cascade-ai 0.46.0 → 0.48.0

package/README.md

@@ -778,7 +778,7 @@ web/
 | ✓ | Session rollback button (desktop) + `/rollback` (CLI) |
 | ✓ | Cost-per-feature attribution (`costByFeature` in results, CLI cost panel, desktop chat) |
 | ✓ | Project world state (encrypted local log feeding T1 planning) |
-| ✓ | Cascade Cloud (hosted chat — GitHub/Google login, bring-your-own-key, `app.cascadeai.in`) |
+| ✓ | Cascade Cloud (hosted chat — GitHub/Google login, bring-your-own-key, `cascadeai.in`) |
 | 🔜 | VSCode extension (`cascade-vscode`) — see [docs/ROADMAP.md](docs/ROADMAP.md) |
 | 🔜 | JetBrains extension (`cascade-jetbrains`) — see [docs/ROADMAP.md](docs/ROADMAP.md) |
 | 🔜 | WASM/isolate sandboxing for tool execution — see [docs/ROADMAP.md](docs/ROADMAP.md) |

package/dist/cli.cjs

@@ -116,7 +116,7 @@ var __export = (target, all) => {
 var CASCADE_VERSION, CASCADE_CONFIG_FILE, CASCADE_DB_FILE, CASCADE_DASHBOARD_SECRET_FILE, GLOBAL_CONFIG_DIR, GLOBAL_DB_FILE, GLOBAL_KEYSTORE_FILE, GLOBAL_CREDENTIALS_FILE, GLOBAL_RUNTIME_DB_FILE, DEFAULT_DASHBOARD_PORT, DEFAULT_CONTEXT_LIMIT, DEFAULT_AUTO_SUMMARIZE_AT, MODELS, T1_MODEL_PRIORITY, T2_MODEL_PRIORITY, T3_MODEL_PRIORITY, VISION_MODEL_PRIORITY, COMPLEXITY_T2_COUNT, THEME_NAMES, DEFAULT_THEME, OLLAMA_BASE_URL, LM_STUDIO_BASE_URL, AZURE_BASE_URL_TEMPLATE, TOOL_NAMES, DEFAULT_APPROVAL_REQUIRED;
 var init_constants = __esm({
   "src/constants.ts"() {
-    CASCADE_VERSION = "0.46.0";
+    CASCADE_VERSION = "0.48.0";
     CASCADE_CONFIG_FILE = ".cascade/config.json";
     CASCADE_DB_FILE = ".cascade/memory.db";
     CASCADE_DASHBOARD_SECRET_FILE = ".cascade/dashboard-secret";
@@ -595,6 +595,60 @@ var init_base = __esm({
   }
 });
 
+// src/providers/model-filter.ts
+function isChatModel(id, methods) {
+  if (methods && methods.length > 0) {
+    const canGenerate = methods.some((m) => /generatecontent|generatemessage|chat|completion/i.test(m));
+    if (!canGenerate) return false;
+  }
+  return !NON_CHAT.test(id);
+}
+var NON_CHAT_PATTERNS, NON_CHAT;
+var init_model_filter = __esm({
+  "src/providers/model-filter.ts"() {
+    NON_CHAT_PATTERNS = [
+      "embed",
+      // text-embedding-3, nomic-embed, mxbai-embed, snowflake-arctic-embed
+      "(?:^|[-_/])tts(?:[-_/]|$)",
+      "text-to-speech",
+      "speech",
+      // text-to-speech
+      "whisper",
+      "transcrib",
+      "(?:^|[-_/])stt(?:[-_/]|$)",
+      // speech-to-text
+      "dall[-]?e",
+      "imagen",
+      "image-generation",
+      "gpt-image",
+      "(?:^|[-_/])image(?:[-_/]|$)",
+      "stable-diffusion",
+      "sdxl",
+      "flux",
+      // image gen
+      "moderation",
+      "(?:^|[-_/])guard(?:[-_/]|$)",
+      // moderation / safety
+      "rerank",
+      "(?:^|[-_/])audio(?:[-_/]|$)",
+      "native-audio",
+      "realtime",
+      // audio / realtime
+      "veo",
+      "sora",
+      "(?:^|[-_/])video(?:[-_/]|$)",
+      // video
+      "(?:^|[-_/])aqa(?:[-_/]|$)",
+      // attributed QA
+      "(?:^|[-_/])live(?:[-_/]|$)",
+      // live-audio variants
+      "(?:^|[-_/])(?:davinci|babbage)(?:[-_/]|$)"
+      // legacy completion-only base models
+    ];
+    NON_CHAT = new RegExp(NON_CHAT_PATTERNS.join("|"), "i");
+  }
+});
+
 // src/providers/anthropic.ts
 var anthropic_exports = {};
 __export(anthropic_exports, {
@@ -611,6 +665,7 @@ var init_anthropic = __esm({
   "src/providers/anthropic.ts"() {
     init_constants();
     init_base();
+    init_model_filter();
     AnthropicProvider = class extends BaseProvider {
       client;
       constructor(config, model) {
@@ -717,7 +772,7 @@ var init_anthropic = __esm({
           if (!Array.isArray(data?.data)) {
             return Object.values(MODELS).filter((m) => m.provider === "anthropic");
           }
-          return data.data.map((m) => {
+          return data.data.filter((m) => isChatModel(m.id)).map((m) => {
             const known = Object.values(MODELS).find((km) => km.id === m.id && km.provider === "anthropic");
             if (known) return known;
             return {
@@ -826,6 +881,7 @@ var init_openai = __esm({
   "src/providers/openai.ts"() {
     init_constants();
     init_base();
+    init_model_filter();
     OpenAIProvider = class extends BaseProvider {
       client;
       /** Once we learn (from the model id or an API error) that this deployment
@@ -947,7 +1003,7 @@ var init_openai = __esm({
       async listModels() {
         try {
           const response = await this.client.models.list();
-          return response.data.map((m) => {
+          return response.data.filter((m) => isChatModel(m.id)).map((m) => {
             const known = Object.values(MODELS).find((km) => km.id === m.id && km.provider === "openai");
             if (known) return known;
             return {
@@ -1155,6 +1211,7 @@ var init_gemini = __esm({
   "src/providers/gemini.ts"() {
     init_constants();
     init_base();
+    init_model_filter();
     GeminiProvider = class extends BaseProvider {
       client;
       constructor(config, model) {
@@ -1248,7 +1305,7 @@ var init_gemini = __esm({
           if (!Array.isArray(data?.models)) {
             return Object.values(MODELS).filter((m) => m.provider === "gemini");
           }
-          return data.models.map((m) => {
+          return data.models.filter((m) => isChatModel(m.name.replace("models/", ""), m.supportedGenerationMethods)).map((m) => {
             const id = m.name.replace("models/", "");
             const known = Object.values(MODELS).find(
               (km) => km.id === id && km.provider === "gemini"
@@ -1471,6 +1528,7 @@ var init_ollama = __esm({
     init_constants();
     init_net();
     init_base();
+    init_model_filter();
     TOOL_CAPABLE_FAMILIES = [
       "llama3.1",
       "llama3.2",
@@ -1623,10 +1681,10 @@ var init_ollama = __esm({
           const response = await fetch(`${this.baseUrl}/api/tags`);
           if (!response.ok) return [];
           const data = await response.json();
-          const supportedKeywords = ["llama3", "llama2", "gemma", "mistral", "mixtral", "qwen", "phi3", "codellama", "deepseek", "llava", "starcoder", "stable-code", "nomic-embed"];
+          const supportedKeywords = ["llama3", "llama2", "gemma", "mistral", "mixtral", "qwen", "phi3", "codellama", "deepseek", "llava", "starcoder", "stable-code"];
           const entries = data.models.filter((m) => {
             const name = m.name.toLowerCase();
-            return supportedKeywords.some((k) => name.includes(k));
+            return supportedKeywords.some((k) => name.includes(k)) && isChatModel(name);
           });
           const shows = await Promise.all(entries.map((m) => this.showModel(m.name)));
           return entries.map((m, i) => {
@@ -1726,6 +1784,7 @@ var OpenAICompatibleProvider;
 var init_openai_compatible = __esm({
   "src/providers/openai-compatible.ts"() {
     init_openai();
+    init_model_filter();
     init_net();
     OpenAICompatibleProvider = class extends OpenAIProvider {
       constructor(config, model) {
@@ -1759,7 +1818,8 @@ var init_openai_compatible = __esm({
           return typeof m === "string" ? m : void 0;
         }).filter((x) => typeof x === "string" && x.length > 0);
         if (ids.length === 0) return [this.model];
-        return ids.map((id) => ({
+        const chatIds = ids.filter((id) => isChatModel(id));
+        return (chatIds.length ? chatIds : ids).map((id) => ({
           id,
           name: id,
           provider: "openai-compatible",
@@ -14510,7 +14570,7 @@ function clearCloudSession(dir = globalDir()) {
 }
 
 // src/cloud/client.ts
-var DEFAULT_CLOUD_URL = "https://app.cascadeai.in";
+var DEFAULT_CLOUD_URL = "https://cascadeai.in";
 var REFRESH_MARGIN_MS = 6e4;
 var FileCloudSessionStore = class {
   constructor(dir) {