missingKey': 'Provider {name} has no API key. Use /settings → Providers → API key.',
     'm.missingModel': 'Provider {name} has no default model. Use /settings → Provider models.',
     'm.doctorOk': '✓ Configuration ready: {pm}. Changes: /settings · temporary model: /model [provider:]model',
     'm.spawnFail': 'Cannot launch the agent: no usable provider/model. Configure one via /settings.',
@@ -330,9 +336,12 @@ const en = {
     'set.newSkillName': 'Skill name (a .md template will be created in ~/.parallel/skills)',
     'set.newSpecialistName': 'Specialist name (a .md template will be created in ~/.parallel/specialists)',
     // provider-pick section headers
-    'wiz.provider.section.configured': 'Configured',
-    'wiz.provider.section.cloud': 'Cloud Providers',
-    'wiz.provider.section.local': 'Local',
+    'wiz.provider.section.configured': 'Configured Providers',
+    'wiz.provider.section.western': 'Western',
+    'wiz.provider.section.chinese': 'Chinese',
+    'wiz.provider.section.gateways': 'Gateways',
+    'wiz.provider.section.inference': 'Inference',
+    'wiz.provider.section.local': 'Local Models',
     'wiz.provider.customDetail': 'Configure manually',
     'wiz.provider.ollamaDetail': 'Local — no API key needed',
     'wiz.provider.ollama.checking': 'Checking Ollama at localhost:11434...',
@@ -347,6 +356,8 @@ const en = {
     'set.providers.back': 'Back',
     'set.providerDetail.title': '{name}',
     'set.providerDetail.key': 'Set API Key',
+    'set.providerDetail.clearKey': 'Clear API Key',
+    'set.providerDetail.endpoint': 'Endpoint',
     'set.providerDetail.models': 'Manage Models',
     'set.providerDetail.pricing': 'Pricing',
     'set.providerDetail.setDefault': 'Set as Default',
@@ -396,6 +407,10 @@ const fr = {
     'wiz.provider.name.ph': 'ex : openrouter, mistral, ollama…',
     'wiz.provider.url.title': 'URL de base (endpoint OpenAI-compatible)',
     'wiz.provider.url.ph': 'ex : https://openrouter.ai/api/v1',
+    'wiz.provider.endpoint.title': 'Vérifier l’endpoint de {name}',
+    'wiz.provider.endpoint.model': 'Modèle : {model}',
+    'wiz.provider.endpoint.use': 'Utiliser cet endpoint',
+    'wiz.provider.endpoint.edit': 'Modifier l’endpoint',
     'wiz.provider.model.title': 'Nom du modèle — exactement comme dans la documentation du provider',
     'wiz.provider.model.ph': 'ex : deepseek-chat, gpt-4o-mini, qwen2.5-coder…',
     'wiz.provider.key.title': 'Clef API pour {name}',
@@ -454,7 +469,7 @@ const fr = {
     'help.l2d': ' pour tous).',
     'help.l3': 'Les agents voient mutuellement leurs statuts et leurs diffs avant chaque action : ils co-éditent les mêmes fichiers sans se bloquer et sans casser le travail des autres.',
     'help.states': "États : 🔨 travaille · 👂 écoute les autres agents (double bordure) · 🧠 réfléchit · ✋ attend ton approbation · ⏹ stop. Sons : 1 bip = agent lancé/terminé, 2 bips = approbation requise (/sound off pour couper).",
-    'help.keys': 'Esc : revenir à la vue agents · Tab : complétion · ↑/↓ : historique · Ctrl+V : coller une image',
+    'help.keys': 'Esc : revenir à la vue agents · Tab : complétion · ↑/↓ : suggestions, historique ou scroll selon la vue · Ctrl+V : coller une image',
     'cmd.ask': 'Agent ask : répond et conseille sans modifier',
     'cmd.task': 'Agent task : exécute, modifie, valide, résume',
     'cmd.agents': 'Vue panneaux agents (temps réel)',
@@ -491,6 +506,8 @@ const fr = {
     'cmd.doctor': 'Vérifier provider/modèle/clef API et afficher les corrections',
     'cmd.settings': 'Réglages globaux — providers, clefs API, langue, défauts (persistés)',
     'cmd.ssettings': 'Réglages de session — modèle, approbations, son (cette session)',
+    'cmd.project': 'Changer de dossier projet ou rouvrir le choix de dossier',
+    'cmd.wizard': 'Relancer le wizard de configuration',
     'cmd.approvals': 'Approbations shell : ask, auto-safe ou yolo (cette session)',
     'cmd.sound': 'Repères sonores (cette session)',
     'cmd.save': 'Sauvegarder la session maintenant (avec un nom optionnel)',
@@ -587,7 +604,7 @@ const fr = {
     'm.usageSession': 'Usage : /sessions pour lister, puis /session <n> ou /session latest',
     'm.sessionLoaded': '📂 Session du {date} chargée.',
     'm.missingProvider': 'Aucun provider configuré. Ouvre /settings → Ajouter un provider, ou relance une première config.',
-    'm.missingKey': 'Le provider {name} n’a pas de clef API. Utilise /key <clef API> ou /settings → Clef API.',
+    'm.missingKey': 'Le provider {name} n’a pas de clef API. Utilise /settings → Providers → Clef API.',
     'm.missingModel': 'Le provider {name} n’a pas de modèle par défaut. Utilise /settings → Modèles du provider.',
     'm.doctorOk': '✓ Configuration prête : {pm}. Changements : /settings · modèle temporaire : /model [provider:]modèle',
     'm.spawnFail': "Impossible de lancer l'agent : aucun provider/modèle utilisable. Configure-en un via /settings.",
@@ -658,9 +675,12 @@ const fr = {
     'set.newSkillName': 'Nom du skill (un modèle .md sera créé dans ~/.parallel/skills)',
     'set.newSpecialistName': 'Nom du spécialiste (un modèle .md sera créé dans ~/.parallel/specialists)',
     // provider-pick section headers
-    'wiz.provider.section.configured': 'Configurés',
-    'wiz.provider.section.cloud': 'Fournisseurs cloud',
-    'wiz.provider.section.local': 'Local',
+    'wiz.provider.section.configured': 'Providers configurés',
+    'wiz.provider.section.western': 'Occidental',
+    'wiz.provider.section.chinese': 'Chinois',
+    'wiz.provider.section.gateways': 'Passerelles',
+    'wiz.provider.section.inference': 'Inférence',
+    'wiz.provider.section.local': 'Modèles locaux',
     'wiz.provider.customDetail': 'Configurer manuellement',
     'wiz.provider.ollamaDetail': 'Local — pas de clef API requise',
     'wiz.provider.ollama.checking': 'Vérification de Ollama sur localhost:11434…',
@@ -675,6 +695,8 @@ const fr = {
     'set.providers.back': 'Retour',
     'set.providerDetail.title': '{name}',
     'set.providerDetail.key': 'Définir la clef API',
+    'set.providerDetail.clearKey': 'Effacer la clef API',
+    'set.providerDetail.endpoint': 'Endpoint',
     'set.providerDetail.models': 'Gérer les modèles',
     'set.providerDetail.pricing': 'Tarifs',
     'set.providerDetail.setDefault': 'Définir par défaut',
@@ -724,6 +746,10 @@ const es = {
     'wiz.provider.name.ph': 'ej.: openrouter, mistral, ollama…',
     'wiz.provider.url.title': 'URL base (endpoint compatible con OpenAI)',
     'wiz.provider.url.ph': 'ej.: https://openrouter.ai/api/v1',
+    'wiz.provider.endpoint.title': 'Revisar endpoint de {name}',
+    'wiz.provider.endpoint.model': 'Modelo: {model}',
+    'wiz.provider.endpoint.use': 'Usar este endpoint',
+    'wiz.provider.endpoint.edit': 'Editar endpoint',
     'wiz.provider.model.title': 'Nombre del modelo — exactamente como en la documentación del proveedor',
     'wiz.provider.model.ph': 'ej.: deepseek-chat, gpt-4o-mini, qwen2.5-coder…',
     'wiz.provider.key.title': 'Clave API para {name}',
@@ -782,7 +808,7 @@ const es = {
     'help.l2d': ' para todos).',
     'help.l3': 'Los agentes ven los estados y diffs de los demás antes de cada acción: co-editan los mismos archivos sin bloquearse y sin romper el trabajo ajeno.',
     'help.states': 'Estados: 🔨 trabajando · 👂 escuchando a otros agentes (borde doble) · 🧠 pensando · ✋ esperando tu aprobación · ⏹ stop. Sonidos: 1 bip = agente lanzado/terminado, 2 bips = aprobación requerida (/sound off para silenciar).',
-    'help.keys': 'Esc: volver a la vista de agentes · Tab: autocompletar · ↑/↓: historial · Ctrl+V: pegar imagen',
+    'help.keys': 'Esc: volver a la vista de agentes · Tab: autocompletar · ↑/↓: sugerencias, historial o scroll de vista · Ctrl+V: pegar imagen',
     'cmd.ask': 'Agente ask: responde y aconseja sin editar',
     'cmd.task': 'Agente task: ejecuta, edita, valida y resume',
     'cmd.agents': 'Vista de paneles de agentes (tiempo real)',
@@ -819,6 +845,8 @@ const es = {
     'cmd.doctor': 'Comprobar proveedor/modelo/clave API y mostrar correcciones',
     'cmd.settings': 'Ajustes globales — proveedores, claves API, idioma, valores por defecto (persistentes)',
     'cmd.ssettings': 'Ajustes de sesión — modelo, aprobaciones, sonido (solo esta sesión)',
+    'cmd.project': 'Cambiar carpeta del proyecto o reabrir selector de carpeta',
+    'cmd.wizard': 'Relanzar el asistente de configuración',
     'cmd.approvals': 'Aprobaciones shell: ask, auto-safe o yolo (esta sesión)',
     'cmd.sound': 'Señales sonoras (esta sesión)',
     'cmd.save': 'Guardar la sesión ahora (con un nombre opcional)',
@@ -915,7 +943,7 @@ const es = {
     'm.usageSession': 'Uso: /sessions para listar, luego /session <n> o /session latest',
     'm.sessionLoaded': '📂 Sesión del {date} cargada.',
     'm.missingProvider': 'No hay proveedor configurado. Abre /settings → Añadir proveedor, o reinicia con configuración inicial.',
-    'm.missingKey': 'El proveedor {name} no tiene clave API. Usa /key <clave API> o /settings → Clave API.',
+    'm.missingKey': 'El proveedor {name} no tiene clave API. Usa /settings → Providers → Clave API.',
     'm.missingModel': 'El proveedor {name} no tiene modelo por defecto. Usa /settings → Modelos del proveedor.',
     'm.doctorOk': '✓ Configuración lista: {pm}. Cambios: /settings · modelo temporal: /model [proveedor:]modelo',
     'm.spawnFail': 'No se puede lanzar el agente: ningún provider/modelo utilizable. Configura uno con /settings.',
@@ -986,9 +1014,12 @@ const es = {
     'set.newSkillName': 'Nombre del skill (se creará una plantilla .md en ~/.parallel/skills)',
     'set.newSpecialistName': 'Nombre del especialista (se creará una plantilla .md en ~/.parallel/specialists)',
     // provider-pick section headers
-    'wiz.provider.section.configured': 'Configurados',
-    'wiz.provider.section.cloud': 'Proveedores en la nube',
-    'wiz.provider.section.local': 'Local',
+    'wiz.provider.section.configured': 'Proveedores configurados',
+    'wiz.provider.section.western': 'Occidental',
+    'wiz.provider.section.chinese': 'Chino',
+    'wiz.provider.section.gateways': 'Pasarelas',
+    'wiz.provider.section.inference': 'Inferencia',
+    'wiz.provider.section.local': 'Modelos locales',
     'wiz.provider.customDetail': 'Configurar manualmente',
     'wiz.provider.ollamaDetail': 'Local — no requiere clave API',
     'wiz.provider.ollama.checking': 'Comprobando Ollama en localhost:11434…',
@@ -1003,6 +1034,8 @@ const es = {
     'set.providers.back': 'Volver',
     'set.providerDetail.title': '{name}',
     'set.providerDetail.key': 'Establecer clave API',
+    'set.providerDetail.clearKey': 'Borrar clave API',
+    'set.providerDetail.endpoint': 'Endpoint',
     'set.providerDetail.models': 'Gestionar modelos',
     'set.providerDetail.pricing': 'Precios',
     'set.providerDetail.setDefault': 'Establecer como predeterminado',
@@ -1052,6 +1085,10 @@ const zh = {
     'wiz.provider.name.ph': '例如 openrouter、mistral、ollama…',
     'wiz.provider.url.title': '基础 URL（OpenAI 兼容端点）',
     'wiz.provider.url.ph': '例如 https://openrouter.ai/api/v1',
+    'wiz.provider.endpoint.title': '检查 {name} 端点',
+    'wiz.provider.endpoint.model': '模型：{model}',
+    'wiz.provider.endpoint.use': '使用此端点',
+    'wiz.provider.endpoint.edit': '编辑端点',
     'wiz.provider.model.title': '模型名称 — 与提供商文档中的写法完全一致',
     'wiz.provider.model.ph': '例如 deepseek-chat、gpt-4o-mini、qwen2.5-coder…',
     'wiz.provider.key.title': '{name} 的 API 密钥',
@@ -1110,7 +1147,7 @@ const zh = {
     'help.l2d': ' 发给所有）。',
     'help.l3': '每次行动前，智能体都能看到彼此的状态和差异：它们共同编辑同一批文件，互不阻塞，也不破坏彼此的工作。',
     'help.states': '状态：🔨 工作中 · 👂 倾听其他智能体（双边框）· 🧠 思考中 · ✋ 等待你的批准 · ⏹ 停止。声音：1 声 = 智能体启动/完成，2 声 = 需要批准（/sound off 静音）。',
-    'help.keys': 'Esc：返回智能体视图 · Tab：补全 · ↑/↓：历史 · Ctrl+V：粘贴图片',
+    'help.keys': 'Esc：返回智能体视图 · Tab：补全 · ↑/↓：建议、历史或视图滚动 · Ctrl+V：粘贴图片',
     'cmd.ask': 'Ask 智能体：只回答和建议，不编辑',
     'cmd.task': 'Task 智能体：执行、编辑、验证并总结',
     'cmd.agents': '智能体面板视图（实时）',
@@ -1147,6 +1184,8 @@ const zh = {
     'cmd.doctor': '检查提供商/模型/API 密钥配置并显示修复提示',
     'cmd.settings': '全局设置 — 提供商、API 密钥、语言、默认值（持久化）',
     'cmd.ssettings': '会话设置 — 模型、批准、声音（仅本会话）',
+    'cmd.project': '更改项目文件夹或重新打开文件夹选择器',
+    'cmd.wizard': '重新启动设置向导',
     'cmd.approvals': 'Shell 批准：ask、auto-safe 或 yolo（本会话）',
     'cmd.sound': '声音提示（本会话）',
     'cmd.save': '立即保存会话（可附加名称）',
@@ -1243,7 +1282,7 @@ const zh = {
     'm.usageSession': '用法：先用 /sessions 列出，然后 /session <n> 或 /session latest',
     'm.sessionLoaded': '📂 已加载 {date} 的会话。',
     'm.missingProvider': '未配置提供商。打开 /settings → 添加提供商，或重新进行首次配置。',
-    'm.missingKey': '提供商 {name} 没有 API 密钥。使用 /key <API 密钥> 或 /settings → API 密钥。',
+    'm.missingKey': '提供商 {name} 没有 API 密钥。使用 /settings → Providers → API 密钥。',
     'm.missingModel': '提供商 {name} 没有默认模型。使用 /settings → 提供商模型。',
     'm.doctorOk': '✓ 配置就绪：{pm}。修改：/settings · 临时模型：/model [提供商:]模型',
     'm.spawnFail': '无法启动智能体：没有可用的 provider/模型。请通过 /settings 配置。',
@@ -1314,9 +1353,12 @@ const zh = {
     'set.newSkillName': '技能名称（将在 ~/.parallel/skills 创建 .md 模板）',
     'set.newSpecialistName': '专家名称（将在 ~/.parallel/specialists 创建 .md 模板）',
     // provider-pick section headers
-    'wiz.provider.section.configured': '已配置',
-    'wiz.provider.section.cloud': '云提供商',
-    'wiz.provider.section.local': '本地',
+    'wiz.provider.section.configured': '已配置的提供商',
+    'wiz.provider.section.western': '西方',
+    'wiz.provider.section.chinese': '中文',
+    'wiz.provider.section.gateways': '网关',
+    'wiz.provider.section.inference': '推理',
+    'wiz.provider.section.local': '本地模型',
     'wiz.provider.customDetail': '手动配置',
     'wiz.provider.ollamaDetail': '本地 — 无需 API 密钥',
     'wiz.provider.ollama.checking': '正在检查 Ollama（localhost:11434）…',
@@ -1331,6 +1373,8 @@ const zh = {
     'set.providers.back': '返回',
     'set.providerDetail.title': '{name}',
     'set.providerDetail.key': '设置 API 密钥',
+    'set.providerDetail.clearKey': '清除 API 密钥',
+    'set.providerDetail.endpoint': '端点',
     'set.providerDetail.models': '管理模型',
     'set.providerDetail.pricing': '定价',
     'set.providerDetail.setDefault': '设为默认',

package/dist/index.js

@@ -6,7 +6,7 @@ import path from 'node:path';
 import { render } from 'ink';
 import { App } from './ui/App.js';
 import { Controller } from './controller.js';
-import { loadConfig, setConfigHome } from './config.js';
+import { loadConfig, providerReady, setConfigHome } from './config.js';
 import { setLang } from './i18n.js';
 const argv = process.argv.slice(2);
 function takeFlagValue(flag) {
@@ -47,10 +47,13 @@ Environment variables:
   PARALLEL_API_KEY / DEEPSEEK_API_KEY   API key
   PARALLEL_MODEL                        Default model (e.g. deepseek-chat)
   PARALLEL_BASE_URL                     OpenAI-compatible endpoint
+  PARALLEL_NO_ALT_SCREEN=1              Disable the alternate terminal screen.
 
 Inside the TUI:
   <task> + Enter         Launch agent N+1 — even while the others are working
   @a1 <message>          Real-time instruction to an agent (@all for everyone)
+  /project [folder]      Change project folder or reopen the folder picker
+  /wizard                Relaunch the setup wizard
   /help                  All commands
 `);
     process.exit(0);
@@ -104,7 +107,7 @@ if (headless) {
     // No human in the loop: commands are auto-approved.
     ctl.setSessionApprovalMode('yolo');
     const provider = ctl.sessionProvider();
-    if (!provider || !provider.apiKey) {
+    if (!provider || !providerReady(provider)) {
         console.error('Headless mode needs a configured provider + API key. Run `parallel` interactively once, or set PARALLEL_API_KEY.');
         process.exit(1);
     }

package/dist/pricing.js

@@ -5,60 +5,162 @@
  * Prices drift over time; overrides always win.
  */
 const BUILTIN = {
-    // DeepSeek
-    'deepseek-v4-flash': { input: 0.27, output: 1.1 },
-    'deepseek-v4-pro': { input: 0.55, output: 2.19 },
-    'deepseek-chat': { input: 0.27, output: 1.1 },
-    'deepseek-reasoner': { input: 0.55, output: 2.19 },
+    // === Western ===
     // OpenAI
-    'gpt-4o': { input: 2.5, output: 10 },
-    'gpt-4o-mini': { input: 0.15, output: 0.6 },
-    'gpt-4.1': { input: 2, output: 8 },
-    'gpt-4.1-mini': { input: 0.4, output: 1.6 },
-    'gpt-4.1-nano': { input: 0.1, output: 0.4 },
-    'o3-mini': { input: 1.1, output: 4.4 },
-    'o4-mini': { input: 1.1, output: 4.4 },
+    'gpt-5.5': { input: 5.00, output: 30.00, cacheHit: 0.50 },
+    'gpt-5.5-pro': { input: 30.00, output: 180.00, cacheHit: 3.00 },
+    'gpt-5.4': { input: 1.25, output: 10.00, cacheHit: 0.125 }, // approx
+    'gpt-5.3-codex': { input: 1.25, output: 10.00, cacheHit: 0.125 }, // approx
+    'gpt-4o': { input: 2.50, output: 10.00 },
+    'gpt-4o-mini': { input: 0.15, output: 0.60 },
+    'o4-mini': { input: 1.10, output: 4.40 },
+    'o3': { input: 5.00, output: 20.00 },
+    'o3-mini': { input: 0.55, output: 2.20 },
+    'o1': { input: 15.00, output: 60.00 },
+    'o1-mini': { input: 1.10, output: 4.40 },
     // Anthropic
-    'claude-opus-4': { input: 15, output: 75 },
-    'claude-sonnet-4': { input: 3, output: 15 },
-    'claude-haiku-4': { input: 1, output: 5 },
-    'claude-3-5-haiku': { input: 0.8, output: 4 },
+    'claude-opus-4-8': { input: 5.00, output: 25.00, cacheHit: 0.50 },
+    'claude-opus-4-7': { input: 5.00, output: 25.00, cacheHit: 0.50 },
+    'claude-sonnet-4-6': { input: 3.00, output: 15.00, cacheHit: 0.30 },
+    'claude-haiku-4-5': { input: 1.00, output: 5.00, cacheHit: 0.10 },
+    // Google Gemini
+    'gemini-3.1-pro': { input: 2.00, output: 12.00, cacheHit: 0.20 }, // tiered: >200K = 2×/1.5×
+    'gemini-3.5-flash': { input: 1.50, output: 9.00, cacheHit: 0.15 },
+    'gemini-3-flash': { input: 0.60, output: 3.00, cacheHit: 0.06 }, // approx
+    'gemini-3.1-flash-lite': { input: 0.10, output: 0.40, cacheHit: 0.01 },
+    // xAI Grok
+    'grok-4': { input: 3.00, output: 15.00 }, // approx
+    'grok-4-fast-reasoning': { input: 0.20, output: 0.50 }, // approx
+    'grok-3': { input: 3.00, output: 15.00 },
+    'grok-code-fast-1': { input: 0.20, output: 1.50 }, // approx
     // Mistral
-    'mistral-large': { input: 2, output: 6 },
-    'codestral': { input: 0.3, output: 0.9 },
-    'devstral': { input: 0.1, output: 0.3 },
-    // Alibaba
-    'qwen2.5-coder': { input: 0.09, output: 0.09 },
-    'qwen-max': { input: 1.6, output: 6.4 },
-    // xAI (Grok)
-    'grok-4': { input: 4.0, output: 16.0 },
-    'grok-3-beta': { input: 3.0, output: 12.0 },
-    'grok-3-mini': { input: 0.55, output: 2.2 },
-    // Perplexity
-    'sonar-pro': { input: 3.0, output: 15.0 },
-    'sonar': { input: 1.0, output: 1.0 },
-    'sonar-reasoning': { input: 2.0, output: 16.0 },
+    'mistral-large-2': { input: 2.00, output: 6.00 },
+    'magistral-medium': { input: 2.00, output: 5.00 },
+    'codestral-latest': { input: 0.30, output: 0.90 },
+    'mistral-small-latest': { input: 0.20, output: 0.60 },
     // Cohere
-    'command-a': { input: 2.5, output: 10.0 },
-    'command-r-plus': { input: 2.5, output: 10.0 },
-    'command-r': { input: 0.5, output: 1.5 },
-    // DeepInfra
-    'llama-4-maverick': { input: 0.2, output: 0.6 }, // approximate
-    'wizardlm-2-8x22b': { input: 0.5, output: 0.5 }, // approximate
-    // Fireworks
-    'llama-4-scout': { input: 0.1, output: 0.3 }, // approximate
-    'mixtral-8x22b': { input: 0.9, output: 0.9 }, // approximate
+    'command-a': { input: 2.50, output: 10.00 },
+    'command-r-plus': { input: 2.50, output: 10.00 },
+    // Perplexity
+    'sonar-pro': { input: 3.00, output: 15.00 },
+    'sonar-deep-research': { input: 2.00, output: 8.00 },
+    // === Chinese ===
+    // DeepSeek
+    'deepseek-v4-pro': { input: 0.435, output: 0.87, cacheHit: 0.0036 },
+    'deepseek-v4-flash': { input: 0.14, output: 0.28, cacheHit: 0.003 },
+    'deepseek-chat': { input: 0.27, output: 1.10, cacheHit: 0.027 },
+    'deepseek-reasoner': { input: 0.55, output: 2.19, cacheHit: 0.14 },
+    // MiniMax — list pricing
+    'MiniMax-M3': { input: 0.60, output: 2.40, cacheHit: 0.12 },
+    'MiniMax-M2.7': { input: 0.30, output: 1.20, cacheHit: 0.06 },
+    'MiniMax-M2.7-highspeed': { input: 0.60, output: 2.40, cacheHit: 0.06 },
+    // Z.ai / GLM
+    'glm-5.2': { input: 1.00, output: 3.20 },
+    'glm-5.1': { input: 1.00, output: 3.20 },
+    'glm-4.7': { input: 0.60, output: 2.20 },
+    'glm-4.7-flash': { input: 0.00, output: 0.00 }, // free
+    'glm-5v-turbo': { input: 0.30, output: 0.90 },
+    // Alibaba / Qwen
+    'qwen3.7-max': { input: 1.25, output: 3.75 }, // promo price, cache: yes
+    'qwen3.6-max-preview': { input: 1.04, output: 6.24 }, // tiered, lower bound
+    'qwen3.6-plus': { input: 0.40, output: 1.20 },
+    'qwen3.5-coder': { input: 0.20, output: 0.60 },
+    // Moonshot / Kimi
+    'kimi-k2.6': { input: 0.95, output: 4.00, cacheHit: 0.16 },
+    'kimi-k2.7-code': { input: 0.95, output: 4.00, cacheHit: 0.19 },
+    'kimi-k2.5': { input: 0.60, output: 3.00, cacheHit: 0.10 },
+    'moonshot-v1-128k': { input: 0.85, output: 1.70 },
+    // Xiaomi / MiMo
+    'mimo-v2-pro': { input: 0.30, output: 1.00 }, // approx
+    'mimo-v2-omni': { input: 0.40, output: 1.50 }, // approx
+    // StepFun
+    'step-2-16k': { input: 0.50, output: 1.50 }, // approx
+    // === Gateways ===
+    // OpenRouter
+    'openai/gpt-5.5': { input: 5.00, output: 30.00 },
+    'anthropic/claude-sonnet-4-6': { input: 3.00, output: 15.00 },
+    'google/gemini-3.5-flash': { input: 1.50, output: 9.00 },
+    'deepseek/deepseek-v4-pro': { input: 0.435, output: 0.87 },
+    'meta-llama/llama-4-maverick': { input: 0.50, output: 1.60 },
+    'mistralai/mistral-large-2': { input: 2.00, output: 6.00 },
+    // SiliconFlow
+    'deepseek-ai/DeepSeek-V4-Pro': { input: 0.435, output: 0.87 },
+    'deepseek-ai/DeepSeek-R1': { input: 0.55, output: 2.19 },
+    'Qwen/Qwen3-Coder-480B': { input: 0.20, output: 0.60 },
+    'glm-4/GLM-5.2': { input: 1.00, output: 3.20 },
+    'moonshotai/Kimi-K2.6': { input: 0.95, output: 4.00 },
+    // Atlas Cloud
+    'deepseek-v4-pro @atlas': { input: 0.435, output: 0.87 },
+    'deepseek-r1 @atlas': { input: 0.55, output: 2.19 },
+    'qwen3.7-max @atlas': { input: 1.25, output: 3.75 },
+    'glm-5.2 @atlas': { input: 1.00, output: 3.20 },
+    'kimi-k2.6 @atlas': { input: 0.95, output: 4.00 },
+    'llama-4-maverick @atlas': { input: 0.50, output: 1.60 },
+    // Requesty — 0% markup, same as direct
+    'gpt-5.5 @requesty': { input: 5.00, output: 30.00 },
+    'claude-sonnet-4-6 @requesty': { input: 3.00, output: 15.00 },
+    'gemini-3.5-flash @requesty': { input: 1.50, output: 9.00 },
+    'deepseek-v4-pro @requesty': { input: 0.435, output: 0.87 },
+    'llama-4-maverick @requesty': { input: 0.50, output: 1.60 },
+    'mistral-large-2 @requesty': { input: 2.00, output: 6.00 },
+    // Vercel AI Gateway — list price, no markup
+    'gpt-5.5 @vercel': { input: 5.00, output: 30.00 },
+    'claude-sonnet-4-6 @vercel': { input: 3.00, output: 15.00 },
+    'gemini-3.5-flash @vercel': { input: 1.50, output: 9.00 },
+    'deepseek-v4-pro @vercel': { input: 0.435, output: 0.87 },
+    'llama-4-maverick @vercel': { input: 0.50, output: 1.60 },
+    // === Inference hosts ===
+    // Groq
+    'qwen-2.5-coder-32b': { input: 0.30, output: 0.50 },
+    'deepseek-r1-distill-llama-70b': { input: 0.30, output: 2.00 },
+    'kimi-k2.6 @groq': { input: 1.00, output: 3.00 },
+    'llama-3.3-70b-versatile': { input: 0.15, output: 0.30 },
     // Cerebras
-    'llama-3.3-70b @cerebras': { input: 0.5, output: 1.5 }, // approximate
-    'llama-3.1-8b': { input: 0.05, output: 0.1 }, // approximate
+    'llama-4-maverick-17b-128e-instruct': { input: 0.65, output: 0.85 },
+    'qwen3-coder-480b @cerebras': { input: 2.00, output: 2.00 },
+    'kimi-k2.6 @cerebras': { input: 1.65, output: 6.50 },
+    'llama-3.3-70b @cerebras': { input: 0.10, output: 0.20 },
+    // Together AI
+    'meta-llama/Llama-4-Maverick-17B-128E-Instruct': { input: 0.27, output: 0.85 },
+    'deepseek-ai/DeepSeek-V3 @together': { input: 0.30, output: 0.30 },
+    'Qwen/Qwen3-Coder-480B @together': { input: 0.40, output: 1.20 },
+    'moonshotai/Kimi-K2.6 @together': { input: 0.60, output: 2.50 },
+    // Fireworks
+    'accounts/fireworks/models/llama4-maverick-17b': { input: 0.22, output: 0.88 },
+    'accounts/fireworks/models/deepseek-v3': { input: 0.90, output: 0.90 },
+    'accounts/fireworks/models/qwen3-coder-480b': { input: 0.45, output: 1.80 },
+    'accounts/fireworks/models/kimi-k2.6': { input: 0.60, output: 2.50 },
+    // DeepInfra
+    'meta-llama/Llama-4-Maverick-17B-128E': { input: 0.20, output: 0.60 },
+    'deepseek-ai/DeepSeek-V3 @deepinfra': { input: 0.26, output: 0.38 },
+    'Qwen/Qwen3-Coder-480B @deepinfra': { input: 0.30, output: 1.00 },
+    'moonshotai/Kimi-K2.6 @deepinfra': { input: 0.75, output: 3.50, cacheHit: 0.15 },
     // Novita
-    'deepseek-r1': { input: 2.0, output: 8.0 },
-    'deepseek-v3': { input: 1.25, output: 5.0 },
-    'llama-3.1-70b': { input: 0.35, output: 0.4 }, // approximate
+    'meta-llama/llama-4-maverick-17b-128e': { input: 0.20, output: 0.60 },
+    'deepseek/deepseek-v3': { input: 0.10, output: 0.28 },
+    'qwen/qwen3-coder-480b': { input: 0.30, output: 1.20 },
+    'moonshotai/kimi-k2.6': { input: 0.57, output: 2.30 },
     // Hyperbolic
-    'qwen3-235b': { input: 0.5, output: 1.5 }, // approximate
-    // Local endpoints are free
-    'ollama': { input: 0, output: 0 },
+    'meta-llama/Llama-4-Maverick-17B-128E @hyperbolic': { input: 0.20, output: 0.60 },
+    'deepseek-ai/DeepSeek-V3 @hyperbolic': { input: 0.25, output: 0.85 },
+    'Qwen/Qwen3-Coder-480B @hyperbolic': { input: 0.30, output: 1.20 },
+    'moonshotai/Kimi-K2.6 @hyperbolic': { input: 0.80, output: 3.00 },
+    // SambaNova
+    'Meta-Llama-4-Maverick-17B-128E-Instruct': { input: 0.20, output: 0.30 },
+    'DeepSeek-V3 @sambanova': { input: 1.00, output: 1.50 },
+    'Llama-3.3-70B-Instruct': { input: 0.10, output: 0.20 },
+    // === Local (free) ===
+    // Ollama
+    'qwen3-coder:480b': { input: 0.00, output: 0.00 },
+    'glm-4.7 @ollama': { input: 0.00, output: 0.00 },
+    'deepseek-v3 @ollama': { input: 0.00, output: 0.00 },
+    'kimi-k2 @ollama': { input: 0.00, output: 0.00 },
+    'llama3.2': { input: 0.00, output: 0.00 },
+    'mistral @ollama': { input: 0.00, output: 0.00 },
+    'codellama': { input: 0.00, output: 0.00 },
+    'gemma3': { input: 0.00, output: 0.00 },
+    // vLLM / SGLang
+    'your-model-here': { input: 0.00, output: 0.00 },
 };
 /**
  * Resolve the price of a model: provider override first, then built-in table
@@ -66,22 +168,28 @@ const BUILTIN = {
  * "claude-sonnet-4-20250514" or "openai/gpt-4o-mini"). null = unknown.
  */
 export function priceFor(provider, model) {
-    const override = provider?.prices?.[model];
+    const overrideKey = provider?.prices
+        ? Object.keys(provider.prices).find((key) => key.toLowerCase() === model.toLowerCase())
+        : undefined;
+    const override = overrideKey ? provider?.prices?.[overrideKey] : undefined;
     if (override)
         return override;
     const m = model.toLowerCase();
     // strip an optional "vendor/" prefix (OpenRouter-style ids)
     const bare = m.includes('/') ? m.slice(m.lastIndexOf('/') + 1) : m;
-    if (BUILTIN[bare])
-        return BUILTIN[bare];
+    const builtins = Object.entries(BUILTIN).map(([key, price]) => [key.toLowerCase(), price]);
+    const exact = builtins.find(([key]) => key === bare);
+    if (exact)
+        return exact[1];
     // longest prefix wins so "deepseek-chat" beats nothing else
     let best = null;
-    for (const key of Object.keys(BUILTIN)) {
-        if ((bare.startsWith(key) || bare.includes(key)) && (!best || key.length > best.length))
-            best = key;
+    for (const entry of builtins) {
+        const [key] = entry;
+        if ((bare.startsWith(key) || bare.includes(key)) && (!best || key.length > best[0].length))
+            best = entry;
     }
     if (best)
-        return BUILTIN[best];
+        return best[1];
     // local endpoints (ollama, llama.cpp, vLLM on localhost) → free
     if (provider && /localhost|127\.0\.0\.1|0\.0\.0\.0/.test(provider.baseUrl))
         return { input: 0, output: 0 };