@cursor/july 0.1.30 → 0.1.31

package/src/internal/model-pricing.ts

@@ -1,426 +0,0 @@
-/**
- * Best-effort dollar-cost estimation for agent turns.
- *
- * The Cursor harness streams token usage (`turn-ended`) but no dollar cost,
- * and server-billed cost is only queryable for cloud agents — so the
- * framework prices token usage against a built-in snapshot of Cursor's
- * published per-model API rates. Estimates are marked `source: "estimate"`
- * everywhere they surface (traces, playground, A/B metrics, CLI); the billed
- * amount on the Cursor dashboard remains the source of truth.
- *
- * Unknown models produce no estimate rather than a wrong one. Deployments
- * can extend or correct the table without a package upgrade via the
- * `AGENT_SERVE_MODEL_PRICING` environment variable (a JSON object merged
- * over the defaults, keyed by model id).
- */
-
-import type { ModelSetting, TurnCost, TurnUsage } from "../types.js";
-
-/** USD per million tokens for one pricing tier. */
-export interface ModelPricingTier {
-  inputUsdPerMTok: number;
-  outputUsdPerMTok: number;
-  cacheReadUsdPerMTok: number;
-  /** Defaults to {@link inputUsdPerMTok} when omitted (provider convention). */
-  cacheWriteUsdPerMTok?: number;
-}
-
-export interface ModelPricing extends ModelPricingTier {
-  /** Rates when the model's `fast` parameter is `"true"`. */
-  fast?: ModelPricingTier;
-}
-
-const M = 1_000_000;
-
-/**
- * Snapshot of Cursor API per-model token rates (USD per million tokens).
- * Keyed by normalized model id; family aliases (`sonnet`, `gpt`, …) point at
- * the model they currently resolve to.
- */
-export const DEFAULT_MODEL_PRICING: Readonly<Record<string, ModelPricing>> = {
-  // xAI / Cursor-served Grok
-  "grok-4.5": {
-    inputUsdPerMTok: 2,
-    outputUsdPerMTok: 6,
-    cacheReadUsdPerMTok: 0.5,
-    fast: { inputUsdPerMTok: 4, outputUsdPerMTok: 18, cacheReadUsdPerMTok: 1 },
-  },
-  "grok-4.20": {
-    inputUsdPerMTok: 2,
-    outputUsdPerMTok: 6,
-    cacheReadUsdPerMTok: 0.2,
-  },
-  "grok-4.3": {
-    inputUsdPerMTok: 1.25,
-    outputUsdPerMTok: 2.5,
-    cacheReadUsdPerMTok: 0.125,
-  },
-  "grok-code": {
-    inputUsdPerMTok: 0.2,
-    outputUsdPerMTok: 1.5,
-    cacheReadUsdPerMTok: 0.02,
-  },
-
-  // Anthropic
-  "opus-5": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 25,
-    cacheReadUsdPerMTok: 0.5,
-    cacheWriteUsdPerMTok: 6.25,
-    fast: {
-      inputUsdPerMTok: 10,
-      outputUsdPerMTok: 50,
-      cacheReadUsdPerMTok: 1,
-      cacheWriteUsdPerMTok: 12.5,
-    },
-  },
-  "opus-4.8": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 25,
-    cacheReadUsdPerMTok: 0.5,
-    cacheWriteUsdPerMTok: 6.25,
-    fast: {
-      inputUsdPerMTok: 10,
-      outputUsdPerMTok: 50,
-      cacheReadUsdPerMTok: 1,
-      cacheWriteUsdPerMTok: 12.5,
-    },
-  },
-  "opus-4.7": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 25,
-    cacheReadUsdPerMTok: 0.5,
-    cacheWriteUsdPerMTok: 6.25,
-    fast: {
-      inputUsdPerMTok: 30,
-      outputUsdPerMTok: 150,
-      cacheReadUsdPerMTok: 3,
-      cacheWriteUsdPerMTok: 37.5,
-    },
-  },
-  "opus-4.6": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 25,
-    cacheReadUsdPerMTok: 0.5,
-    cacheWriteUsdPerMTok: 6.25,
-  },
-  "opus-4.5": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 25,
-    cacheReadUsdPerMTok: 0.5,
-    cacheWriteUsdPerMTok: 6.25,
-  },
-  "sonnet-5": {
-    inputUsdPerMTok: 3,
-    outputUsdPerMTok: 15,
-    cacheReadUsdPerMTok: 0.3,
-    cacheWriteUsdPerMTok: 3.75,
-  },
-  "sonnet-4.6": {
-    inputUsdPerMTok: 3,
-    outputUsdPerMTok: 15,
-    cacheReadUsdPerMTok: 0.3,
-    cacheWriteUsdPerMTok: 3.75,
-  },
-  "sonnet-4.5": {
-    inputUsdPerMTok: 3,
-    outputUsdPerMTok: 15,
-    cacheReadUsdPerMTok: 0.3,
-    cacheWriteUsdPerMTok: 3.75,
-  },
-  "haiku-4.5": {
-    inputUsdPerMTok: 1,
-    outputUsdPerMTok: 5,
-    cacheReadUsdPerMTok: 0.1,
-    cacheWriteUsdPerMTok: 1.25,
-  },
-  "fable-5": {
-    inputUsdPerMTok: 10,
-    outputUsdPerMTok: 50,
-    cacheReadUsdPerMTok: 1,
-    cacheWriteUsdPerMTok: 12.5,
-  },
-
-  // OpenAI
-  "gpt-5.6": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 30,
-    cacheReadUsdPerMTok: 0.5,
-    cacheWriteUsdPerMTok: 6.25,
-    fast: {
-      inputUsdPerMTok: 10,
-      outputUsdPerMTok: 60,
-      cacheReadUsdPerMTok: 1,
-      cacheWriteUsdPerMTok: 12.5,
-    },
-  },
-  "gpt-5.5": {
-    inputUsdPerMTok: 5,
-    outputUsdPerMTok: 30,
-    cacheReadUsdPerMTok: 0.5,
-  },
-  "gpt-5.4": {
-    inputUsdPerMTok: 2.5,
-    outputUsdPerMTok: 15,
-    cacheReadUsdPerMTok: 0.25,
-  },
-  "gpt-5.2": {
-    inputUsdPerMTok: 1.75,
-    outputUsdPerMTok: 14,
-    cacheReadUsdPerMTok: 0.175,
-  },
-  "gpt-mini": {
-    inputUsdPerMTok: 0.75,
-    outputUsdPerMTok: 4.5,
-    cacheReadUsdPerMTok: 0.075,
-  },
-  "codex-5.3": {
-    inputUsdPerMTok: 1.75,
-    outputUsdPerMTok: 14,
-    cacheReadUsdPerMTok: 0.175,
-  },
-  "codex-5.2": {
-    inputUsdPerMTok: 1.75,
-    outputUsdPerMTok: 14,
-    cacheReadUsdPerMTok: 0.175,
-  },
-
-  // Google
-  "gemini-3.1-pro": {
-    inputUsdPerMTok: 2,
-    outputUsdPerMTok: 12,
-    cacheReadUsdPerMTok: 0.2,
-  },
-  "gemini-3-pro": {
-    inputUsdPerMTok: 2,
-    outputUsdPerMTok: 12,
-    cacheReadUsdPerMTok: 0.2,
-  },
-  "gemini-3.6-flash": {
-    inputUsdPerMTok: 1.5,
-    outputUsdPerMTok: 7.5,
-    cacheReadUsdPerMTok: 0.15,
-  },
-  "gemini-3.5-flash": {
-    inputUsdPerMTok: 1.5,
-    outputUsdPerMTok: 9,
-    cacheReadUsdPerMTok: 0.15,
-  },
-
-  // Cursor Composer
-  "composer-2.5": {
-    inputUsdPerMTok: 0.5,
-    outputUsdPerMTok: 2.5,
-    cacheReadUsdPerMTok: 0.2,
-    fast: {
-      inputUsdPerMTok: 3,
-      outputUsdPerMTok: 15,
-      cacheReadUsdPerMTok: 0.5,
-    },
-  },
-  "composer-2": {
-    inputUsdPerMTok: 0.5,
-    outputUsdPerMTok: 2.5,
-    cacheReadUsdPerMTok: 0.2,
-  },
-
-  // Moonshot / Zhipu
-  "kimi-k3": {
-    inputUsdPerMTok: 3,
-    outputUsdPerMTok: 15,
-    cacheReadUsdPerMTok: 0.3,
-  },
-  "kimi-k2.7": {
-    inputUsdPerMTok: 0.95,
-    outputUsdPerMTok: 4,
-    cacheReadUsdPerMTok: 0.19,
-  },
-  "kimi-k2.5": {
-    inputUsdPerMTok: 0.6,
-    outputUsdPerMTok: 3,
-    cacheReadUsdPerMTok: 0.1,
-  },
-  "glm-5.2": {
-    inputUsdPerMTok: 1.4,
-    outputUsdPerMTok: 4.4,
-    cacheReadUsdPerMTok: 0.26,
-  },
-};
-
-/** Family aliases → the model id they currently resolve to. */
-const MODEL_ID_ALIASES: Readonly<Record<string, string>> = {
-  grok: "grok-4.20",
-  "grok-latest": "grok-4.3",
-  "grok-code-latest": "grok-code",
-  opus: "opus-5",
-  "opus-latest": "opus-5",
-  sonnet: "sonnet-4.6",
-  "sonnet-latest": "sonnet-5",
-  haiku: "haiku-4.5",
-  "haiku-latest": "haiku-4.5",
-  fable: "fable-5",
-  gpt: "gpt-5.6",
-  "gpt-latest": "gpt-5.6",
-  "gpt-5-6-sol": "gpt-5.6",
-  "gpt-5-5": "gpt-5.5",
-  "gpt-mini-latest": "gpt-mini",
-  codex: "codex-5.3",
-  "codex-latest": "codex-5.3",
-  gemini: "gemini-3.1-pro",
-  "gemini-latest": "gemini-3.1-pro",
-  "gemini-pro": "gemini-3.1-pro",
-  "gemini-pro-latest": "gemini-3.1-pro",
-  "gemini-flash": "gemini-3.6-flash",
-  "gemini-flash-latest": "gemini-3.6-flash",
-  composer: "composer-2.5",
-  "composer-latest": "composer-2.5",
-  "composer-2-5": "composer-2.5",
-  kimi: "kimi-k2.7",
-  "kimi-latest": "kimi-k2.7",
-};
-
-/** Env var holding a JSON pricing override merged over the defaults. */
-export const MODEL_PRICING_ENV_VAR = "AGENT_SERVE_MODEL_PRICING";
-
-let cachedEnvOverride:
-  | { raw: string | undefined; table: Record<string, ModelPricing> }
-  | undefined;
-
-function envOverrideTable(): Record<string, ModelPricing> {
-  const raw = process.env[MODEL_PRICING_ENV_VAR];
-  if (cachedEnvOverride !== undefined && cachedEnvOverride.raw === raw) {
-    return cachedEnvOverride.table;
-  }
-  let table: Record<string, ModelPricing> = {};
-  if (raw !== undefined && raw.trim() !== "") {
-    try {
-      const parsed = JSON.parse(raw) as unknown;
-      if (typeof parsed === "object" && parsed !== null) {
-        for (const [key, value] of Object.entries(parsed)) {
-          const tier = parsePricing(value);
-          if (tier !== undefined) {
-            table[normalizeModelId(key)] = tier;
-          }
-        }
-      }
-    } catch {
-      table = {};
-    }
-  }
-  cachedEnvOverride = { raw, table };
-  return table;
-}
-
-function parseTier(value: unknown): ModelPricingTier | undefined {
-  if (typeof value !== "object" || value === null) {
-    return undefined;
-  }
-  const row = value as Record<string, unknown>;
-  const input = row.inputUsdPerMTok;
-  const output = row.outputUsdPerMTok;
-  const cacheRead = row.cacheReadUsdPerMTok;
-  if (
-    !isRate(input) ||
-    !isRate(output) ||
-    (cacheRead !== undefined && !isRate(cacheRead))
-  ) {
-    return undefined;
-  }
-  const cacheWrite = row.cacheWriteUsdPerMTok;
-  return {
-    inputUsdPerMTok: input,
-    outputUsdPerMTok: output,
-    cacheReadUsdPerMTok: isRate(cacheRead) ? cacheRead : input,
-    cacheWriteUsdPerMTok: isRate(cacheWrite) ? cacheWrite : undefined,
-  };
-}
-
-function parsePricing(value: unknown): ModelPricing | undefined {
-  const base = parseTier(value);
-  if (base === undefined) {
-    return undefined;
-  }
-  const fast = parseTier((value as Record<string, unknown>).fast);
-  return { ...base, fast };
-}
-
-function isRate(value: unknown): value is number {
-  return typeof value === "number" && Number.isFinite(value) && value >= 0;
-}
-
-export function normalizeModelId(id: string): string {
-  return id.trim().toLowerCase();
-}
-
-function modelId(model: ModelSetting): string {
-  return typeof model === "string" ? model : model.id;
-}
-
-function modelHasFastParam(model: ModelSetting): boolean {
-  if (typeof model === "string") {
-    return false;
-  }
-  return (
-    model.params?.some(
-      (param) => param.id === "fast" && param.value === "true"
-    ) === true
-  );
-}
-
-/**
- * Rates for a model selection, honoring its `fast` parameter. `undefined`
- * when the model is not in the table (no estimate is better than a wrong
- * one).
- */
-export function resolveModelPricing(
-  model: ModelSetting
-): ModelPricingTier | undefined {
-  const overrides = envOverrideTable();
-  const rawId = normalizeModelId(modelId(model));
-  const canonicalId = MODEL_ID_ALIASES[rawId] ?? rawId;
-  const pricing =
-    overrides[rawId] ??
-    overrides[canonicalId] ??
-    DEFAULT_MODEL_PRICING[rawId] ??
-    DEFAULT_MODEL_PRICING[canonicalId];
-  if (pricing === undefined) {
-    return undefined;
-  }
-  if (modelHasFastParam(model) && pricing.fast !== undefined) {
-    return pricing.fast;
-  }
-  return pricing;
-}
-
-/**
- * Estimate the dollar cost of one turn's token usage. `undefined` when the
- * model has no known rates.
- */
-export function estimateTurnCost(
-  model: ModelSetting,
-  usage: TurnUsage
-): TurnCost | undefined {
-  const tier = resolveModelPricing(model);
-  if (tier === undefined) {
-    return undefined;
-  }
-  const cacheWriteRate = tier.cacheWriteUsdPerMTok ?? tier.inputUsdPerMTok;
-  const inputUsd = (usage.inputTokens / M) * tier.inputUsdPerMTok;
-  const outputUsd = (usage.outputTokens / M) * tier.outputUsdPerMTok;
-  const cacheReadUsd = (usage.cacheReadTokens / M) * tier.cacheReadUsdPerMTok;
-  const cacheWriteUsd = (usage.cacheWriteTokens / M) * cacheWriteRate;
-  return {
-    totalUsd: roundUsd(inputUsd + outputUsd + cacheReadUsd + cacheWriteUsd),
-    inputUsd: roundUsd(inputUsd),
-    outputUsd: roundUsd(outputUsd),
-    cacheReadUsd: roundUsd(cacheReadUsd),
-    cacheWriteUsd: roundUsd(cacheWriteUsd),
-    model: normalizeModelId(modelId(model)),
-    source: "estimate",
-  };
-}
-
-/** Keep persisted estimates readable: micro-dollar precision. */
-function roundUsd(usd: number): number {
-  return Math.round(usd * 1e6) / 1e6;
-}