pi-clinepass 0.1.2 → 0.1.3

package/CHANGELOG.md

@@ -2,6 +2,10 @@
 
 All notable changes to this project will be documented in this file.
 
+## 0.1.3 - 2026-08-31
+
+- Corrected model input modalities: `glm-5.3` and `deepseek-v4-flash` are text-only, `mimo-v2.5` and `qwen3.8-max` accept images
+
 ## 0.1.2 - 2026-08-30
 
 - New paid model: `cline-pass/glm-5.3-flash` ($0.15/$0.50/$0.03)

package/package.json

@@ -1,54 +1,54 @@
-{
-  "name": "pi-clinepass",
-  "version": "0.1.2",
-  "description": "ClinePass for pi — static measured-price catalog, server-truth billing meter, free models",
-  "keywords": [
-    "ai",
-    "cline",
-    "cline-pass",
-    "clinepass",
-    "pi",
-    "pi-coding-agent",
-    "pi-package",
-    "pi-extension"
-  ],
-  "license": "MIT",
-  "author": "fifidayone",
-  "repository": {
-    "type": "git",
-    "url": "git+https://github.com/fifidayone/pi-clinepass.git"
-  },
-  "type": "module",
-  "main": "src/index.ts",
-  "types": "src/index.ts",
-  "files": [
-    "src",
-    "CHANGELOG.md",
-    "README.md",
-    "LICENSE"
-  ],
-  "engines": {
-    "node": ">=22"
-  },
-  "scripts": {
-    "test": "vitest run",
-    "test:watch": "vitest",
-    "typecheck": "tsc --noEmit"
-  },
-  "peerDependencies": {
-    "@earendil-works/pi-ai": "*",
-    "@earendil-works/pi-coding-agent": "*",
-    "@earendil-works/pi-tui": "*"
-  },
-  "devDependencies": {
-    "@types/node": "^26.3.0",
-    "typescript": "^7.0.2",
-    "vitest": "^4.1.11"
-  },
-  "pi": {
-    "extensions": [
-      "./src/index.ts"
-    ],
-    "image": "https://raw.githubusercontent.com/fifidayone/pi-clinepass/main/assets/report.png"
-  }
-}
+{
+  "name": "pi-clinepass",
+  "version": "0.1.3",
+  "description": "ClinePass for pi — static measured-price catalog, server-truth billing meter, free models",
+  "keywords": [
+    "ai",
+    "cline",
+    "cline-pass",
+    "clinepass",
+    "pi",
+    "pi-coding-agent",
+    "pi-package",
+    "pi-extension"
+  ],
+  "license": "MIT",
+  "author": "fifidayone",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/fifidayone/pi-clinepass.git"
+  },
+  "type": "module",
+  "main": "src/index.ts",
+  "types": "src/index.ts",
+  "files": [
+    "src",
+    "CHANGELOG.md",
+    "README.md",
+    "LICENSE"
+  ],
+  "engines": {
+    "node": ">=22"
+  },
+  "scripts": {
+    "test": "vitest run",
+    "test:watch": "vitest",
+    "typecheck": "tsc --noEmit"
+  },
+  "peerDependencies": {
+    "@earendil-works/pi-ai": "*",
+    "@earendil-works/pi-coding-agent": "*",
+    "@earendil-works/pi-tui": "*"
+  },
+  "devDependencies": {
+    "@types/node": "^26.3.0",
+    "typescript": "^7.0.2",
+    "vitest": "^4.1.11"
+  },
+  "pi": {
+    "extensions": [
+      "./src/index.ts"
+    ],
+    "image": "https://raw.githubusercontent.com/fifidayone/pi-clinepass/main/assets/report.png"
+  }
+}

package/src/catalog.ts

@@ -1,174 +1,174 @@
-/**
- * ClinePass static model catalog.
- *
- * Prices are the measured billing rates from the Cline `/usages` API
- * (verified against real usage), not the published reference table which
- * drifts from actual billing. `cacheWrite` is kept at 0: pi's cost model
- * requires the field, but we do not display or track cache-write pricing.
- *
- * Context is capped to 921,600 for models whose gateway reports ~1M
- * (pool max, not single-node) to leave headroom; max output is capped to
- * 131,072. Models under 1M keep their tested values.
- */
-
-export type ThinkingLevel = "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
-
-export type ThinkingLevelMap = Readonly<Record<ThinkingLevel, string | null>>;
-
-export interface ClinePassModel {
-  id: string;
-  name: string;
-  reasoning: boolean;
-  input: readonly ("text" | "image")[];
-  cost: {
-    input: number;
-    output: number;
-    cacheRead: number;
-    cacheWrite: number;
-  };
-  contextWindow: number;
-  maxTokens: number;
-  thinkingLevelMap: ThinkingLevelMap;
-  compat: {
-    supportsDeveloperRole: false;
-    cacheControlFormat: "anthropic";
-    supportsLongCacheRetention: false;
-  };
-}
-
-/** All 7 levels supported (off → "none"). Used when `supportedEfforts: null` + mandatory false. */
-const ALL_THINKING: ThinkingLevelMap = {
-  off: "none",
-  minimal: "minimal",
-  low: "low",
-  medium: "medium",
-  high: "high",
-  xhigh: "xhigh",
-  max: "max",
-};
-
-/** All levels except off (mandatory reasoning). Used when `mandatory: true` + `supportedEfforts: null`. */
-const ALL_MANDATORY: ThinkingLevelMap = {
-  off: null,
-  minimal: "minimal",
-  low: "low",
-  medium: "medium",
-  high: "high",
-  xhigh: "xhigh",
-  max: "max",
-};
-
-/** Only max/high/low + off. Used for deepseek, glm-5.3, kimi-k3, free models. */
-const MAX_HIGH_LOW: ThinkingLevelMap = {
-  off: "none",
-  minimal: null,
-  low: "low",
-  medium: null,
-  high: "high",
-  xhigh: null,
-  max: "max",
-};
-
-/** Max/high/low but off not allowed (mandatory). */
-const MAX_HIGH_LOW_MANDATORY: ThinkingLevelMap = {
-  off: null,
-  minimal: null,
-  low: "low",
-  medium: null,
-  high: "high",
-  xhigh: null,
-  max: "max",
-};
-
-/** Only xhigh/high + off. Used for glm-5.2. */
-const XHIGH_HIGH: ThinkingLevelMap = {
-  off: "none",
-  minimal: null,
-  low: null,
-  medium: null,
-  high: "high",
-  xhigh: "xhigh",
-  max: null,
-};
-
-/** xhigh/medium/low + off. Used for qwen3.8-max. */
-const XHIGH_MEDIUM_LOW: ThinkingLevelMap = {
-  off: "none",
-  minimal: null,
-  low: "low",
-  medium: "medium",
-  high: null,
-  xhigh: "xhigh",
-  max: null,
-};
-
-/**
- * ClinePass rejects the `developer` role; the API caps at `max_tokens` /
- * `max_completion_tokens` with reasoning excluded from the completion cap,
- * so pi's default `max_completion_tokens` is correct and left unset.
- * `cacheControlFormat: "anthropic"` makes pi send `{type:"ephemeral"}`
- * (no ttl) — verified to engage real caching on ClinePass.
- * `supportsLongCacheRetention: false` keeps ttl from ever being sent
- * (ttl causes HTTP 500).
- */
-const COMPAT = {
-  supportsDeveloperRole: false as const,
-  cacheControlFormat: "anthropic" as const,
-  supportsLongCacheRetention: false as const,
-};
-
-function model(
-  id: string,
-  name: string,
-  cost: [input: number, output: number, cacheRead: number],
-  contextWindow: number,
-  maxTokens: number,
-  input: readonly ("text" | "image")[] = ["text"],
-  thinkingLevelMap: ThinkingLevelMap = ALL_THINKING,
-): ClinePassModel {
-  return {
-    id,
-    name,
-    reasoning: true,
-    input,
-    cost: { input: cost[0], output: cost[1], cacheRead: cost[2], cacheWrite: 0 },
-    contextWindow,
-    maxTokens,
-    thinkingLevelMap,
-    compat: COMPAT,
-  };
-}
-
-/** Measured prices ($/1M tokens: input/output/cacheRead) — latest verified. */
-export const MODELS: readonly ClinePassModel[] = [
-  // ── Free models (Cline free tier, cost 0) ──────────────────────────────
-  model("cline-free/longcat-2.0", "LongCat 2.0", [0, 0, 0], 921_600, 131_072, ["text"], ALL_THINKING),
-  model("z-ai/glm-5.3-flash", "GLM-5.3 Flash", [0, 0, 0], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW_MANDATORY),
-  model("deepseek/deepseek-v4-flash", "DeepSeek V4 Flash", [0, 0, 0], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW),
-  model("poolside/laguna-s-2.1:free", "Laguna S-2.1", [0, 0, 0], 262_144, 131_072, ["text"], ALL_THINKING),
-  // ── ClinePass models (measured billing prices) ─────────────────────────
-  model("cline-pass/glm-5.3-flash", "GLM-5.3 Flash", [0.15, 0.5, 0.03], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW_MANDATORY),
-  model("cline-pass/glm-5.3", "GLM-5.3", [1.4, 4.4, 0.26], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW_MANDATORY),
-  model("cline-pass/glm-5.2", "GLM-5.2", [1.4, 4.4, 0.26], 921_600, 131_072, ["text"], XHIGH_HIGH),
-  model("cline-pass/kimi-k2.7-code", "Kimi K2.7 Code", [1.58, 6.67, 0.32], 262_144, 131_072, ["text", "image"], ALL_MANDATORY),
-  model("cline-pass/kimi-k2.6", "Kimi K2.6", [1.58, 6.67, 0.27], 262_144, 131_072, ["text", "image"], ALL_THINKING),
-  model("cline-pass/kimi-k3", "Kimi K3", [6.0, 30.0, 0.6], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW),
-  model("cline-pass/deepseek-v4-pro", "DeepSeek V4 Pro", [1.65, 4.95, 0.06], 921_600, 131_072, ["text"], MAX_HIGH_LOW),
-  model("cline-pass/deepseek-v4-flash", "DeepSeek V4 Flash", [0.44, 1.32, 0.014], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW),
-  model("cline-pass/mimo-v2.5", "MiMo-V2.5", [0.14, 0.28, 0.0028], 921_600, 131_072, ["text"], ALL_THINKING),
-  model("cline-pass/mimo-v2.5-pro", "MiMo-V2.5-Pro", [0.435, 0.87, 0.0036], 921_600, 131_072, ["text"], ALL_THINKING),
-  model("cline-pass/minimax-m3", "MiniMax M3", [0.5, 2.0, 0.1], 921_600, 131_072, ["text", "image"], ALL_THINKING),
-  model("cline-pass/qwen3.7-plus", "Qwen3.7 Plus", [0.67, 2.67, 0.07], 921_600, 131_072, ["text", "image"], ALL_THINKING),
-  model("cline-pass/qwen3.7-max", "Qwen3.7 Max", [4.17, 12.5, 0.83], 921_600, 131_072, ["text"], ALL_THINKING),
-  model("cline-pass/qwen3.8-max", "Qwen3.8 Max", [2.75, 8.25, 0.34], 921_600, 131_072, ["text"], XHIGH_MEDIUM_LOW),
-];
-
-export function modelIds(): string[] {
-  return MODELS.map((m) => m.id);
-}
-
-/** True for the Cline free-tier models — derived from the catalog itself
- * (cost 0), so adding a free model never needs a second edit here. */
-export function isFreeModel(id: string): boolean {
-  return MODELS.some((m) => m.id === id && m.cost.input === 0);
-}
+/**
+ * ClinePass static model catalog.
+ *
+ * Prices are the measured billing rates from the Cline `/usages` API
+ * (verified against real usage), not the published reference table which
+ * drifts from actual billing. `cacheWrite` is kept at 0: pi's cost model
+ * requires the field, but we do not display or track cache-write pricing.
+ *
+ * Context is capped to 921,600 for models whose gateway reports ~1M
+ * (pool max, not single-node) to leave headroom; max output is capped to
+ * 131,072. Models under 1M keep their tested values.
+ */
+
+export type ThinkingLevel = "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "max";
+
+export type ThinkingLevelMap = Readonly<Record<ThinkingLevel, string | null>>;
+
+export interface ClinePassModel {
+  id: string;
+  name: string;
+  reasoning: boolean;
+  input: readonly ("text" | "image")[];
+  cost: {
+    input: number;
+    output: number;
+    cacheRead: number;
+    cacheWrite: number;
+  };
+  contextWindow: number;
+  maxTokens: number;
+  thinkingLevelMap: ThinkingLevelMap;
+  compat: {
+    supportsDeveloperRole: false;
+    cacheControlFormat: "anthropic";
+    supportsLongCacheRetention: false;
+  };
+}
+
+/** All 7 levels supported (off → "none"). Used when `supportedEfforts: null` + mandatory false. */
+const ALL_THINKING: ThinkingLevelMap = {
+  off: "none",
+  minimal: "minimal",
+  low: "low",
+  medium: "medium",
+  high: "high",
+  xhigh: "xhigh",
+  max: "max",
+};
+
+/** All levels except off (mandatory reasoning). Used when `mandatory: true` + `supportedEfforts: null`. */
+const ALL_MANDATORY: ThinkingLevelMap = {
+  off: null,
+  minimal: "minimal",
+  low: "low",
+  medium: "medium",
+  high: "high",
+  xhigh: "xhigh",
+  max: "max",
+};
+
+/** Only max/high/low + off. Used for deepseek, glm-5.3, kimi-k3, free models. */
+const MAX_HIGH_LOW: ThinkingLevelMap = {
+  off: "none",
+  minimal: null,
+  low: "low",
+  medium: null,
+  high: "high",
+  xhigh: null,
+  max: "max",
+};
+
+/** Max/high/low but off not allowed (mandatory). */
+const MAX_HIGH_LOW_MANDATORY: ThinkingLevelMap = {
+  off: null,
+  minimal: null,
+  low: "low",
+  medium: null,
+  high: "high",
+  xhigh: null,
+  max: "max",
+};
+
+/** Only xhigh/high + off. Used for glm-5.2. */
+const XHIGH_HIGH: ThinkingLevelMap = {
+  off: "none",
+  minimal: null,
+  low: null,
+  medium: null,
+  high: "high",
+  xhigh: "xhigh",
+  max: null,
+};
+
+/** xhigh/medium/low + off. Used for qwen3.8-max. */
+const XHIGH_MEDIUM_LOW: ThinkingLevelMap = {
+  off: "none",
+  minimal: null,
+  low: "low",
+  medium: "medium",
+  high: null,
+  xhigh: "xhigh",
+  max: null,
+};
+
+/**
+ * ClinePass rejects the `developer` role; the API caps at `max_tokens` /
+ * `max_completion_tokens` with reasoning excluded from the completion cap,
+ * so pi's default `max_completion_tokens` is correct and left unset.
+ * `cacheControlFormat: "anthropic"` makes pi send `{type:"ephemeral"}`
+ * (no ttl) — verified to engage real caching on ClinePass.
+ * `supportsLongCacheRetention: false` keeps ttl from ever being sent
+ * (ttl causes HTTP 500).
+ */
+const COMPAT = {
+  supportsDeveloperRole: false as const,
+  cacheControlFormat: "anthropic" as const,
+  supportsLongCacheRetention: false as const,
+};
+
+function model(
+  id: string,
+  name: string,
+  cost: [input: number, output: number, cacheRead: number],
+  contextWindow: number,
+  maxTokens: number,
+  input: readonly ("text" | "image")[] = ["text"],
+  thinkingLevelMap: ThinkingLevelMap = ALL_THINKING,
+): ClinePassModel {
+  return {
+    id,
+    name,
+    reasoning: true,
+    input,
+    cost: { input: cost[0], output: cost[1], cacheRead: cost[2], cacheWrite: 0 },
+    contextWindow,
+    maxTokens,
+    thinkingLevelMap,
+    compat: COMPAT,
+  };
+}
+
+/** Measured prices ($/1M tokens: input/output/cacheRead) — latest verified. */
+export const MODELS: readonly ClinePassModel[] = [
+  // ── Free models (Cline free tier, cost 0) ──────────────────────────────
+  model("cline-free/longcat-2.0", "LongCat 2.0", [0, 0, 0], 921_600, 131_072, ["text"], ALL_THINKING),
+  model("z-ai/glm-5.3-flash", "GLM-5.3 Flash", [0, 0, 0], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW_MANDATORY),
+  model("deepseek/deepseek-v4-flash", "DeepSeek V4 Flash", [0, 0, 0], 921_600, 131_072, ["text"], MAX_HIGH_LOW),
+  model("poolside/laguna-s-2.1:free", "Laguna S-2.1", [0, 0, 0], 262_144, 131_072, ["text"], ALL_THINKING),
+  // ── ClinePass models (measured billing prices) ─────────────────────────
+  model("cline-pass/glm-5.3-flash", "GLM-5.3 Flash", [0.15, 0.5, 0.03], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW_MANDATORY),
+  model("cline-pass/glm-5.3", "GLM-5.3", [1.4, 4.4, 0.26], 921_600, 131_072, ["text"], MAX_HIGH_LOW_MANDATORY),
+  model("cline-pass/glm-5.2", "GLM-5.2", [1.4, 4.4, 0.26], 921_600, 131_072, ["text"], XHIGH_HIGH),
+  model("cline-pass/kimi-k2.7-code", "Kimi K2.7 Code", [1.58, 6.67, 0.32], 262_144, 131_072, ["text", "image"], ALL_MANDATORY),
+  model("cline-pass/kimi-k2.6", "Kimi K2.6", [1.58, 6.67, 0.27], 262_144, 131_072, ["text", "image"], ALL_THINKING),
+  model("cline-pass/kimi-k3", "Kimi K3", [6.0, 30.0, 0.6], 921_600, 131_072, ["text", "image"], MAX_HIGH_LOW),
+  model("cline-pass/deepseek-v4-pro", "DeepSeek V4 Pro", [1.65, 4.95, 0.06], 921_600, 131_072, ["text"], MAX_HIGH_LOW),
+  model("cline-pass/deepseek-v4-flash", "DeepSeek V4 Flash", [0.44, 1.32, 0.014], 921_600, 131_072, ["text"], MAX_HIGH_LOW),
+  model("cline-pass/mimo-v2.5", "MiMo-V2.5", [0.14, 0.28, 0.0028], 921_600, 131_072, ["text", "image"], ALL_THINKING),
+  model("cline-pass/mimo-v2.5-pro", "MiMo-V2.5-Pro", [0.435, 0.87, 0.0036], 921_600, 131_072, ["text"], ALL_THINKING),
+  model("cline-pass/minimax-m3", "MiniMax M3", [0.5, 2.0, 0.1], 921_600, 131_072, ["text", "image"], ALL_THINKING),
+  model("cline-pass/qwen3.7-plus", "Qwen3.7 Plus", [0.67, 2.67, 0.07], 921_600, 131_072, ["text", "image"], ALL_THINKING),
+  model("cline-pass/qwen3.7-max", "Qwen3.7 Max", [4.17, 12.5, 0.83], 921_600, 131_072, ["text"], ALL_THINKING),
+  model("cline-pass/qwen3.8-max", "Qwen3.8 Max", [2.75, 8.25, 0.34], 921_600, 131_072, ["text", "image"], XHIGH_MEDIUM_LOW),
+];
+
+export function modelIds(): string[] {
+  return MODELS.map((m) => m.id);
+}
+
+/** True for the Cline free-tier models — derived from the catalog itself
+ * (cost 0), so adding a free model never needs a second edit here. */
+export function isFreeModel(id: string): boolean {
+  return MODELS.some((m) => m.id === id && m.cost.input === 0);
+}