@blockrun/franklin 3.25.4 → 3.26.1

package/dist/agent/llm.d.ts

@@ -161,6 +161,7 @@ export declare class ModelClient {
         stopReason: string;
     }>;
     private signPayment;
+    private recordSettledPayment;
     private signBasePayment;
     private signSolanaPayment;
     private extractPaymentReq;

package/dist/agent/llm.js

@@ -591,17 +591,18 @@ export class ModelClient {
             if (response.status === 402) {
                 if (this.debug)
                     console.error('[franklin] Payment required — signing...');
-                const paymentHeader = await this.signPayment(response, request.model);
-                if (!paymentHeader) {
+                const signedPayment = await this.signPayment(response, request.model);
+                if (!signedPayment) {
                     yield { kind: 'error', payload: { message: 'Payment signing failed' } };
                     return;
                 }
                 response = await withAbortableTimeout(() => fetchWithUnwrappedCause(endpoint, {
                     method: 'POST',
-                    headers: { ...headers, ...paymentHeader },
+                    headers: { ...headers, ...signedPayment.headers },
                     body,
                     signal: requestController.signal,
                 }), requestController, createModelTimeoutError('request', request.model, requestTimeoutMs), requestTimeoutMs);
+                this.recordSettledPayment(signedPayment, response);
             }
             if (!response.ok) {
                 const errorBody = await response.text().catch(() => 'unknown error');
@@ -653,17 +654,18 @@ export class ModelClient {
                         signal: requestController.signal,
                     }), requestController, createModelTimeoutError('request', request.model, requestTimeoutMs), requestTimeoutMs);
                     if (response.status === 402) {
-                        const paymentHeader = await this.signPayment(response, request.model);
-                        if (!paymentHeader) {
+                        const signedPayment = await this.signPayment(response, request.model);
+                        if (!signedPayment) {
                             yield { kind: 'error', payload: { message: 'Payment signing failed' } };
                             return;
                         }
                         response = await withAbortableTimeout(() => fetchWithUnwrappedCause(endpoint, {
                             method: 'POST',
-                            headers: { ...headers, ...paymentHeader },
+                            headers: { ...headers, ...signedPayment.headers },
                             body: retryBody,
                             signal: requestController.signal,
                         }), requestController, createModelTimeoutError('request', request.model, requestTimeoutMs), requestTimeoutMs);
+                        this.recordSettledPayment(signedPayment, response);
                     }
                     if (!response.ok) {
                         const retryBodyText = await response.text().catch(() => 'unknown error');
@@ -1101,6 +1103,15 @@ export class ModelClient {
             return null;
         }
     }
+    recordSettledPayment(payment, response) {
+        // A post-signature 402 means the gateway rejected the payment rather than
+        // settling it. Keep both the session cost and cost_log anchored to calls
+        // the gateway accepted after the paid retry returned.
+        if (response.status === 402)
+            return;
+        this.lastPaidUsd += payment.amountUsd;
+        appendSettlementRow(payment.endpoint, payment.amountUsd, payment.meta);
+    }
     async signBasePayment(response, model) {
         // Refresh wallet cache after TTL to pick up balance/key changes
         if (!this.cachedBaseWallet || (Date.now() - this.walletCacheTime > ModelClient.WALLET_CACHE_TTL)) {
@@ -1116,25 +1127,24 @@ export class ModelClient {
             throw new Error('No payment requirements in 402 response');
         const paymentRequired = parsePaymentRequired(paymentHeader);
         const details = extractPaymentDetails(paymentRequired);
-        this.lastPaidUsd = paymentAmountToUsd(details.amount);
-        // Mirror the SDK's appendCostLog write so cost_log.jsonl becomes a
-        // true wallet-truth ledger covering both SDK helper traffic AND the
-        // agent's main LLM stream (which uses this signer, not the SDK).
-        // Match SDK schema (model/wallet/network/client_kind) so every row
-        // is independently queryable.
-        appendSettlementRow('/v1/messages', this.lastPaidUsd, {
-            model,
-            wallet: wallet.address,
-            network: details.network || 'base-mainnet',
-            client_kind: 'AgentClient',
-        });
+        const amountUsd = paymentAmountToUsd(details.amount);
         const payload = await createPaymentPayload(wallet.privateKey, wallet.address, details.recipient, details.amount, details.network || 'eip155:8453', {
             resourceUrl: details.resource?.url || this.apiUrl,
             resourceDescription: details.resource?.description || 'BlockRun AI API call',
             maxTimeoutSeconds: details.maxTimeoutSeconds || 300,
             extra: details.extra,
         });
-        return { 'PAYMENT-SIGNATURE': payload };
+        return {
+            headers: { 'PAYMENT-SIGNATURE': payload },
+            endpoint: '/v1/messages',
+            amountUsd,
+            meta: {
+                model,
+                wallet: wallet.address,
+                network: details.network || 'base-mainnet',
+                client_kind: 'AgentClient',
+            },
+        };
     }
     async signSolanaPayment(response, model) {
         if (!this.cachedSolanaWallet || (Date.now() - this.walletCacheTime > ModelClient.WALLET_CACHE_TTL)) {
@@ -1149,13 +1159,7 @@ export class ModelClient {
             throw new Error('No payment requirements in 402 response');
         const paymentRequired = parsePaymentRequired(paymentHeader);
         const details = extractPaymentDetails(paymentRequired, SOLANA_NETWORK);
-        this.lastPaidUsd = paymentAmountToUsd(details.amount);
-        appendSettlementRow('/v1/messages', this.lastPaidUsd, {
-            model,
-            wallet: wallet.address,
-            network: details.network || 'solana-mainnet',
-            client_kind: 'AgentClient',
-        });
+        const amountUsd = paymentAmountToUsd(details.amount);
         const secretBytes = await solanaKeyToBytes(wallet.privateKey);
         const feePayer = details.extra?.feePayer || details.recipient;
         const payload = await createSolanaPaymentPayload(secretBytes, wallet.address, details.recipient, details.amount, feePayer, {
@@ -1164,7 +1168,17 @@ export class ModelClient {
             maxTimeoutSeconds: details.maxTimeoutSeconds || 300,
             extra: details.extra,
         });
-        return { 'PAYMENT-SIGNATURE': payload };
+        return {
+            headers: { 'PAYMENT-SIGNATURE': payload },
+            endpoint: '/v1/messages',
+            amountUsd,
+            meta: {
+                model,
+                wallet: wallet.address,
+                network: details.network || 'solana-mainnet',
+                client_kind: 'AgentClient',
+            },
+        };
     }
     async extractPaymentReq(response) {
         let header = response.headers.get('payment-required');

package/dist/pricing.js

@@ -64,17 +64,19 @@ export const MODEL_PRICING = {
     'xai/grok-4-fast-reasoning': { input: 0.2, output: 0.5 },
     'xai/grok-4-1-fast': { input: 0.2, output: 0.5 },
     'xai/grok-4-1-fast-reasoning': { input: 0.2, output: 0.5 },
-    'xai/grok-4-0709': { input: 0.2, output: 1.5 },
+    'xai/grok-4-0709': { input: 3.0, output: 15.0 }, // gateway lists $3/$15 (was mispriced here at $0.2/$1.5)
     'xai/grok-3-mini': { input: 0.3, output: 0.5 },
     'xai/grok-2-vision': { input: 2.0, output: 10.0 },
     'xai/grok-3': { input: 3.0, output: 15.0 },
+    'xai/grok-4.3': { input: 1.5, output: 4.0 }, // public flagship 2026-06-04, OpenRouter resale
+    'xai/grok-build-0.1': { input: 1.5, output: 3.0 }, // agentic coding, OpenRouter resale
     // DeepSeek (gateway re-aliased these to V4 Flash on 2026-05-03; price
     // dropped from $0.28/$0.42 to $0.20/$0.40, context bumped 128K→1M).
     'deepseek/deepseek-chat': { input: 0.20, output: 0.40 },
     'deepseek/deepseek-reasoner': { input: 0.20, output: 0.40 },
     // V4 Pro (1.6T MoE / 49B active, 1M ctx, 65K out). 75% launch promo
     // through 2026-05-31 — list is $2.00/$4.00, promo is $0.50/$1.00.
-    'deepseek/deepseek-v4-pro': { input: 0.50, output: 1.00 },
+    'deepseek/deepseek-v4-pro': { input: 0.435, output: 0.87 }, // 75% promo became permanent list after 2026-05-31
     // Minimax
     'minimax/minimax-m3': { input: 0.3, output: 1.2 },
     'minimax/minimax-m2.7': { input: 0.3, output: 1.2 },
@@ -86,10 +88,10 @@ export const MODEL_PRICING = {
     'moonshot/kimi-k2.5': { input: 0.6, output: 3.0 },
     'nvidia/kimi-k2.5': { input: 0.55, output: 2.5 },
     // PROMOTION (active ~2026-04): flat $0.001/call for all GLM models
-    'zai/glm-5': { input: 0, output: 0, perCall: 0.001 },
-    'zai/glm-5.1': { input: 0, output: 0, perCall: 0.001 },
-    'zai/glm-5-turbo': { input: 0, output: 0, perCall: 0.001 },
-    'zai/glm-5.1-turbo': { input: 0, output: 0, perCall: 0.001 }, // client alias for zai/glm-5-turbo
+    'zai/glm-5': { input: 0.6, output: 1.92 }, // flat promo ended 2026-06-06 — per-token now
+    'zai/glm-5.1': { input: 1.40, output: 4.40 }, // launch promo ended 2026-06-05 — per-token now
+    'zai/glm-5-turbo': { input: 1.2, output: 4.0 }, // flat promo ended 2026-06-06 — per-token now
+    'zai/glm-5.1-turbo': { input: 1.2, output: 4.0 }, // client alias for zai/glm-5-turbo
 };
 /** Opus pricing for savings calculations — tracks the current flagship. */
 export const OPUS_PRICING = MODEL_PRICING['anthropic/claude-opus-4.8'];

package/dist/proxy/server.js

@@ -767,12 +767,13 @@ async function handleBasePayment(response, url, method, headers, body, privateKe
     const paymentRequired = parsePaymentRequired(paymentHeader);
     const details = extractPaymentDetails(paymentRequired);
     const paidUsd = paymentAmountToUsd(details.amount);
-    appendSettlementRow(extractEndpointPath(url), paidUsd, {
+    const endpoint = extractEndpointPath(url);
+    const settlementMeta = {
         model,
         wallet: fromAddress,
         network: details.network || 'base-mainnet',
         client_kind: 'ProxyClient',
-    });
+    };
     const paymentPayload = await createPaymentPayload(privateKey, fromAddress, details.recipient, details.amount, details.network || 'eip155:8453', {
         resourceUrl: details.resource?.url || url,
         resourceDescription: details.resource?.description || 'BlockRun AI API call',
@@ -787,6 +788,9 @@ async function handleBasePayment(response, url, method, headers, body, privateKe
         },
         body: body || undefined,
     }, timeoutMs, `Paid proxy request for ${model}`);
+    if (paid.status === 402)
+        return { response: paid, paidUsd: 0 };
+    appendSettlementRow(endpoint, paidUsd, settlementMeta);
     return { response: paid, paidUsd };
 }
 // ======================================================================
@@ -800,12 +804,13 @@ async function handleSolanaPayment(response, url, method, headers, body, private
     const paymentRequired = parsePaymentRequired(paymentHeader);
     const details = extractPaymentDetails(paymentRequired, SOLANA_NETWORK);
     const paidUsd = paymentAmountToUsd(details.amount);
-    appendSettlementRow(extractEndpointPath(url), paidUsd, {
+    const endpoint = extractEndpointPath(url);
+    const settlementMeta = {
         model,
         wallet: fromAddress,
         network: details.network || 'solana-mainnet',
         client_kind: 'ProxyClient',
-    });
+    };
     const secretKey = await solanaKeyToBytes(privateKey);
     const feePayer = details.extra?.feePayer || details.recipient;
     const paymentPayload = await createSolanaPaymentPayload(secretKey, fromAddress, details.recipient, details.amount, feePayer, {
@@ -822,6 +827,9 @@ async function handleSolanaPayment(response, url, method, headers, body, private
         },
         body: body || undefined,
     }, timeoutMs, `Paid proxy request for ${model}`);
+    if (paid.status === 402)
+        return { response: paid, paidUsd: 0 };
+    appendSettlementRow(endpoint, paidUsd, settlementMeta);
     return { response: paid, paidUsd };
 }
 /**

package/dist/router/index.js

@@ -36,7 +36,7 @@ function loadLearnedWeights() {
 }
 // ─── Tier Model Configs ───
 // Auto-routing strategy (post-DeepSeek-V4-Pro launch promo, 2026-05-03):
-// V4 Pro at $0.50/$1.00 with 1M context is the new sweet spot for SIMPLE +
+// V4 Pro at $0.435/$0.87 with 1M context is the new sweet spot for SIMPLE +
 // MEDIUM agent work — Sonnet-quality reasoning at ~1/6 the price. Reserve
 // Opus only for genuinely complex multi-file/multi-decision tasks where
 // the model's wider context handling and tighter tool-use discipline still
@@ -78,7 +78,7 @@ const AUTO_TIERS = {
  * if none of them have vision, escalates to COMPLEX (Opus is always vision).
  *
  * Note: only applied when the caller signals needsVision=true. Without that
- * hint the classic per-tier defaults still rule — V4 Pro's $0.50/$1.00 promo
+ * hint the classic per-tier defaults still rule — V4 Pro's $0.435/$0.87 price
  * is the right SIMPLE/MEDIUM pick for text-only turns and we don't want to
  * blanket-upgrade everyone to a vision model.
  */

package/dist/ui/model-picker.js

@@ -50,8 +50,11 @@ export const MODEL_SHORTCUTS = {
     flash: 'google/gemini-2.5-flash',
     'gemini-3': 'google/gemini-3.1-pro',
     'gemini-3.1': 'google/gemini-3.1-pro',
-    // xAI
-    grok: 'xai/grok-3',
+    // xAI — grok-4.3 is the public flagship since 2026-06-04 (grok-3 and the
+    // fast families are hidden on the gateway; explicit IDs still resolve).
+    grok: 'xai/grok-4.3',
+    'grok-4.3': 'xai/grok-4.3',
+    'grok-build': 'xai/grok-build-0.1',
     'grok-3': 'xai/grok-3',
     'grok-4': 'xai/grok-4-0709',
     'grok-fast': 'xai/grok-4-1-fast-reasoning',
@@ -95,6 +98,7 @@ export const MODEL_SHORTCUTS = {
     'm3': 'minimax/minimax-m3',
     'm2.7': 'minimax/minimax-m2.7',
     glm: 'zai/glm-5.1',
+    'glm-5': 'zai/glm-5',
     'glm-turbo': 'zai/glm-5-turbo',
     'glm5': 'zai/glm-5.1',
     kimi: 'moonshot/kimi-k2.6',
@@ -122,13 +126,6 @@ export function resolveModel(input) {
  * edit this one place.
  */
 export const PICKER_CATEGORIES = [
-    {
-        category: '🔥 Promo (flat $0.001/call)',
-        models: [
-            { id: 'zai/glm-5.1', shortcut: 'glm', label: 'GLM-5.1', price: '$0.001/call', highlight: true },
-            { id: 'zai/glm-5-turbo', shortcut: 'glm-turbo', label: 'GLM-5 Turbo', price: '$0.001/call', highlight: true },
-        ],
-    },
     {
         category: '🧠 Smart routing (auto-pick)',
         models: [
@@ -136,7 +133,7 @@ export const PICKER_CATEGORIES = [
             // Premium are kept as shortcut aliases (`eco`, `premium`) and resolve
             // through the router for back-compat with older configs/sessions, but
             // they're hidden from new users — Auto already covers the cheap end
-            // (V4 Pro at $0.50/$1.00 for SIMPLE/MEDIUM) and the quality end (Opus
+            // (V4 Pro at $0.435/$0.87 for SIMPLE/MEDIUM) and the quality end (Opus
             // for COMPLEX), so a separate Eco/Premium picker entry just adds
             // choice paralysis without distinct value.
             { id: 'blockrun/auto', shortcut: 'auto', label: 'Auto', price: 'routed' },
@@ -156,7 +153,7 @@ export const PICKER_CATEGORIES = [
             { id: 'openai/gpt-5.5', shortcut: 'gpt', label: 'GPT-5.5', price: '$5/$30', highlight: true },
             { id: 'google/gemini-3.1-pro', shortcut: 'gemini-3', label: 'Gemini 3.1 Pro', price: '$2/$12' },
             { id: 'google/gemini-2.5-pro', shortcut: 'gemini', label: 'Gemini 2.5 Pro', price: '$1.25/$10' },
-            { id: 'xai/grok-4-0709', shortcut: 'grok-4', label: 'Grok 4', price: '$0.2/$1.5' },
+            { id: 'xai/grok-4.3', shortcut: 'grok', label: 'Grok 4.3', price: '$1.5/$4' },
         ],
     },
     {
@@ -164,9 +161,10 @@ export const PICKER_CATEGORIES = [
         models: [
             { id: 'openai/o3', shortcut: 'o3', label: 'O3', price: '$2/$8' },
             { id: 'openai/gpt-5.3-codex', shortcut: 'codex', label: 'GPT-5.3 Codex', price: '$1.75/$14' },
-            // V4 Pro on launch promo (75% off through 2026-05-31). 1M context,
-            // 1.6T MoE → punches up to GPT-5.5/Opus on hard tasks at <1/10 the price.
-            { id: 'deepseek/deepseek-v4-pro', shortcut: 'deepseek-v4-pro', label: 'DeepSeek V4 Pro', price: '$0.5/$1 (promo)', highlight: true },
+            // V4 Pro: the 75% launch promo became DeepSeek's permanent list price
+            // after 2026-05-31. 1M context, 1.6T MoE → punches up to GPT-5.5/Opus
+            // on hard tasks at <1/10 the price.
+            { id: 'deepseek/deepseek-v4-pro', shortcut: 'deepseek-v4-pro', label: 'DeepSeek V4 Pro', price: '$0.435/$0.87', highlight: true },
             { id: 'deepseek/deepseek-reasoner', shortcut: 'r1', label: 'DeepSeek V4 Flash R.', price: '$0.2/$0.4' },
             { id: 'xai/grok-4-1-fast-reasoning', shortcut: 'grok-fast', label: 'Grok 4.1 Fast R.', price: '$0.2/$0.5' },
         ],
@@ -180,6 +178,9 @@ export const PICKER_CATEGORIES = [
             // Re-aliased to V4 Flash Chat upstream — context 1M, price 30% lower.
             { id: 'deepseek/deepseek-chat', shortcut: 'deepseek', label: 'DeepSeek V4 Flash Chat', price: '$0.2/$0.4' },
             { id: 'moonshot/kimi-k2.6', shortcut: 'kimi', label: 'Kimi K2.6', price: '$0.95/$4' },
+            // GLM flat-rate promos fully ended 2026-06-06 — whole family per-token
+            // now (glm-5 $0.60/$1.92; `glm` shortcut still pins flagship glm-5.1).
+            { id: 'zai/glm-5', shortcut: 'glm-5', label: 'GLM-5', price: '$0.6/$1.92' },
             // Minimax M2.7 hidden to make room for V4 Pro in Reasoning + V4 Flash
             // (free) without exceeding the picker's 24-entry cap. Shortcut `minimax`
             // still resolves to it.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blockrun/franklin",
-  "version": "3.25.4",
+  "version": "3.26.1",
   "description": "Franklin Agent — The AI agent with a wallet. Spends USDC autonomously to get real work done. Pay per action, no subscriptions.",
   "type": "module",
   "exports": {