190proof 1.0.85 → 1.0.87

package/README.md

@@ -245,7 +245,13 @@ interface ParsedResponseMessage {
   role: "assistant";
   content: string | null;
   function_call: FunctionCall | null;
+  function_calls: FunctionCall[];
   files: File[]; // For models that return files (e.g., image generation)
+  usage: {
+    prompt_tokens: number;
+    completion_tokens: number;
+    total_tokens: number;
+  } | null; // null when streaming
 }
 ```
 

package/dist/index.d.mts

@@ -31,6 +31,8 @@ declare enum GPTModel {
 }
 declare enum GroqModel {
     LLAMA_3_70B_8192 = "llama3-70b-8192",
+    LLAMA_3_3_70B_VERSATILE = "llama-3.3-70b-versatile",
+    QWEN3_32B = "qwen/qwen3-32b",
     DEEPSEEK_R1_DISTILL_LLAMA_70B = "deepseek-r1-distill-llama-70b"
 }
 declare enum GeminiModel {
@@ -62,6 +64,11 @@ interface ParsedResponseMessage {
     function_call: FunctionCall | null;
     function_calls: FunctionCall[];
     files: File[];
+    usage: {
+        prompt_tokens: number;
+        completion_tokens: number;
+        total_tokens: number;
+    } | null;
 }
 interface FunctionCall {
     name: string;

package/dist/index.d.ts

@@ -31,6 +31,8 @@ declare enum GPTModel {
 }
 declare enum GroqModel {
     LLAMA_3_70B_8192 = "llama3-70b-8192",
+    LLAMA_3_3_70B_VERSATILE = "llama-3.3-70b-versatile",
+    QWEN3_32B = "qwen/qwen3-32b",
     DEEPSEEK_R1_DISTILL_LLAMA_70B = "deepseek-r1-distill-llama-70b"
 }
 declare enum GeminiModel {
@@ -62,6 +64,11 @@ interface ParsedResponseMessage {
     function_call: FunctionCall | null;
     function_calls: FunctionCall[];
     files: File[];
+    usage: {
+        prompt_tokens: number;
+        completion_tokens: number;
+        total_tokens: number;
+    } | null;
 }
 interface FunctionCall {
     name: string;

package/dist/index.js

@@ -74,6 +74,8 @@ var GPTModel = /* @__PURE__ */ ((GPTModel2) => {
 })(GPTModel || {});
 var GroqModel = /* @__PURE__ */ ((GroqModel2) => {
   GroqModel2["LLAMA_3_70B_8192"] = "llama3-70b-8192";
+  GroqModel2["LLAMA_3_3_70B_VERSATILE"] = "llama-3.3-70b-versatile";
+  GroqModel2["QWEN3_32B"] = "qwen/qwen3-32b";
   GroqModel2["DEEPSEEK_R1_DISTILL_LLAMA_70B"] = "deepseek-r1-distill-llama-70b";
   return GroqModel2;
 })(GroqModel || {});
@@ -200,7 +202,8 @@ function parseStreamedResponse(identifier, paragraph, toolCallAccumulators, allo
     content: paragraph || null,
     function_call: functionCalls[0] || null,
     function_calls: functionCalls,
-    files: []
+    files: [],
+    usage: null
   };
 }
 function truncatePayload(payload) {
@@ -372,7 +375,10 @@ async function callOpenAIStream(id, openAiPayload, openAiConfig, chunkTimeoutMs)
     const { done, value } = await reader.read();
     clearTimeout(abortTimeout);
     if (done) {
-      logger_default.error(id, `Stream ended prematurely after ${chunkIndex + 1} chunks`);
+      logger_default.error(
+        id,
+        `Stream ended prematurely after ${chunkIndex + 1} chunks`
+      );
       throw new Error("Stream error: ended prematurely");
     }
     let chunk = new TextDecoder().decode(value);
@@ -477,7 +483,12 @@ async function callOpenAI(id, openAiPayload, openAiConfig) {
     content: choice.message.content || null,
     function_call: functionCalls[0] || null,
     function_calls: functionCalls,
-    files: []
+    files: [],
+    usage: data.usage ? {
+      prompt_tokens: data.usage.prompt_tokens,
+      completion_tokens: data.usage.completion_tokens,
+      total_tokens: data.usage.total_tokens
+    } : null
   };
 }
 async function callOpenAiWithRetries(id, openAiPayload, openAiConfig, retries = 5, chunkTimeoutMs = 15e3) {
@@ -493,7 +504,12 @@ async function callOpenAiWithRetries(id, openAiPayload, openAiConfig, retries =
     "OpenAI",
     async () => {
       if (useStreaming) {
-        return callOpenAIStream(id, openAiPayload, openAiConfig, chunkTimeoutMs);
+        return callOpenAIStream(
+          id,
+          openAiPayload,
+          openAiConfig,
+          chunkTimeoutMs
+        );
       } else {
         return callOpenAI(id, openAiPayload, openAiConfig);
       }
@@ -526,7 +542,10 @@ function jigAnthropicMessages(messages) {
   var _a, _b;
   let jiggedMessages = messages.slice();
   if (((_a = jiggedMessages[0]) == null ? void 0 : _a.role) !== "user") {
-    jiggedMessages = [{ role: "user", content: "..." }, ...jiggedMessages];
+    jiggedMessages = [
+      { role: "user", content: "..." },
+      ...jiggedMessages
+    ];
   }
   jiggedMessages = jiggedMessages.reduce((acc, message) => {
     if (acc.length === 0)
@@ -571,24 +590,28 @@ async function prepareAnthropicPayload(_identifier, payload) {
     for (const file of message.files || []) {
       if (ALLOWED_IMAGE_MIME_TYPES.includes(file.mimeType)) {
         if (file.url) {
-          contentBlocks.push({
-            type: "image",
-            source: {
-              type: "base64",
-              media_type: "image/png",
-              data: await getNormalizedBase64PNG(file.url, file.mimeType)
-            }
-          });
+          if (message.role == "user") {
+            contentBlocks.push({
+              type: "image",
+              source: {
+                type: "base64",
+                media_type: "image/png",
+                data: await getNormalizedBase64PNG(file.url, file.mimeType)
+              }
+            });
+          }
           contentBlocks.push({ type: "text", text: `Image (${file.url})` });
         } else if (file.data) {
-          contentBlocks.push({
-            type: "image",
-            source: {
-              type: "base64",
-              media_type: file.mimeType,
-              data: file.data
-            }
-          });
+          if (message.role == "user") {
+            contentBlocks.push({
+              type: "image",
+              source: {
+                type: "base64",
+                media_type: file.mimeType,
+                data: file.data
+              }
+            });
+          }
         }
       } else if (file.url) {
         contentBlocks.push({
@@ -687,7 +710,11 @@ ${text}` : text;
     }
   }
   if (!textResponse && !functionCalls.length) {
-    logger_default.error(id, "Missing text & functions in Anthropic API response:", data);
+    logger_default.error(
+      id,
+      "Missing text & functions in Anthropic API response:",
+      data
+    );
     throw new Error("Missing text & functions in Anthropic API response");
   }
   return {
@@ -695,19 +722,32 @@ ${text}` : text;
     content: textResponse,
     function_call: functionCalls[0] || null,
     function_calls: functionCalls,
-    files: []
+    files: [],
+    usage: data.usage ? {
+      prompt_tokens: data.usage.input_tokens,
+      completion_tokens: data.usage.output_tokens,
+      total_tokens: data.usage.input_tokens + data.usage.output_tokens
+    } : null
   };
 }
 async function callAnthropicWithRetries(id, payload, config, retries = 5) {
-  return withRetries(id, "Anthropic", () => callAnthropic(id, payload, config), {
-    retries
-  });
+  return withRetries(
+    id,
+    "Anthropic",
+    () => callAnthropic(id, payload, config),
+    {
+      retries
+    }
+  );
 }
 function jigGoogleMessages(messages) {
   var _a, _b;
   let jiggedMessages = messages.slice();
   if (((_a = jiggedMessages[0]) == null ? void 0 : _a.role) === "model") {
-    jiggedMessages = [{ role: "user", parts: [{ text: "..." }] }, ...jiggedMessages];
+    jiggedMessages = [
+      { role: "user", parts: [{ text: "..." }] },
+      ...jiggedMessages
+    ];
   }
   jiggedMessages = jiggedMessages.reduce((acc, message) => {
     if (acc.length === 0)
@@ -779,7 +819,7 @@ async function prepareGoogleAIPayload(_identifier, payload) {
   return preparedPayload;
 }
 async function callGoogleAI(id, payload) {
-  var _a, _b, _c, _d, _e, _f;
+  var _a, _b, _c, _d, _e, _f, _g, _h, _i;
   const googleMessages = jigGoogleMessages(payload.messages);
   const history = googleMessages.slice(0, -1);
   const lastMessage = googleMessages.slice(-1)[0];
@@ -843,13 +883,15 @@ async function callGoogleAI(id, payload) {
     content: text || null,
     files,
     function_call: (functionCalls == null ? void 0 : functionCalls[0]) || null,
-    function_calls: functionCalls || []
+    function_calls: functionCalls || [],
+    usage: response.usageMetadata ? {
+      prompt_tokens: (_g = response.usageMetadata.promptTokenCount) != null ? _g : 0,
+      completion_tokens: (_h = response.usageMetadata.candidatesTokenCount) != null ? _h : 0,
+      total_tokens: (_i = response.usageMetadata.totalTokenCount) != null ? _i : 0
+    } : null
   };
 }
-var CONTENT_VIOLATION_REASONS = /* @__PURE__ */ new Set([
-  "PROHIBITED_CONTENT",
-  "SAFETY"
-]);
+var CONTENT_VIOLATION_REASONS = /* @__PURE__ */ new Set(["PROHIBITED_CONTENT", "SAFETY"]);
 function removeImagesFromGooglePayload(payload) {
   let removedImages = false;
   for (const message of payload.messages) {
@@ -889,7 +931,14 @@ async function callGoogleAIWithRetries(id, payload, retries = 5) {
         errorDetails.errorCode = error2.code;
       if (error2.details)
         errorDetails.errorDetails = error2.details;
-      logger_default.error(id, `Retry #${attempt} error: ${error2.message}`, errorDetails);
+      const fileUris = payload.messages.flatMap((m) => m.parts).filter((p) => "fileData" in p).map((p) => p.fileData.fileUri);
+      if (fileUris.length)
+        errorDetails.fileUris = fileUris;
+      logger_default.error(
+        id,
+        `Retry #${attempt} error: ${error2.message}`,
+        errorDetails
+      );
       const violationReason = CONTENT_VIOLATION_REASONS.has(error2.finishReason) && error2.finishReason || CONTENT_VIOLATION_REASONS.has((_a = error2.promptFeedback) == null ? void 0 : _a.blockReason) && ((_b = error2.promptFeedback) == null ? void 0 : _b.blockReason);
       if (violationReason) {
         if (!hasTriedWithoutImages) {
@@ -969,7 +1018,12 @@ async function callGroq(id, payload) {
     content: answer.content || null,
     function_call: functionCalls[0] || null,
     function_calls: functionCalls,
-    files: []
+    files: [],
+    usage: response.data.usage ? {
+      prompt_tokens: response.data.usage.prompt_tokens,
+      completion_tokens: response.data.usage.completion_tokens,
+      total_tokens: response.data.usage.total_tokens
+    } : null
   };
 }
 async function callGroqWithRetries(id, payload, retries = 5) {
@@ -1007,7 +1061,11 @@ async function callWithRetries(id, aiPayload, aiConfig, retries = 5, chunkTimeou
       );
     }
     if (isGroqPayload(aiPayload)) {
-      return await callGroqWithRetries(id, prepareGroqPayload(aiPayload), retries);
+      return await callGroqWithRetries(
+        id,
+        prepareGroqPayload(aiPayload),
+        retries
+      );
     }
     if (isGoogleAIPayload(aiPayload)) {
       return await callGoogleAIWithRetries(
@@ -1019,13 +1077,21 @@ async function callWithRetries(id, aiPayload, aiConfig, retries = 5, chunkTimeou
     throw new Error("Invalid AI payload: Unknown model type.");
   } catch (error2) {
     if (aiPayload.fallbackModel) {
-      logger_default.log(
+      logger_default.error(
         id,
-        `Primary model ${aiPayload.model} failed, falling back to ${aiPayload.fallbackModel}`
+        `Primary model ${aiPayload.model} failed, falling back to ${aiPayload.fallbackModel}`,
+        {
+          error: error2 instanceof Error ? error2.message : error2,
+          cause: error2 instanceof Error && error2.cause instanceof Error ? error2.cause.message : void 0
+        }
       );
       return callWithRetries(
         id,
-        { ...aiPayload, model: aiPayload.fallbackModel, fallbackModel: void 0 },
+        {
+          ...aiPayload,
+          model: aiPayload.fallbackModel,
+          fallbackModel: void 0
+        },
         aiConfig,
         retries,
         chunkTimeoutMs