From 2619d4175da4425fa45a70ccf2081ca8045f4103 Mon Sep 17 00:00:00 2001 From: Yi Ding <yi.s.ding@gmail.com> Date: Tue, 18 Jul 2023 23:10:14 -0700 Subject: [PATCH] Fixed the context window issue. Putting in a max tokens into the request was a mistake. --- apps/simple/vectorIndexCustomize.ts | 18 ++++++++++++++++++ packages/core/src/LLM.ts | 3 +-- 2 files changed, 19 insertions(+), 2 deletions(-) create mode 100644 apps/simple/vectorIndexCustomize.ts diff --git a/apps/simple/vectorIndexCustomize.ts b/apps/simple/vectorIndexCustomize.ts new file mode 100644 index 000000000..49d5ca4d8 --- /dev/null +++ b/apps/simple/vectorIndexCustomize.ts @@ -0,0 +1,18 @@ +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); diff --git a/packages/core/src/LLM.ts b/packages/core/src/LLM.ts index 20b706610..0913d64fc 100644 --- a/packages/core/src/LLM.ts +++ b/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; -- GitLab