@axlsdk/axl 0.25.2 → 0.25.3

package/dist/index.d.cts

@@ -7378,8 +7378,8 @@ type DirectOpenAIPricingContext = {
     };
 };
 /**
- * Estimate OpenAI call cost from token usage. Exact match first, then
- * longest-prefix match for versioned models (e.g. `gpt-4o-2024-05-13`).
+ * Estimate OpenAI call cost from token usage using exact catalog IDs only.
+ * Explicitly listed snapshots (e.g. `gpt-4o-2024-05-13`) have their own rows.
  *
  * Returns `undefined` when the model is not in the table — callers must treat
  * that as "unknown cost", never as free (a silent `0` would break
@@ -7409,8 +7409,8 @@ declare class OpenAIProvider extends OpenAICompatibleProvider {
         dangerouslyAllowInsecureHttp?: boolean;
         rateLimit?: RateLimitConfig;
     });
-    /** Report a clamped `effort` — Chat Completions caps `'max'` at `'xhigh'` and
-     *  rejects `'none'` before gpt-5.1 (clamped to `'minimal'`). */
+    /** Report a clamped `effort` — Chat caps `'max'` at `'xhigh'` on GPT-5.6 and
+     *  GPT-6.1 Sol, and maps `'none'` to the model's supported floor. */
     effortResolution(options: Pick<ChatOptions, 'model' | 'effort' | 'thinkingBudget' | 'includeThoughts' | 'providerOptions'>): EffortResolution | undefined;
     protected validateFinalRequestBody(body: Record<string, unknown>): void;
     protected computeCost(model: string, usage: ProviderResponse['usage'], _reportedCost: number | undefined, context?: DirectOpenAIPricingContext): number | undefined;

package/dist/index.d.ts

@@ -7378,8 +7378,8 @@ type DirectOpenAIPricingContext = {
     };
 };
 /**
- * Estimate OpenAI call cost from token usage. Exact match first, then
- * longest-prefix match for versioned models (e.g. `gpt-4o-2024-05-13`).
+ * Estimate OpenAI call cost from token usage using exact catalog IDs only.
+ * Explicitly listed snapshots (e.g. `gpt-4o-2024-05-13`) have their own rows.
  *
  * Returns `undefined` when the model is not in the table — callers must treat
  * that as "unknown cost", never as free (a silent `0` would break
@@ -7409,8 +7409,8 @@ declare class OpenAIProvider extends OpenAICompatibleProvider {
         dangerouslyAllowInsecureHttp?: boolean;
         rateLimit?: RateLimitConfig;
     });
-    /** Report a clamped `effort` — Chat Completions caps `'max'` at `'xhigh'` and
-     *  rejects `'none'` before gpt-5.1 (clamped to `'minimal'`). */
+    /** Report a clamped `effort` — Chat caps `'max'` at `'xhigh'` on GPT-5.6 and
+     *  GPT-6.1 Sol, and maps `'none'` to the model's supported floor. */
     effortResolution(options: Pick<ChatOptions, 'model' | 'effort' | 'thinkingBudget' | 'includeThoughts' | 'providerOptions'>): EffortResolution | undefined;
     protected validateFinalRequestBody(body: Record<string, unknown>): void;
     protected computeCost(model: string, usage: ProviderResponse['usage'], _reportedCost: number | undefined, context?: DirectOpenAIPricingContext): number | undefined;

package/dist/index.js

@@ -348,7 +348,7 @@ var sharedGuard = typeof installedGuard === "object" && installedGuard !== null
 var buildPath = typeof __AXL_BUILD_PATH__ === "string" ? __AXL_BUILD_PATH__ : "<source>";
 var thisCopy = {
   path: buildPath.startsWith("file:") ? fileURLToPath(buildPath) : buildPath,
-  version: true ? "0.25.2" : "unknown"
+  version: true ? "0.25.3" : "unknown"
 };
 function warnCrossCopy(other, joined) {
   if (other === thisCopy) return;
@@ -3617,7 +3617,7 @@ var OPENAI_PRICING = {
   "gpt-4.1-mini": [4e-7, 16e-7, 0.25],
   "gpt-4.1-nano": [1e-7, 4e-7, 0.25],
   o3: [2e-6, 8e-6, 0.25],
-  "o3-mini": [11e-7, 44e-7, 0.25],
+  "o3-mini": [11e-7, 44e-7, 0.5],
   "o4-mini": [11e-7, 44e-7, 0.25],
   "gpt-5": [125e-8, 1e-5, 0.1],
   "gpt-5-mini": [25e-8, 2e-6, 0.1],
@@ -3640,6 +3640,13 @@ var DIRECT_OPENAI_CATALOG = [
     long: { input: 20 / M, cachedInput: 2 / M, cacheWrite: 25 / M, output: 75 / M },
     contextBoundary: LONG_CONTEXT_BOUNDARY
   },
+  {
+    aliases: ["gpt-6.1-sol"],
+    // Rechecked 2026-10-15 against https://developers.openai.com/api/docs/pricing.
+    short: { input: 2 / M, cachedInput: 0.1 / M, cacheWrite: 2.5 / M, output: 10 / M },
+    long: { input: 4 / M, cachedInput: 0.2 / M, cacheWrite: 5 / M, output: 15 / M },
+    contextBoundary: LONG_CONTEXT_BOUNDARY
+  },
   {
     aliases: ["gpt-6-sol"],
     short: { input: 2 / M, cachedInput: 0.2 / M, cacheWrite: 2.5 / M, output: 10 / M },
@@ -4019,12 +4026,12 @@ var BASE_EFFORTS = ["minimal", "low", "medium", "high"];
 function bothEndpoints(value) {
   return { chat: value, responses: value };
 }
-function gpt6(id, efforts, chatTools) {
+function gpt6(id, responsesEfforts, chatTools, chatEfforts = responsesEfforts) {
   return {
     id,
     family: "gpt-6",
     reasoning: true,
-    efforts: bothEndpoints(efforts),
+    efforts: { chat: chatEfforts, responses: responsesEfforts },
     defaultEffort: "medium",
     // Same policy as GPT-5.x: a portable `temperature` is dropped while
     // reasoning is active (including the active default). A raw
@@ -4052,6 +4059,13 @@ function gpt56(id) {
 var OPENAI_EXACT_MODELS = new Map(
   [
     gpt6("gpt-6-astra", ["low", "medium", "high", "xhigh", "max"], "never"),
+    // Live Chat Completions rejects max; Responses accepts it (2026-10-15).
+    gpt6("gpt-6.1-sol", ["low", "medium", "high", "xhigh", "max"], "never", [
+      "low",
+      "medium",
+      "high",
+      "xhigh"
+    ]),
     gpt6("gpt-6-sol", ["none", "low", "medium", "high", "xhigh", "max"], "effort-none-only"),
     gpt6("gpt-6-luna", ["none", "low", "medium", "high", "xhigh", "max"], "effort-none-only"),
     gpt56("gpt-5.6"),
@@ -4195,15 +4209,15 @@ function validateOpenAIFinalRequestBody(body, endpoint) {
     }
   }
   if (activeReasoning) {
+    const remediation = (key) => descriptor.efforts[endpoint].includes("none") ? `Remove ${key} or set reasoning effort to 'none'.` : `Remove ${key}; ${model} does not support reasoning effort 'none'.`;
     const forbidden = endpoint === "chat" ? ["temperature", "top_p", "top_logprobs", "logprobs"] : ["temperature", "top_p", "top_logprobs"];
     for (const key of forbidden) {
-      if (body[key] !== void 0)
-        fail(key, `Remove ${key} or set reasoning effort to 'none' on Sol or Luna.`);
+      if (body[key] !== void 0) fail(key, remediation(key));
     }
     if (endpoint === "responses" && Array.isArray(body.include) && body.include.includes("message.output_text.logprobs")) {
       fail(
         "message.output_text.logprobs",
-        "Remove it from include or set reasoning effort to none on Sol or Luna."
+        remediation("message.output_text.logprobs from include")
       );
     }
   }
@@ -4256,8 +4270,8 @@ var OpenAIProvider = class extends OpenAICompatibleProvider {
   constructor(options = {}) {
     super({ profile: OPENAI_PROFILE, ...options });
   }
-  /** Report a clamped `effort` — Chat Completions caps `'max'` at `'xhigh'` and
-   *  rejects `'none'` before gpt-5.1 (clamped to `'minimal'`). */
+  /** Report a clamped `effort` — Chat caps `'max'` at `'xhigh'` on GPT-5.6 and
+   *  GPT-6.1 Sol, and maps `'none'` to the model's supported floor. */
   effortResolution(options) {
     return resolveOpenAIEffortResolution(
       options,