tokentracker-cli 0.96.2 → 0.96.4

package/src/lib/pricing/curated-overrides.json

@@ -2,7 +2,8 @@
   "_meta": {
     "note": "Curated price overrides. Always wins over LiteLLM. Two reasons to live here: (1) self-defined alias names that LiteLLM will never carry (kiro-*, hy3-*, composer-*, kimi-for-coding, free-tier OpenRouter routes); (2) prices we want to pin even if LiteLLM has the model (e.g. cache_write fields LiteLLM often omits). Units: USD per million tokens. Edit this file to override pricing without redeploying.",
     "units": "usd_per_million_tokens",
-    "deepseek_v4_pricing_note": "Official DeepSeek V4 pricing verified on 2026-08-21. Peak hours are 01:00-04:00 and 06:00-10:00 UTC; all other hours are off-peak at half price. Exact entries store peak rates. computeRowCost applies the time-of-use discount from each row's UTC hour_start."
+    "deepseek_v4_pricing_note": "Official DeepSeek V4 pricing verified on 2026-08-21. Peak hours are 01:00-04:00 and 06:00-10:00 UTC; all other hours are off-peak at half price. Exact entries store peak rates. computeRowCost applies the time-of-use discount from each row's UTC hour_start.",
+    "iflytek_maas_pricing_note": "iFlytek MaaS inference prices supplied for AStudio on 2026-08-31. AStudio homepage: https://agent.xfyun.cn/. Official pricing source: https://maas.xfyun.cn/modelSquare. Source values are RMB per million tokens, converted at 7.2 RMB/USD, then rounded to two decimal places for billing. Models without a dedicated cache-hit price use the normal input rate; cache writes also use the normal input rate. iFlytek MaaS prices are source-scoped and do not inherit DeepSeek time-of-use discounts."
   },
   "exact": {
     "claude-fable-5": {
@@ -420,6 +421,71 @@
   "alias": {
     "auto": "composer-1"
   },
+  "source_exact": {
+    "acode": {
+      "xopglm53": { "input": 1.11, "output": 3.89, "cache_read": 0.28, "cache_write": 1.11 },
+      "xopdeepseekv4pro0813": { "input": 1.25, "output": 3.75, "cache_read": 0.04, "cache_write": 1.25 },
+      "xopdeepseekv4flash0731": { "input": 0.14, "output": 0.28, "cache_read": 0.03, "cache_write": 0.14 },
+      "xopkimik27code": { "input": 0.90, "output": 3.75, "cache_read": 0.90, "cache_write": 0.90 },
+      "xopglm52": { "input": 1.11, "output": 3.89, "cache_read": 0.28, "cache_write": 1.11 },
+      "xopdeepseekv4flash": { "input": 0.14, "output": 0.28, "cache_read": 0.03, "cache_write": 0.14 },
+      "xopkimik26": { "input": 0.90, "output": 3.75, "cache_read": 0.18, "cache_write": 0.90 },
+      "xopdeepseekv4pro": { "input": 1.67, "output": 3.33, "cache_read": 0.14, "cache_write": 1.67 },
+      "xopqwen36v35b": { "input": 0.15, "output": 0.90, "cache_read": 0.15, "cache_write": 0.15 },
+      "xophunyuan7bmt": { "input": 0.07, "output": 0.28, "cache_read": 0.07, "cache_write": 0.07 },
+      "xoppaddleocrv16": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
+      "xsparkx2flash": { "input": 0.14, "output": 0.28, "cache_read": 0.14, "cache_write": 0.14 },
+      "xopglm51": { "input": 1.11, "output": 3.89, "cache_read": 0.22, "cache_write": 1.11 },
+      "xsparkx2": { "input": 0.42, "output": 0.42, "cache_read": 0.42, "cache_write": 0.42 },
+      "xop35qwen2b": { "input": 0.03, "output": 0.06, "cache_read": 0.03, "cache_write": 0.03 },
+      "xopqwen35397b": { "input": 0.17, "output": 1.00, "cache_read": 0.17, "cache_write": 0.17 },
+      "xminimaxm25": { "input": 0.29, "output": 1.17, "cache_read": 0.29, "cache_write": 0.29 },
+      "xopglm5": { "input": 0.83, "output": 3.06, "cache_read": 0.17, "cache_write": 0.83 },
+      "xopkimik25": { "input": 0.56, "output": 2.92, "cache_read": 0.56, "cache_write": 0.56 },
+      "xopdeepseekv32": { "input": 0.14, "output": 0.21, "cache_read": 0.14, "cache_write": 0.14 },
+      "xop3qwencodernext": { "input": 0.35, "output": 1.39, "cache_read": 0.35, "cache_write": 0.35 },
+      "xopglmv47flash": { "input": 0.14, "output": 0.21, "cache_read": 0.14, "cache_write": 0.14 },
+      "xopglm47blth2": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
+      "xop3qwen32bvl": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "xopdeepseekocr": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
+      "xophunyuanocr": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
+      "xop3qwen80bnext": { "input": 0.08, "output": 0.33, "cache_read": 0.08, "cache_write": 0.08 },
+      "xop3qwen235b2507": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
+      "xop3qwen30b2507": { "input": 0.06, "output": 0.63, "cache_read": 0.06, "cache_write": 0.06 },
+      "xop3qwen235b": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
+      "xop3qwen30b": { "input": 0.06, "output": 0.63, "cache_read": 0.06, "cache_write": 0.06 },
+      "xop3qwen32b": { "input": 0.17, "output": 1.67, "cache_read": 0.17, "cache_write": 0.17 },
+      "xdeepseekv3": { "input": 0.22, "output": 0.89, "cache_read": 0.22, "cache_write": 0.22 },
+      "xdeepseekr1": { "input": 0.44, "output": 1.78, "cache_read": 0.44, "cache_write": 0.44 },
+      "xdeepseekr1qwen32b": { "input": 0.22, "output": 0.67, "cache_read": 0.22, "cache_write": 0.22 },
+      "xopkimik2blth": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
+      "xopkimik2blins": { "input": 0.56, "output": 2.22, "cache_read": 0.56, "cache_write": 0.56 },
+      "xop3qwen8breranker": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
+      "xop3qwen8bembedding": { "input": 0.00, "output": 0.00, "cache_read": 0.00, "cache_write": 0.00 },
+      "xop3qwen0b6": { "input": 0.04, "output": 0.42, "cache_read": 0.04, "cache_write": 0.04 },
+      "xop3qwen4b": { "input": 0.04, "output": 0.42, "cache_read": 0.04, "cache_write": 0.04 },
+      "xqwen257bchat": { "input": 0.07, "output": 0.14, "cache_read": 0.07, "cache_write": 0.07 },
+      "xop3qwen14b": { "input": 0.14, "output": 1.39, "cache_read": 0.14, "cache_write": 0.14 },
+      "xop3qwen8b": { "input": 0.07, "output": 0.69, "cache_read": 0.07, "cache_write": 0.07 },
+      "xsparkprox": { "input": 1.11, "output": 5.56, "cache_read": 1.11, "cache_write": 1.11 },
+      "xspark13b6k": { "input": 0.28, "output": 0.83, "cache_read": 0.28, "cache_write": 0.28 },
+      "spark mini": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "spark mini instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "spark tiny": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "internlm2.5_7b_chat": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "internlm2.5_1.8b_chat": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "qwen_v2.5_7b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "xsqwen2d53b": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "qwen_v2.5_3b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "qwen_v2.5_1.5b_instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "qwen_v2.5_1.5b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "qwen_v2.5_0.5b_instruct": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "qwen_v2.5_0.5b_base": { "input": 0.28, "output": 1.11, "cache_read": 0.28, "cache_write": 0.28 },
+      "xsqwenv2s1b5c": { "input": 0.14, "output": 0.28, "cache_read": 0.14, "cache_write": 0.14 },
+      "xsqwenv2s0b5c": { "input": 0.28, "output": 0.56, "cache_read": 0.28, "cache_write": 0.28 },
+      "xqwen14bchat": { "input": 0.28, "output": 0.83, "cache_read": 0.28, "cache_write": 0.28 }
+    }
+  },
   "fuzzy": [
     {
       "match": "gpt-6-astra",

package/src/lib/pricing/index.js

@@ -35,7 +35,6 @@ const DEEPSEEK_TIME_PRICED_MODELS = [
   "deepseek-v4-flash",
   "deepseek-v4-pro",
 ];
-
 // Sync seed load. Done at require-time so callers that haven't awaited
 // ensurePricingLoaded() (e.g. tests, vite mock startup, edge functions) still
 // get LiteLLM-backed pricing instead of all-zero. ensurePricingLoaded() will
@@ -163,6 +162,9 @@ function isDeepSeekOffPeak(row) {
 
 function getRowPricing(row) {
   const pricing = getModelPricing(row?.model, { source: row?.source });
+  // AStudio uses iFlytek MaaS fixed prices and does not inherit DeepSeek public API
+  // time-based discounts.
+  if (String(row?.source || "").toLowerCase() === "acode") return pricing;
   if (!isDeepSeekTimePricedModel(row?.model) || !isDeepSeekOffPeak(row)) return pricing;
   return {
     ...pricing,
@@ -192,7 +194,8 @@ function computeRowCost(row) {
     reportedCost > 0
   ) return reportedCost;
   const pricing = getRowPricing(row);
-  const reasoningIncludedInOutput = row.source === "codex" || row.source === "every-code";
+  const reasoningIncludedInOutput =
+    row.source === "codex" || row.source === "acode" || row.source === "every-code";
   const reasoningCost = reasoningIncludedInOutput
     ? 0
     : (row.reasoning_output_tokens || 0) * (pricing.output || 0);

package/src/lib/pricing/matcher.js

@@ -1,6 +1,7 @@
 // Pure pricing-lookup logic. No I/O, no async. Tested in isolation.
 //
 // Resolve order:
+//   0. CURATED source exact match (source-specific pricing)
 //   1. CURATED exact match (self-defined aliases like kiro-*, hy3-*)
 //   2. LiteLLM exact match (mainstream claude/gpt-5/gemini)
 //   3. CURATED alias (e.g. "auto" -> "composer-1")
@@ -164,6 +165,15 @@ function normalizeWorkbuddyModel(model) {
   return model;
 }
 
+function normalizeIFlytekMaasModel(model) {
+  if (!model || typeof model !== "string") return model;
+  const trimmed = model.trim();
+  const lower = trimmed.toLowerCase();
+  // The default Agent slug maps to the Spark X2 service in the supplied price list.
+  if (lower === "xsparkx2agent") return "xsparkx2";
+  return trimmed;
+}
+
 // Unsloth Studio can mix local engines and paid API providers in one database.
 // The parser qualifies metered models with their provider and marks local,
 // subscription-backed, or ambiguous custom routes as unpriced. Returning a
@@ -182,6 +192,7 @@ function normalizeUnslothModel(model) {
 // raw model name is preserved for storage/display). Add a source here when its
 // model strings don't match the LiteLLM/curated keys verbatim.
 const SOURCE_MODEL_NORMALIZERS = {
+  acode: normalizeIFlytekMaasModel,
   antigravity: normalizeAntigravityModel,
   claude: normalizeClaudeModel,
   cursor: normalizeCursorModel,
@@ -242,6 +253,22 @@ function lookupPricing(model, { curated, litellm, source } = {}) {
   const lower = lookupModel.toLowerCase();
   const dotForm = buildDotRestoredModel(lookupModel);
 
+  // 0. CURATED source exact. Source-specific prices apply only to their source,
+  // preventing collisions with public prices for same-named models from other CLIs.
+  const sourceKey = typeof source === "string" ? source.toLowerCase() : "";
+  // AStudio does not disclose its routed model. Stop before generic aliases
+  // and fuzzy matching can turn an unresolved router into a priced model.
+  if (sourceKey === "acode" && (lower === "auto" || lower.endsWith("-auto"))) {
+    return { hit: false, source: "miss", value: null };
+  }
+  const sourceExact = curated.source_exact?.[sourceKey];
+  if (sourceExact) {
+    const sourceValue = lookupExactCaseInsensitive(sourceExact, lookupModel);
+    if (sourceValue) {
+      return { hit: true, source: "curated:source-exact", value: sourceValue };
+    }
+  }
+
   // 1. CURATED exact
   if (curated.exact && curated.exact[lookupModel]) {
     return { hit: true, source: "curated:exact", value: curated.exact[lookupModel] };
@@ -382,6 +409,7 @@ module.exports = {
   lookupPricing,
   stripReasoningSuffix,
   normalizeAntigravityModel,
+  normalizeIFlytekMaasModel,
   normalizeClaudeModel,
   normalizeCursorModel,
   normalizeZedModel,