@blockrun/clawrouter 0.11.9 → 0.11.11

package/dist/index.d.ts

@@ -770,6 +770,8 @@ type UsageEntry = {
     baselineCost: number;
     savings: number;
     latencyMs: number;
+    /** Input (prompt) tokens reported by the provider */
+    inputTokens?: number;
     /** Partner service ID (e.g., "x_users_lookup") — only set for partner API calls */
     partnerId?: string;
     /** Partner service name (e.g., "AttentionVC") — only set for partner API calls */

package/dist/index.js

@@ -5335,6 +5335,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
   let maxTokens = 4096;
   let routingProfile = null;
   let accumulatedContent = "";
+  let responseInputTokens;
   const isChatCompletion = req.url?.includes("/chat/completions");
   const sessionId = getSessionId(req.headers);
   let effectiveSessionId = sessionId;
@@ -6189,6 +6190,10 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
         const jsonStr = jsonBody.toString();
         try {
           const rsp = JSON.parse(jsonStr);
+          if (rsp.usage && typeof rsp.usage === "object") {
+            const u = rsp.usage;
+            if (typeof u.prompt_tokens === "number") responseInputTokens = u.prompt_tokens;
+          }
           const baseChunk = {
             id: rsp.id ?? `chatcmpl-${Date.now()}`,
             object: "chat.completion.chunk",
@@ -6337,6 +6342,10 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
         if (rspJson.choices?.[0]?.message?.content) {
           accumulatedContent = rspJson.choices[0].message.content;
         }
+        if (rspJson.usage && typeof rspJson.usage === "object") {
+          if (typeof rspJson.usage.prompt_tokens === "number")
+            responseInputTokens = rspJson.usage.prompt_tokens;
+        }
       } catch {
       }
     }
@@ -6385,7 +6394,8 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
       cost: costWithBuffer,
       baselineCost: baselineWithBuffer,
       savings: accurateCosts.savings,
-      latencyMs: Date.now() - startTime
+      latencyMs: Date.now() - startTime,
+      ...responseInputTokens !== void 0 && { inputTokens: responseInputTokens }
     };
     logUsage(entry).catch(() => {
     });
@@ -6789,8 +6799,8 @@ function injectModelsConfig(logger) {
     "openai/gpt-5.2",
     "openai/gpt-4o",
     "openai/o3",
-    "google/gemini-2.5-pro",
-    "google/gemini-2.5-flash",
+    "google/gemini-3.1-pro",
+    "google/gemini-3-flash-preview",
     "deepseek/deepseek-chat",
     "moonshot/kimi-k2.5",
     "xai/grok-3",