From 37eec3d452b6eba904a50f327223e28f5e5ebfdd Mon Sep 17 00:00:00 2001 From: Finley Ge <32237950+FinleyGe@users.noreply.github.com> Date: Wed, 20 Aug 2025 18:42:15 +0800 Subject: [PATCH] perf: customizable embedding chunk size via env var (#5494) * perf: customizable embedding chunk size via env var * Update .env.template --------- Co-authored-by: Archer <545436317@qq.com> --- packages/service/core/ai/embedding/index.ts | 2 +- projects/app/.env.template | 4 ++++ 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/packages/service/core/ai/embedding/index.ts b/packages/service/core/ai/embedding/index.ts index ba31eaf29..f0ddd2325 100644 --- a/packages/service/core/ai/embedding/index.ts +++ b/packages/service/core/ai/embedding/index.ts @@ -24,7 +24,7 @@ export async function getVectorsByText({ model, input, type, headers }: GetVecto const formatInput = Array.isArray(input) ? input : [input]; // 20 size every request - const chunkSize = 20; + const chunkSize = parseInt(process.env.EMBEDDING_CHUNK_SIZE || '10'); const chunks = []; for (let i = 0; i < formatInput.length; i += chunkSize) { chunks.push(formatInput.slice(i, i + chunkSize)); diff --git a/projects/app/.env.template b/projects/app/.env.template index 186d06512..79aef450f 100644 --- a/projects/app/.env.template +++ b/projects/app/.env.template @@ -13,6 +13,9 @@ ROOT_KEY=fdafasd # 强制将图片转成 base64 传递给模型 MULTIPLE_DATA_TO_BASE64=true +# 临时解决-emb 模型单次并发量 +EMBEDDING_CHUNK_SIZE=10 + # 是否隐藏版权信息配置,只有值为 'true' 时隐藏 HIDE_CHAT_COPYRIGHT_SETTING= @@ -96,3 +99,4 @@ CONFIG_JSON_PATH= SIGNOZ_BASE_URL= SIGNOZ_SERVICE_NAME= SIGNOZ_STORE_LEVEL=warn +