@blockrun/llm 3.24.3 → 3.24.5

package/README.md

@@ -987,6 +987,7 @@ token price.
 
 | Model | Context |
 |---|---|
+| `google/gemini-4-argon` | 1M |
 | `google/gemini-3.1-pro` | 1M |
 | `google/gemini-3.8-flash` | 1M |
 | `google/gemini-3.6-flash` | 1M |

package/dist/index.cjs

@@ -571,7 +571,7 @@ function withCost(body, response, bookedUsd, scheme) {
 // src/client.ts
 var import_accounts2 = require("viem/accounts");
 
-// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+2cd5a1026f755_easf3bov3hzdfoup33nwewqudy/node_modules/@blockrun/router-core/dist/index.js
+// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+8773a9954b091_rjdkm4uoakz3dfyojiwn3sv45i/node_modules/@blockrun/router-core/dist/index.js
 var UNSEGMENTED_SCRIPT = /[؀-ۿ　-ヿ㐀-䶿一-鿿가-힯豈-﫿＀-￯]/;
 var WORD_CHAR = /[\p{L}\p{N}]/u;
 var ASCII_PHRASE = /^[a-z0-9' -]+$/;
@@ -1213,10 +1213,9 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsVision: false
   },
   "deepseek/deepseek-v4.1-flash": {
-    // supportsTools: not probed — fails closed
     contextWindow: 1048576,
     maxOutputTokens: 65536,
-    supportsTools: false,
+    supportsTools: true,
     supportsVision: false
   },
   "google/gemini-2.5-flash": {
@@ -1279,6 +1278,13 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "google/gemini-4-argon": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1048576,
+    maxOutputTokens: 65536,
+    supportsTools: false,
+    supportsVision: true
+  },
   "minimax/minimax-m2.7": {
     contextWindow: 204800,
     maxOutputTokens: 16384,
@@ -1500,10 +1506,9 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsVision: true
   },
   "openai/gpt-6.1-sol": {
-    // supportsTools: not probed — fails closed
     contextWindow: 105e4,
     maxOutputTokens: 128e3,
-    supportsTools: false,
+    supportsTools: true,
     supportsVision: true
   },
   "openai/o1": {
@@ -1758,6 +1763,14 @@ var model_profiles_generated_default = {
     errorRate: 0,
     samples: 3
   },
+  "deepseek/deepseek-v4.1-flash": {
+    measuredAt: "2026-10-20T18:12:51Z",
+    latencyMs: 2383.5,
+    p95LatencyMs: 2547.8,
+    outputTokensPerSecond: 216.53,
+    errorRate: 0,
+    samples: 3
+  },
   "google/gemini-2.5-flash": {
     measuredAt: "2026-08-29T16:51:33Z",
     latencyMs: 5416.4,
@@ -2046,6 +2059,14 @@ var model_profiles_generated_default = {
     errorRate: 0,
     samples: 3
   },
+  "openai/gpt-6.1-sol": {
+    measuredAt: "2026-10-20T18:12:51Z",
+    latencyMs: 11324,
+    p95LatencyMs: 12747.6,
+    outputTokensPerSecond: 45.11,
+    errorRate: 0,
+    samples: 3
+  },
   "openai/o1": {
     measuredAt: "2026-08-29T16:51:33Z",
     latencyMs: 4324.9,
@@ -2861,11 +2882,13 @@ var VERIFIED_TOOL_FALLBACKS = [
   "anthropic/claude-fable-5.1",
   "anthropic/claude-opus-5.5",
   "anthropic/claude-sonnet-5.5",
+  "deepseek/deepseek-v4.1-flash",
   "google/gemini-3.8-flash",
   "openai/gpt-5.1",
   "openai/gpt-6-astra",
   "openai/gpt-6-luna",
   "openai/gpt-6-sol",
+  "openai/gpt-6.1-sol",
   "xai/grok-4.6",
   "xai/grok-4.7",
   "xiaomi/mimo-v2.5"
@@ -5066,7 +5089,7 @@ function getCostSummary() {
 }
 
 // src/version.ts
-var SDK_VERSION = "3.24.3";
+var SDK_VERSION = "3.24.5";
 var USER_AGENT = `blockrun-ts/${SDK_VERSION}`;
 
 // src/client.ts
@@ -5134,8 +5157,15 @@ var LLMClient = class _LLMClient {
   modelPricingCache = null;
   modelPricingPromise = null;
   // Pre-auth cache: avoids the 402 round-trip on repeat requests to the same model.
-  // Key = "endpoint:model", value = cached payment header + timestamp.
+  // Key = "endpoint:model", value = cached payment header + timestamp + the
+  // length of the request body it was quoted for.
   // TTL: 1 hour — pre-auth quotes are stable server-side on that horizon.
+  //
+  // A quote prices ONE request (input size + max_tokens), and exact settles
+  // the whole signed amount, so reusing it for a smaller request overpaid and
+  // for a larger one underpaid. The gateway refuses both with a 402 (an
+  // over-signed payment since blockrun#874), so a reuse is only attempted for
+  // a body the same size as the one that was quoted.
   preAuthCache = /* @__PURE__ */ new Map();
   static PRE_AUTH_TTL_MS = 36e5;
   /**
@@ -5742,9 +5772,10 @@ var LLMClient = class _LLMClient {
     if (options?.responseFormat !== void 0) body.response_format = options.responseFormat;
     if (options?.stop !== void 0) body.stop = options.stop;
     const cacheKey2 = `/v1/chat/completions:${requestModel}`;
+    const requestBody = JSON.stringify(body);
     const cached = this.preAuthCache.get(cacheKey2);
     const now = Date.now();
-    if (cached && now - cached.cachedAt < _LLMClient.PRE_AUTH_TTL_MS) {
+    if (cached && now - cached.cachedAt < _LLMClient.PRE_AUTH_TTL_MS && cached.bodyLength === requestBody.length) {
       let signed;
       try {
         signed = await this.signPayment(cached.paymentHeader);
@@ -5813,7 +5844,7 @@ var LLMClient = class _LLMClient {
       }
     }
     if (!paymentHeader) throw new PaymentError("402 response but no payment requirements found");
-    this.preAuthCache.set(cacheKey2, { paymentHeader, cachedAt: now });
+    this.preAuthCache.set(cacheKey2, { paymentHeader, cachedAt: now, bodyLength: requestBody.length });
     const { paymentPayload, costUsd } = await this.signPayment(paymentHeader);
     const streamResp = await this.fetchWithTimeout(url, {
       method: "POST",

package/dist/index.js

@@ -440,7 +440,7 @@ function withCost(body, response, bookedUsd, scheme) {
 // src/client.ts
 import { privateKeyToAccount } from "viem/accounts";
 
-// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+2cd5a1026f755_easf3bov3hzdfoup33nwewqudy/node_modules/@blockrun/router-core/dist/index.js
+// node_modules/.pnpm/@blockrun+router-core@https+++codeload.github.com+BlockRunAI+router-core+tar.gz+8773a9954b091_rjdkm4uoakz3dfyojiwn3sv45i/node_modules/@blockrun/router-core/dist/index.js
 var UNSEGMENTED_SCRIPT = /[؀-ۿ　-ヿ㐀-䶿一-鿿가-힯豈-﫿＀-￯]/;
 var WORD_CHAR = /[\p{L}\p{N}]/u;
 var ASCII_PHRASE = /^[a-z0-9' -]+$/;
@@ -1082,10 +1082,9 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsVision: false
   },
   "deepseek/deepseek-v4.1-flash": {
-    // supportsTools: not probed — fails closed
     contextWindow: 1048576,
     maxOutputTokens: 65536,
-    supportsTools: false,
+    supportsTools: true,
     supportsVision: false
   },
   "google/gemini-2.5-flash": {
@@ -1148,6 +1147,13 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsTools: true,
     supportsVision: true
   },
+  "google/gemini-4-argon": {
+    // supportsTools: not probed — fails closed
+    contextWindow: 1048576,
+    maxOutputTokens: 65536,
+    supportsTools: false,
+    supportsVision: true
+  },
   "minimax/minimax-m2.7": {
     contextWindow: 204800,
     maxOutputTokens: 16384,
@@ -1369,10 +1375,9 @@ var DEFAULT_MODEL_CAPABILITIES = Object.freeze({
     supportsVision: true
   },
   "openai/gpt-6.1-sol": {
-    // supportsTools: not probed — fails closed
     contextWindow: 105e4,
     maxOutputTokens: 128e3,
-    supportsTools: false,
+    supportsTools: true,
     supportsVision: true
   },
   "openai/o1": {
@@ -1627,6 +1632,14 @@ var model_profiles_generated_default = {
     errorRate: 0,
     samples: 3
   },
+  "deepseek/deepseek-v4.1-flash": {
+    measuredAt: "2026-10-20T18:12:51Z",
+    latencyMs: 2383.5,
+    p95LatencyMs: 2547.8,
+    outputTokensPerSecond: 216.53,
+    errorRate: 0,
+    samples: 3
+  },
   "google/gemini-2.5-flash": {
     measuredAt: "2026-08-29T16:51:33Z",
     latencyMs: 5416.4,
@@ -1915,6 +1928,14 @@ var model_profiles_generated_default = {
     errorRate: 0,
     samples: 3
   },
+  "openai/gpt-6.1-sol": {
+    measuredAt: "2026-10-20T18:12:51Z",
+    latencyMs: 11324,
+    p95LatencyMs: 12747.6,
+    outputTokensPerSecond: 45.11,
+    errorRate: 0,
+    samples: 3
+  },
   "openai/o1": {
     measuredAt: "2026-08-29T16:51:33Z",
     latencyMs: 4324.9,
@@ -2730,11 +2751,13 @@ var VERIFIED_TOOL_FALLBACKS = [
   "anthropic/claude-fable-5.1",
   "anthropic/claude-opus-5.5",
   "anthropic/claude-sonnet-5.5",
+  "deepseek/deepseek-v4.1-flash",
   "google/gemini-3.8-flash",
   "openai/gpt-5.1",
   "openai/gpt-6-astra",
   "openai/gpt-6-luna",
   "openai/gpt-6-sol",
+  "openai/gpt-6.1-sol",
   "xai/grok-4.6",
   "xai/grok-4.7",
   "xiaomi/mimo-v2.5"
@@ -4935,7 +4958,7 @@ function getCostSummary() {
 }
 
 // src/version.ts
-var SDK_VERSION = "3.24.3";
+var SDK_VERSION = "3.24.5";
 var USER_AGENT = `blockrun-ts/${SDK_VERSION}`;
 
 // src/client.ts
@@ -5003,8 +5026,15 @@ var LLMClient = class _LLMClient {
   modelPricingCache = null;
   modelPricingPromise = null;
   // Pre-auth cache: avoids the 402 round-trip on repeat requests to the same model.
-  // Key = "endpoint:model", value = cached payment header + timestamp.
+  // Key = "endpoint:model", value = cached payment header + timestamp + the
+  // length of the request body it was quoted for.
   // TTL: 1 hour — pre-auth quotes are stable server-side on that horizon.
+  //
+  // A quote prices ONE request (input size + max_tokens), and exact settles
+  // the whole signed amount, so reusing it for a smaller request overpaid and
+  // for a larger one underpaid. The gateway refuses both with a 402 (an
+  // over-signed payment since blockrun#874), so a reuse is only attempted for
+  // a body the same size as the one that was quoted.
   preAuthCache = /* @__PURE__ */ new Map();
   static PRE_AUTH_TTL_MS = 36e5;
   /**
@@ -5611,9 +5641,10 @@ var LLMClient = class _LLMClient {
     if (options?.responseFormat !== void 0) body.response_format = options.responseFormat;
     if (options?.stop !== void 0) body.stop = options.stop;
     const cacheKey2 = `/v1/chat/completions:${requestModel}`;
+    const requestBody = JSON.stringify(body);
     const cached = this.preAuthCache.get(cacheKey2);
     const now = Date.now();
-    if (cached && now - cached.cachedAt < _LLMClient.PRE_AUTH_TTL_MS) {
+    if (cached && now - cached.cachedAt < _LLMClient.PRE_AUTH_TTL_MS && cached.bodyLength === requestBody.length) {
       let signed;
       try {
         signed = await this.signPayment(cached.paymentHeader);
@@ -5682,7 +5713,7 @@ var LLMClient = class _LLMClient {
       }
     }
     if (!paymentHeader) throw new PaymentError("402 response but no payment requirements found");
-    this.preAuthCache.set(cacheKey2, { paymentHeader, cachedAt: now });
+    this.preAuthCache.set(cacheKey2, { paymentHeader, cachedAt: now, bodyLength: requestBody.length });
     const { paymentPayload, costUsd } = await this.signPayment(paymentHeader);
     const streamResp = await this.fetchWithTimeout(url, {
       method: "POST",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blockrun/llm",
-  "version": "3.24.3",
+  "version": "3.24.5",
   "type": "module",
   "description": "TypeScript SDK for BlockRun - every frontier model behind one OpenAI-compatible client. Pay with a BlockRun API key or per-call in USDC. Smart routing picks the cheapest capable model. No rate limits.",
   "main": "dist/index.cjs",
@@ -58,7 +58,7 @@
   },
   "devDependencies": {
     "@blockrun/core": "^0.1.0",
-    "@blockrun/router-core": "https://codeload.github.com/BlockRunAI/router-core/tar.gz/2cd5a1026f755cd39a3f0d6e61cab23a5ad391a9",
+    "@blockrun/router-core": "https://codeload.github.com/BlockRunAI/router-core/tar.gz/8773a9954b091bd0090acdfec67c75a006c0e156",
     "@eslint/js": "^10.0.1",
     "@solana/kit": "^5.5.1",
     "@types/node": "^26.4.1",