@jambonz/llm 0.2.2 → 0.2.4

package/dist/index.cjs

@@ -117,14 +117,15 @@ async function* streamFromOpenAI(client, req, options) {
     model: req.model,
     messages,
     stream: true,
-    stream_options: { include_usage: true }
+    ...options.includeStreamOptions !== false ? { stream_options: { include_usage: true } } : {}
   };
   if (tools) {
     body.tools = tools;
     body.tool_choice = "auto";
   }
   if (req.temperature !== void 0) body.temperature = req.temperature;
-  if (req.maxTokens !== void 0) body.max_tokens = req.maxTokens;
+  const effectiveMaxTokens = req.maxTokens ?? options.defaultMaxTokens;
+  if (effectiveMaxTokens !== void 0) body.max_tokens = effectiveMaxTokens;
   let stream;
   try {
     stream = await client.chat.completions.create(body, { signal: req.signal });
@@ -1609,7 +1610,7 @@ var VertexOpenAIAdapter = class {
     if (!auth.location) {
       throw new Error("VertexOpenAIAdapter: location is required");
     }
-    const baseURL = `https://${auth.location}-aiplatform.googleapis.com/v1/projects/${auth.projectId}/locations/${auth.location}/endpoints/openapi`;
+    const baseURL = `https://${auth.location}-aiplatform.googleapis.com/v1beta1/projects/${auth.projectId}/locations/${auth.location}/endpoints/openapi`;
     const googleAuth = new import_google_auth_library.GoogleAuth({
       credentials: auth.credentials,
       scopes: ["https://www.googleapis.com/auth/cloud-platform"]
@@ -1631,7 +1632,8 @@ var VertexOpenAIAdapter = class {
   }
   stream(req) {
     return streamFromOpenAI(this.ensureClient(), req, {
-      knownModels: vertexOpenAIManifest.knownModels
+      knownModels: vertexOpenAIManifest.knownModels,
+      defaultMaxTokens: 4096
     });
   }
   appendAssistantToolCall(history, toolCalls) {