nexrall-code 0.5.98 → 0.5.99

package/dist/index.js

@@ -103755,6 +103755,8 @@ ${ctx.repeatError}
             return `Browser: type "${String(input.text ?? "").slice(0, 40)}"`;
           if (action === "snapshot")
             return "Browser: read page";
+          if (action === "read_text")
+            return "Browser: read page text";
           return `Browser: ${action || "(unknown action)"}`;
         }
         default:
@@ -104167,7 +104169,15 @@ ${partial}` : "",
       "deepseek-v4-pro": 1048576,
       "deepseek-v4-flash": 1048576,
       // Qwen (DashScope), by real model id (documented max input, same caveat).
-      "qwen3.7-max": 991800
+      "qwen3.7-max": 991800,
+      // Z.ai (GLM), by real model id. contextWindow is documented (Z.ai/
+      // Cloudflare Workers AI model cards, both list 1,048,576) rather than
+      // measured — a ~400K-token request was ACCEPTED (200), not rejected, so
+      // there was no 400 to read a real ceiling out of. maxOutputTokens IS
+      // measured: `max_tokens: 999999` was rejected with the ceiling in the
+      // error body (backend services/providers/modelRegistry.js's glm-5.3 row
+      // has the full verification notes).
+      "glm-5.3": 1048576
     };
     function contextWindowFor2(model) {
       return MODEL_CONTEXT_TOKENS[model ?? "turbo"] ?? 128e3;
@@ -153125,7 +153135,8 @@ var MODEL_LABELS = {
   "gpt-4.1": "GPT-4.1",
   "deepseek-v4-pro": "DeepSeek V4 Pro",
   "deepseek-v4-flash": "DeepSeek V4 Flash",
-  "qwen3.7-max": "Qwen3.7 Max"
+  "qwen3.7-max": "Qwen3.7 Max",
+  "glm-5.3": "GLM 5.3"
 };
 var LEGACY_MODEL_ALIASES = {
   turbo: "claude-sonnet-5",
@@ -153144,7 +153155,8 @@ var SELECTABLE_MODELS = [
   "gpt-4.1",
   "deepseek-v4-pro",
   "deepseek-v4-flash",
-  "qwen3.7-max"
+  "qwen3.7-max",
+  "glm-5.3"
 ];
 var MODEL_COST_MULTIPLIER = {
   "claude-sonnet-5": 1.5,
@@ -153155,13 +153167,18 @@ var MODEL_COST_MULTIPLIER = {
   "gpt-4.1": 1,
   "deepseek-v4-pro": 0.1,
   "deepseek-v4-flash": 0.05,
-  "qwen3.7-max": 1
+  "qwen3.7-max": 1,
+  // $1.40/$4.40 per 1M in/out (docs.z.ai, 2026-08-30) vs gpt-4.1's $2.00/$8.00
+  // (migration 113) — roughly 0.6x on a blended 75/25 in/out turn. Rounded to
+  // a friendly number like every other row; re-check backend/migrations/
+  // 130_glm_5_3_pricing.sql's own caveat if Z.ai ships a GLM-5.3-specific rate.
+  "glm-5.3": 0.6
 };
 function modelWithCostHint(id) {
   const mult = MODEL_COST_MULTIPLIER[id];
   return mult === void 0 ? id : `${id} (${mult}x)`;
 }
-var NO_VISION_MODELS = /* @__PURE__ */ new Set(["deepseek-v4-pro", "deepseek-v4-flash", "qwen3.7-max"]);
+var NO_VISION_MODELS = /* @__PURE__ */ new Set(["deepseek-v4-pro", "deepseek-v4-flash", "qwen3.7-max", "glm-5.3"]);
 var NO_PDF_MODELS = /* @__PURE__ */ new Set([
   "gpt-5.4",
   "gpt-5.4-mini",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "nexrall-code",
-  "version": "0.5.98",
+  "version": "0.5.99",
   "description": "Nexrall Code — AI coding assistant for your terminal (headless agent for scripts, CI and automation)",
   "keywords": [
     "ai",