llmbrake 0.2.1 → 0.2.2

package/dist/prices.d.ts

@@ -6,8 +6,8 @@
  *   - OpenAI:    https://developers.openai.com/api/docs/pricing
  *   - Anthropic: https://docs.anthropic.com/en/docs/about-claude/pricing
  * Anthropic IDs for the current generation were checked against the models overview page (2026-10-17).
- * Claude Haiku 5.5 is priced at its prompts-up-to-100K tier.
- * Long-context, batch, regional/data-residency uplifts and fast/priority tiers are NOT modeled.
+ * Claude Haiku 5.5: short tier by default; computeCostUsd auto-selects `claude-haiku-5-5:long` when prompt_tokens > 100K.
+ * Batch, regional/data-residency uplifts and fast/priority tiers are NOT modeled.
  * Override or extend via `createGuard({ prices: { ...DEFAULT_PRICES, ...yours } })`.
  */
 export interface ModelPrice {

package/dist/prices.js

@@ -18,16 +18,22 @@ export const DEFAULT_PRICES = {
     "claude-opus-5-5": { inputPerMTok: 4.0, cachedInputPerMTok: 0.2, outputPerMTok: 20.0, source: ANTHROPIC },
     "claude-sonnet-5-5": { inputPerMTok: 2.0, cachedInputPerMTok: 0.1, outputPerMTok: 10.0, source: ANTHROPIC },
     "claude-haiku-5-5": { inputPerMTok: 0.1, cachedInputPerMTok: 0.01, outputPerMTok: 0.5, source: ANTHROPIC },
+    // Prompts over 100K tokens (cache reads/writes count toward the threshold); whole request billed at this tier.
+    "claude-haiku-5-5:long": { inputPerMTok: 0.5, cachedInputPerMTok: 0.05, outputPerMTok: 2.5, source: ANTHROPIC },
     "claude-opus-4-8": { inputPerMTok: 5.0, cachedInputPerMTok: 0.5, outputPerMTok: 25.0, source: ANTHROPIC },
     "claude-sonnet-5": { inputPerMTok: 2.0, cachedInputPerMTok: 0.2, outputPerMTok: 10.0, source: ANTHROPIC },
     "claude-haiku-4-5": { inputPerMTok: 1.0, cachedInputPerMTok: 0.1, outputPerMTok: 5.0, source: ANTHROPIC },
 };
 /** Returns USD cost, or null if the model is not in the table. */
 export function computeCostUsd(model, usage, prices) {
-    const p = lookupPrice(model, prices);
+    const prompt = usage.prompt_tokens ?? 0;
+    // Haiku 5.5 bills the whole request at the >100K tier once total input crosses 100K.
+    const pricedModel = (model === "claude-haiku-5-5" || model.startsWith("claude-haiku-5-5-")) && prompt > 100_000
+        ? "claude-haiku-5-5:long"
+        : model;
+    const p = lookupPrice(pricedModel, prices);
     if (!p)
         return null;
-    const prompt = usage.prompt_tokens ?? 0;
     const cached = Math.min(usage.prompt_tokens_details?.cached_tokens ?? 0, prompt);
     const uncached = prompt - cached;
     const out = usage.completion_tokens ?? 0;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llmbrake",
-  "version": "0.2.1",
+  "version": "0.2.2",
   "description": "Per-customer monthly USD budgets for OpenAI-compatible chat calls: track cost, auto-downgrade near the cap, block at the cap.",
   "license": "MIT",
   "author": "LLMBrake contributors",