Fixed the context window issue.

Putting in a max tokens into the request was a mistake.

apps/simple/vectorIndexCustomize.ts

+import { Document, VectorStoreIndex, RetrieverQueryEngine } from "llamaindex";
+import essay from "./essay";
+
+// Customize retrieval and query args
+async function main() {
+  const document = new Document({ text: essay });
+  const index = await VectorStoreIndex.fromDocuments([document]);
+  const retriever = index.asRetriever();
+  retriever.similarityTopK = 5;
+  // TODO: cannot pass responseSynthesizer into retriever query engine
+  const queryEngine = new RetrieverQueryEngine(retriever);
+  const response = await queryEngine.query(
+    "What did the author do growing up?"
+  );
+  console.log(response.response);
+}
+
+main().catch(console.error);

packages/core/src/LLM.ts

@@ -78,8 +78,7 @@ export class OpenAI implements LLM {
     this.temperature = init?.temperature ?? 0;
     this.requestTimeout = init?.requestTimeout ?? null;
     this.maxRetries = init?.maxRetries ?? 10;
-    this.maxTokens =
-      init?.maxTokens ?? Math.floor(ALL_AVAILABLE_MODELS[this.model] / 2);
+    this.maxTokens = init?.maxTokens ?? undefined;
     this.openAIKey = init?.openAIKey ?? null;
     this.session = init?.session ?? getOpenAISession();
     this.callbackManager = init?.callbackManager;