@mars-sea/dsh-commandcode-provider 0.8.2 → 0.8.4

package/CHANGELOG.md

@@ -4,6 +4,27 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.8.4] - 2026-08-27
+
+### Added
+
+- **New model: Qwen 3.8 Flash (`Qwen/Qwen3.8-Flash`, command-code@1.36.0).** A Go-tier 1M-context Vision reasoning model with `low`/`medium`/`xhigh` effort levels, rounding out the Qwen 3.8 family alongside the existing Max and 27B variants.
+- **New model: GLM-5.3 Flash (`z-ai/glm-5.3-flash`, command-code@1.35.0).** A Go-tier 1M-context Vision reasoning model with `low`/`high`/`max` effort levels, added to the catalog as the official replacement when the Ox Alpha preview ended.
+
+### Removed
+
+- **The Ox Alpha preview is over (`stealth/ox-alpha`, command-code@1.34.0).** The model is gone from the catalog, the picker, and the FREE deal list; accounts that were using it should switch to GLM-5.3 Flash (same effort set, Vision-capable, free on every plan).
+
+### Changed
+
+- **Synced with the official command-code@1.36.0 CLI.** `COMMAND_CODE_CLI_VERSION` is now `1.36.0`. Plan tiers now read 39/43/56/61 (Go / cumulative-through-Goat / cumulative-through-Pro / cumulative-through-Provider): the two new Go models enter Go, and `stealth/ox-alpha` leaves every tier. Re-verified against the official sources with no other changes: wire protocol, endpoints, auth flow, `KNOWN_SUBSCRIPTION_PLANS` (the `individual-*` plan map and its monthly-credit totals are identical to 1.33.0), `KNOWN_PEAK_PRICING` (DeepSeek V4 Pro/Flash/Vision, 01–04 and 06–10 UTC), and every other deal on the pricing page.
+
+## [0.8.3] - 2026-08-26
+
+### Fixed
+
+- **Qwen 3.8 Max no longer shows the hourly-pricing (`Peak`/`Half`) marker.** The official pricing page's hover annotation for the three hourly-billed DeepSeek models had been misattributed to this neighboring row; Qwen 3.8 Max is not billed by the hour. Everything else is identical to 0.8.2.
+
 ## [0.8.2] - 2026-08-26
 
 ### Added
@@ -12,7 +33,6 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 
 ### Changed
 
-- **Qwen 3.8 Max now carries the hourly-pricing marker** (`Peak`/`Half` by current UTC hour) — it joined DeepSeek V4 Flash/Vision on time-of-day pricing per the official pricing page.
 - **The model picker's group header now reads "Command Code"** instead of the lowercase route id (`commandcode`), matching the name shown on the settings page.
 - **Synced with the official command-code@1.33.0 CLI** (upstream's only change was adding the two free models above). `COMMAND_CODE_CLI_VERSION` is now `1.33.0`. Re-verified against the official sources with no other changes: wire protocol, endpoints, auth flow, reasoning-effort maps, subscription plans, and peak windows are all identical; plan tiers gained only the two new Go-tier variants.
 

package/lib/client.js

@@ -1197,7 +1197,7 @@ window.__ModuleLoader__.load({
 		};
 		//#endregion
 		//#region package.json
-		var version = "0.8.2";
+		var version = "0.8.4";
 		var repository = {
 			"type": "git",
 			"url": "git+https://github.com/Mars-Sea/dsh-commandcode-provider.git"

package/lib/index.d.ts

@@ -24,7 +24,7 @@ declare const KNOWN_EFFORTS: Readonly<Record<string, readonly string[]>>;
  */
 declare const KNOWN_IMAGE_MODELS: ReadonlySet<string>;
 /**
- * Models the official CLI's model table (command-code@1.33.0) marks
+ * Models the official CLI's model table (command-code@1.36.0) marks
  * `reasoning:!0` but defines no selectable `reasoning_effort` levels — they
  * think automatically, with Command Code driving the depth. This is the
  * authoritative "thinks, effort not adjustable" set: `KNOWN_EFFORTS` (which
@@ -32,10 +32,11 @@ declare const KNOWN_IMAGE_MODELS: ReadonlySet<string>;
  * effort levels, and this snapshot is not surfaced in the picker's compact
  * description — it exists for programmatic consumers.
  *
- * Source: the command-code@1.33.0 bundled model table (dist/cli.mjs),
+ * Source: the command-code@1.36.0 bundled model table (dist/cli.mjs),
  * cross-checked with https://commandcode.ai/docs/reference/cli/models.
  * (`stealth/ox-alpha` left this set in command-code@1.32.1, which gave it
- * selectable `['low', 'high', 'max']` efforts.)
+ * selectable `['low', 'high', 'max']` efforts; the preview then ended in
+ * 1.34.0, removing the model from the catalog entirely.)
  * Keep in sync via the dsh-commandcode-upstream skill.
  */
 declare const KNOWN_THINKING_MODELS: ReadonlySet<string>;
@@ -156,15 +157,19 @@ declare const KNOWN_DEALS: Readonly<Record<string, KnownDeal>>;
  * charges by the hour: peak hours are 01:00–04:00 and 06:00–10:00 UTC (7h/day,
  * full price); the other 17 hours are off-peak at half price. The V4 Flash
  * Vision (exp) variant (command-code@1.32.0) shares the V4 Flash windows and
- * peak prices ($0.44/$1.32). Qwen 3.8 Max joined the hourly set by
- * 2026-08-26 with the same windows (the page's prose still names all three
- * DeepSeek variants even though the V4 Pro table row lost its hover
- * annotation — kept here on the prose's authority). The picker shows the
+ * peak prices ($0.44/$1.32) — each row's hover annotation states exactly 2×
+ * that row's displayed off-peak prices. The picker shows the
  * *current* state as a compact
  * label (`Peak`/`Half`) matching the English noun style of the other markers
  * (`Image`, `FREE`), so a developer can tell at a glance whether calling the
  * model right now is cheap or expensive.
  *
+ * Extraction caution: in the page's HTML each annotation div sits inside its
+ * OWN row's container, immediately before the NEXT row starts — flattening
+ * the page to text makes every annotation look like it belongs to the model
+ * printed after it. Verify membership against the enclosing row and the 2×
+ * price relation, not the flat-text neighbor.
+ *
  * Keep in sync with the official pricing page when the model set or the peak
  * windows change (see the dsh-commandcode-upstream skill).
  */
@@ -182,7 +187,7 @@ declare function peakPricingState(modelId: string, now?: number): 'peak' | 'off-
  * for models without time-of-day pricing.
  */
 declare function peakPricingLabel(modelId: string, now?: number): string | undefined;
-declare const COMMAND_CODE_CLI_VERSION = "1.33.0";
+declare const COMMAND_CODE_CLI_VERSION = "1.36.0";
 declare const DEFAULT_API_BASE = "https://api.commandcode.ai";
 declare const DEFAULT_GENERATE_MAX_TOKENS = 64000;
 declare const DEFAULT_MAX_OUTPUT_TOKENS = 65536;

package/lib/index.js

@@ -216,7 +216,7 @@ var CommandCodeAccountPool = class {
 * and API key or subscription, and Command Code's terms apply.
 *
 * Wire protocol (reverse-engineered by the pi plugin, command-code@1.28.4;
-* re-verified against command-code@1.33.0 — endpoints, request shape, and
+* re-verified against command-code@1.36.0 — endpoints, request shape, and
 * stream events unchanged):
 *   POST {apiBase}/alpha/generate
 *   body: { config, memory, taste, skills, params: { model, messages, tools,
@@ -240,6 +240,11 @@ const KNOWN_EFFORTS = {
 		"medium",
 		"xhigh"
 	],
+	"Qwen/Qwen3.8-Flash": [
+		"low",
+		"medium",
+		"xhigh"
+	],
 	"claude-fable-5": [
 		"low",
 		"medium",
@@ -355,11 +360,6 @@ const KNOWN_EFFORTS = {
 		"max"
 	],
 	"sakana/fugu-ultra": ["high", "xhigh"],
-	"stealth/ox-alpha": [
-		"low",
-		"high",
-		"max"
-	],
 	"xai/grok-4.5": [
 		"low",
 		"medium",
@@ -371,6 +371,11 @@ const KNOWN_EFFORTS = {
 		"high",
 		"xhigh"
 	],
+	"z-ai/glm-5.3-flash": [
+		"low",
+		"high",
+		"max"
+	],
 	"zai-org/GLM-5.2": ["high", "max"],
 	"zai-org/GLM-5.3": [
 		"low",
@@ -399,6 +404,7 @@ const KNOWN_IMAGE_MODELS = /* @__PURE__ */ new Set([
 	"Qwen/Qwen3.7-Flash",
 	"Qwen/Qwen3.7-Plus",
 	"Qwen/Qwen3.8-27B",
+	"Qwen/Qwen3.8-Flash",
 	"Qwen/Qwen3.8-Max",
 	"claude-fable-5",
 	"claude-haiku-4-5-20251001",
@@ -430,15 +436,15 @@ const KNOWN_IMAGE_MODELS = /* @__PURE__ */ new Set([
 	"moonshotai/Kimi-K2.7-Code-Highspeed",
 	"moonshotai/Kimi-K3",
 	"sakana/fugu-ultra",
-	"stealth/ox-alpha",
 	"stepfun/Step-3.7-Flash",
 	"thinkingmachines/inkling",
 	"thinkingmachines/inkling-small",
 	"xai/grok-4.5",
-	"xiaomi/mimo-v2.5"
+	"xiaomi/mimo-v2.5",
+	"z-ai/glm-5.3-flash"
 ]);
 /**
-* Models the official CLI's model table (command-code@1.33.0) marks
+* Models the official CLI's model table (command-code@1.36.0) marks
 * `reasoning:!0` but defines no selectable `reasoning_effort` levels — they
 * think automatically, with Command Code driving the depth. This is the
 * authoritative "thinks, effort not adjustable" set: `KNOWN_EFFORTS` (which
@@ -446,10 +452,11 @@ const KNOWN_IMAGE_MODELS = /* @__PURE__ */ new Set([
 * effort levels, and this snapshot is not surfaced in the picker's compact
 * description — it exists for programmatic consumers.
 *
-* Source: the command-code@1.33.0 bundled model table (dist/cli.mjs),
+* Source: the command-code@1.36.0 bundled model table (dist/cli.mjs),
 * cross-checked with https://commandcode.ai/docs/reference/cli/models.
 * (`stealth/ox-alpha` left this set in command-code@1.32.1, which gave it
-* selectable `['low', 'high', 'max']` efforts.)
+* selectable `['low', 'high', 'max']` efforts; the preview then ended in
+* 1.34.0, removing the model from the catalog entirely.)
 * Keep in sync via the dsh-commandcode-upstream skill.
 */
 const KNOWN_THINKING_MODELS = /* @__PURE__ */ new Set([
@@ -499,6 +506,7 @@ const KNOWN_PLANS = {
 	"Qwen/Qwen3.7-Max": "go",
 	"Qwen/Qwen3.7-Plus": "go",
 	"Qwen/Qwen3.8-27B": "go",
+	"Qwen/Qwen3.8-Flash": "go",
 	"Qwen/Qwen3.8-Max": "go",
 	"deepseek/deepseek-v4-flash": "go",
 	"deepseek/deepseek-v4-flash-vision-exp": "go",
@@ -514,7 +522,6 @@ const KNOWN_PLANS = {
 	"moonshotai/Kimi-K3": "go",
 	"nvidia/nemotron-3-ultra-550b-a55b": "go",
 	"poolside/laguna-s-2.1-free": "go",
-	"stealth/ox-alpha": "go",
 	"stepfun/Step-3.5-Flash": "go",
 	"stepfun/Step-3.7-Flash": "go",
 	"tencent/hy3-paid": "go",
@@ -523,6 +530,7 @@ const KNOWN_PLANS = {
 	"xai/grok-4.5": "go",
 	"xiaomi/mimo-v2.5": "go",
 	"xiaomi/mimo-v2.5-pro": "go",
+	"z-ai/glm-5.3-flash": "go",
 	"zai-org/GLM-5": "go",
 	"zai-org/GLM-5.1": "go",
 	"zai-org/GLM-5.2": "go",
@@ -699,10 +707,6 @@ const KNOWN_DEALS = {
 	"poolside/laguna-s-2.1-free": {
 		label: "FREE",
 		free: true
-	},
-	"stealth/ox-alpha": {
-		label: "FREE",
-		free: true
 	}
 };
 /**
@@ -711,23 +715,26 @@ const KNOWN_DEALS = {
 * charges by the hour: peak hours are 01:00–04:00 and 06:00–10:00 UTC (7h/day,
 * full price); the other 17 hours are off-peak at half price. The V4 Flash
 * Vision (exp) variant (command-code@1.32.0) shares the V4 Flash windows and
-* peak prices ($0.44/$1.32). Qwen 3.8 Max joined the hourly set by
-* 2026-08-26 with the same windows (the page's prose still names all three
-* DeepSeek variants even though the V4 Pro table row lost its hover
-* annotation — kept here on the prose's authority). The picker shows the
+* peak prices ($0.44/$1.32) — each row's hover annotation states exactly 2×
+* that row's displayed off-peak prices. The picker shows the
 * *current* state as a compact
 * label (`Peak`/`Half`) matching the English noun style of the other markers
 * (`Image`, `FREE`), so a developer can tell at a glance whether calling the
 * model right now is cheap or expensive.
 *
+* Extraction caution: in the page's HTML each annotation div sits inside its
+* OWN row's container, immediately before the NEXT row starts — flattening
+* the page to text makes every annotation look like it belongs to the model
+* printed after it. Verify membership against the enclosing row and the 2×
+* price relation, not the flat-text neighbor.
+*
 * Keep in sync with the official pricing page when the model set or the peak
 * windows change (see the dsh-commandcode-upstream skill).
 */
 const KNOWN_PEAK_PRICING = /* @__PURE__ */ new Set([
 	"deepseek/deepseek-v4-pro",
 	"deepseek/deepseek-v4-flash",
-	"deepseek/deepseek-v4-flash-vision-exp",
-	"Qwen/Qwen3.8-Max"
+	"deepseek/deepseek-v4-flash-vision-exp"
 ]);
 /** Peak hours (UTC, hour-of-day range end-exclusive): 01–03 and 06–09. */
 const PEAK_HOUR_RANGES = [[1, 4], [6, 10]];
@@ -752,7 +759,7 @@ function peakPricingLabel(modelId, now = Date.now()) {
 	if (state === void 0) return void 0;
 	return state === "peak" ? "Peak" : "Half";
 }
-const COMMAND_CODE_CLI_VERSION = "1.33.0";
+const COMMAND_CODE_CLI_VERSION = "1.36.0";
 const DEFAULT_API_BASE = "https://api.commandcode.ai";
 const DEFAULT_GENERATE_MAX_TOKENS = 64e3;
 const DEFAULT_MAX_OUTPUT_TOKENS = 65536;