Skip to main content
Version: v0.7.0

BM25 RAG

In this example, we will show how to use the Elasticsearch as in DB-GPT RAG Storage. Using a Elasticsearch database to implement RAG can, to some extent, alleviate the uncertainty and interpretability issues brought about by Elasticsearch database retrieval.

Install Dependencies​

First, you need to install the dbgpt elasticsearch storage library.

uv sync --all-packages --frozen \
--extra "base" \
--extra "proxy_openai" \
--extra "rag" \
--extra "storage_elasticsearch" \
--extra "dbgpts"

Prepare Elasticsearch​

Prepare Elasticsearch database service, reference-Elasticsearch Installation .

Elasticsearch Configuration​

Set rag storage variables below in configs/dbgpt-bm25-rag.toml file, let DB-GPT know how to connect to Elasticsearch.


[rag.storage]
[rag.storage.full_text]
type = "ElasticSearch"
uri = "127.0.0.1"
port = "9200"

Then run the following command to start the webserver:

uv run python packages/dbgpt-app/src/dbgpt_app/dbgpt_server.py --config configs/dbgpt-bm25-rag.toml

Optionally

uv run python packages/dbgpt-app/src/dbgpt_app/dbgpt_server.py --config configs/dbgpt-bm25-rag.toml