🚀 NexusRAG

NexusRAG is a comprehensive knowledge base backend system designed for the implementation of Large Language Models (LLMs). It provides ready-to-use modules for document processing and Retrieval-Augmented Generation (RAG), enabling rapid deployment of large-scale knowledge retrieval systems to your Generative AI (GenAI) applications. These applications can include enterprise virtual employees, educational tools, or personalized assistants.

The system allows users to:

Create and manage personal knowledge bases
Automatically upload and process documents in personal knowledge bases (supports multimodal OCR)
Employ multiple search methods across knowledge bases (full-text search, semantic retrieval, and knowledge graph querying, or hybrid search)
Perform fine-grained management of document segments within knowledge bases
Track document processing status in real-time

Components

Below are the components you can use:

| Type | What Supported | Where | |:------------|:------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|:-----------------------:| | Embedding | Sentence-transformers | /src/clients/embeddings | | Rerank | - BCE<br>- BGE Reranker | /src/clients/reranker | | File Parser | MinerU | 💡Pending merge | | Store | milvus (Docker) | | | | elasticsearch (Docker) | | | | neo4j (Docker) | 💡Pending merge | | Chunking | MarkdownTextSplitter | Built-in | | | RecursiveCharacterTextSplitter | Built-in |

This project can integrate with external embedding and reranker APIs. Note that these external APIs must conform to the OpenAI API format standards.
"Pending merge" means "Under final code review"
Always welcome to contribute more components.

📌 Quick start

1. Prepare Environment and Download Corresponding Model Files

apt update
apt install python-dev libxml2-dev libxslt1-dev antiword unrtf poppler-utils pstotext tesseract-ocr flac ffmpeg lame libmad0 libsox-fmt-mp3 sox libjpeg-dev swig libpulse-dev

Create volume directories under the docker directory：

cd docker
mkdir -p volumes/elasticsearch
mkdir -p volumes/etcd
mkdir -p volumes/milvus
mkdir -p volumes/minio

Start Services

docker-compose up -d

This command will download and start the following containers:

elasticsearch: For full-text search and document indexing
milvus: For vector similarity search
minio: For object storage
etcd: For distributed key-value storage
kibana: For elasticsearch visualization

All images will be downloaded to Docker's default image storage location (/var/lib/docker/). Total size ~2GB, may take 5-10 minutes depending on your network speed.

If the following words are displayed, it indicates that the download is complete.

[+] Running 6/6
⠿ Network docker_default       Created
⠿ Container elasticsearch      Started
⠿ Container milvus-etcd        Started
⠿ Container milvus-minio       Started
⠿ Container milvus-standalone  Started
⠿ Container kibana             Started

Check the service operation status

docker-compose ps

When you see output similar to this, all services have been successfully started, elasticsearch runs on port 9200, and milvus runs on port 19530:

NAME                COMMAND                  SERVICE             STATUS              PORTS
elasticsearch       "/bin/tini -- /usr/l…"   elasticsearch       running             0.0.0.0:9200->9200/tcp, 0.0.0.0:9300->9300/tcp, :::9200->9200/tcp, :::9300->9300/tcp
kibana              "/bin/tini -- /usr/l…"   kibana              running             0.0.0.0:5601->5601/tcp, :::5601->5601/tcp
milvus-etcd         "etcd -advertise-cli…"   etcd                running             2379-2380/tcp
milvus-minio        "/usr/bin/docker-ent…"   minio               running (healthy)   9000/tcp
milvus-standalone   "/tini -- milvus run…"   standalone          running             0.0.0.0:9091->9091/tcp, 0.0.0.0:19530->19530/tcp, :::9091->9091/tcp, :::19530->19530/tcp

2. Modify the config.ini file.

[rag]
knowledgebase_log_dir = /path/to/your/knowledgebase_server_log
knowledgebase_bind_port = 1001

embedding_model_path = /path/to/your/text2vec-large-chinese
embedding_bind_port = 5003

reranker_model_path = /path/to/your/bce-reranker-base_v1
reranker_bind_port = 5001

es_url = http://127.0.0.1:9200
index_name = test
milvus_url = 127.0.0.1:19530

3. Launch Service

cd src

Launch Embedding Service

python ./clients/emdeddings/server.py --config_path ../config.ini --gpu_id 0

Launch Reranker Service

python ./clients/reranker/server.py --config_path ../config.ini --gpu_id 0

Launch Knowledge Base Service

python main.py --config_path ../config.ini --gpu_id 0

🔍 API endpoints introduction

🔧 FAQ

If Elasticsearch fails to start successfully, enter the following in the docker directory within the container:

sudo chown -R 1000:0 ${DOCKER_VOLUME_DIRECTORY:-.}/volumes/elasticsearch

NexusRAG

Install / Use

README