@tryarcanist/cli 0.1.324 → 0.1.326

package/dist/index.js

@@ -7370,7 +7370,7 @@ import { Command } from "commander";
 // package.json
 var package_default = {
   name: "@tryarcanist/cli",
-  version: "0.1.324",
+  version: "0.1.326",
   description: "CLI for Arcanist - create and manage coding agent sessions",
   type: "module",
   bin: {
@@ -23364,6 +23364,9 @@ var DeepSeekModel = {
   V4Flash: "deepseek-v4-flash",
   V4Pro: "deepseek-v4-pro"
 };
+var GlmModel = {
+  Flash53: "glm-5.3-flash"
+};
 var MODEL_PROVIDERS_SET = /* @__PURE__ */ new Set(["openai", "anthropic"]);
 var BACKEND_DESKTOP_IMAGE_FEEDBACK_CONFIGS = {
   [CODEX_AGENT_RUNTIME_BACKEND]: {
@@ -23379,6 +23382,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.4",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 1e6,
     // Default codex model after the gpt-5.5 downgrade; carries gpt-5.5's prior
@@ -23405,6 +23409,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.6",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 105e4,
     // Verified 2026-07-09 against OpenAI's GPT-5.6 migration guide and model
@@ -23433,6 +23438,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.6 Sol",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 105e4,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh", "max"], default: "medium" },
@@ -23458,6 +23464,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.6 Terra",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 105e4,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh", "max"], default: "medium" },
@@ -23485,6 +23492,7 @@ var MODEL_REGISTRY = [
     provider: "openai",
     overloadFallback: OpenAIModel.GPT56Terra,
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 105e4,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh", "max"], default: "medium" },
@@ -23511,6 +23519,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.5",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 1e6,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh"], default: "medium" },
@@ -23532,6 +23541,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.4 Mini",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 4e5,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh"], default: void 0 },
@@ -23552,6 +23562,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.4 Nano",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: false },
     contextWindow: 4e5,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh"], default: void 0 },
@@ -23563,6 +23574,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.3 Codex Spark",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 128e3,
     reasoning: { efforts: ["low", "medium", "high", "xhigh"], default: "high" },
@@ -23576,6 +23588,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.3 Codex",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 4e5,
     reasoning: { efforts: ["low", "medium", "high", "xhigh"], default: "high" },
@@ -23586,6 +23599,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.2",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 4e5,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh"], default: void 0 },
@@ -23596,6 +23610,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.2 Chat",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 128e3,
     reasoning: { efforts: ["none", "low", "medium", "high", "xhigh"], default: void 0 },
@@ -23606,6 +23621,7 @@ var MODEL_REGISTRY = [
     name: "GPT-5.2 Codex",
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "openai",
     capabilities: { codexToolSearch: true },
     contextWindow: 4e5,
     reasoning: { efforts: ["low", "medium", "high", "xhigh"], default: "high" },
@@ -23624,6 +23640,7 @@ var MODEL_REGISTRY = [
     // silently fails those closed instead of routing anywhere.
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "baseten",
     // Required for codex session-start eligibility: the codex backend is the
     // sole consumer of this flag and drops any model without it from the
     // session-start set (see modelSupportsRequiredSessionStartCapabilities).
@@ -23656,6 +23673,7 @@ var MODEL_REGISTRY = [
     // bridge's model guard.
     provider: "openai",
     backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "baseten",
     // Required for codex session-start eligibility (sole consumer of the flag).
     capabilities: { codexToolSearch: true },
     contextWindow: 262144,
@@ -23671,6 +23689,26 @@ var MODEL_REGISTRY = [
     visibility: "internal_probe"
     // Deliberately no overloadFallback — see DeepSeek V4 Flash: an A/B arm must
     // retry the same model on a Baseten 429, never silently swap models.
+  },
+  {
+    id: GlmModel.Flash53,
+    name: "GLM-5.3 Flash",
+    // Same contract as the DeepSeek V4 entries: "openai" is the wire protocol
+    // (OpenAI Responses), not the vendor. GLM-5.3-Flash is served by Baseten.
+    provider: "openai",
+    backends: [CODEX_AGENT_RUNTIME_BACKEND],
+    codexProviderRoute: "baseten",
+    capabilities: { codexToolSearch: true },
+    contextWindow: 1e6,
+    // Only the effort the live streamed Baseten probe proved on 2026-08-27.
+    reasoning: { efforts: ["medium"], default: "medium" },
+    // Verified against https://www.baseten.co/pricing on 2026-08-27 (Model APIs
+    // token pricing, row `zai-org/GLM-5.3-Flash`): $0.15/M input, $0.03/M cached
+    // input, $0.50/M output. Cached-input rate is published, so it is carried
+    // verbatim. No longContext and no flex (OpenAI-only axes).
+    pricing: { inputPerMillion: 0.15, outputPerMillion: 0.5, cacheReadPerMillion: 0.03 },
+    sessionStart: { eligible: true },
+    visibility: "internal_probe"
   }
 ];
 var MODEL_PROVIDER_NAMES = {
@@ -23763,7 +23801,6 @@ function extractModelId(raw) {
 function getProviderForModel(modelId) {
   return MODEL_PROVIDERS[modelId] ?? "openai";
 }
-var BASETEN_SERVED_MODEL_IDS = new Set(Object.values(DeepSeekModel));
 function toModelSelection(raw) {
   const modelID = extractModelId(raw);
   if (!modelID) return void 0;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tryarcanist/cli",
-  "version": "0.1.324",
+  "version": "0.1.326",
   "description": "CLI for Arcanist - create and manage coding agent sessions",
   "type": "module",
   "bin": {