From 120cac391e38fe112d7ce0011c111668b04bfc9d Mon Sep 17 00:00:00 2001 From: hugy718 Date: Thu, 2 Jan 2025 06:55:33 +0000 Subject: [PATCH] Update hakes-worker README --- docker/hakes-worker/no-sgx/README.md | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) diff --git a/docker/hakes-worker/no-sgx/README.md b/docker/hakes-worker/no-sgx/README.md index ba7a6d4..0895bc9 100644 --- a/docker/hakes-worker/no-sgx/README.md +++ b/docker/hakes-worker/no-sgx/README.md @@ -1,4 +1,4 @@ -# HAKES Worker (SGX) +# HAKES Worker ## build @@ -6,6 +6,25 @@ docker build -t hakes-hakesworker-nosgx:v1 -f docker/hakes-worker/no-sgx/Dockerfile . ``` +## prepare the config file + +The config file mounted for hakes-worker container controls how it interacts with other components in HAKES to serve user requests. A sample config is show below. + +```json +{ + "embed_endpoint_type": "hakes-embed", + "embed_endpoint_config": "", + "embed_endpoint_addr": "10.10.10.28:2700", + "search_worker_addrs": ["10.10.10.57:2800"], + "preferred_search_worker": 0, + "store_addr": "10.10.10.26:2600" +} +``` + +* `embed_endpoint_type`, `embed_endpoint_config` and `embed_endpoint_addr` dictate the embed service that converts raw data in user requests to embedding vectors. When the `embed_endpoint_type` is `hakes-embed`, it uses an embed-worker server or function. We also support external embedding services through the [embed-endpoint](../../../embed-endpoint/) module, such as `openai` and `huggingface` for open-ai embedding APIs and huggingface inrference endpoints. +* `search_worker_addrs` points to the search-workers that maintains the vector index and `preferred_search_worker` indicates the request routing to a specific search-worker that the hakes-worker will contact. +* `store_addr` points to the hakes-store service that persists the data. + ## run ```sh